From 38dd7fba1632139ca997b4854474f37f69c92300 Mon Sep 17 00:00:00 2001 From: Paul Rashidi Date: Mon, 11 Jul 2016 18:07:34 -0700 Subject: [PATCH] Initial source and xcode build. --- .gitignore | 32 + AUTHORS | 14 +- CMakeLists.txt | 29 + CONTRIBUTING.md | 94 +- EtcLib/CMakeLists.txt | 24 + EtcLib/Etc/Etc.cpp | 58 + EtcLib/Etc/Etc.h | 43 + EtcLib/Etc/EtcColor.h | 64 + EtcLib/Etc/EtcColorFloatRGBA.h | 321 + EtcLib/Etc/EtcConfig.h | 67 + EtcLib/Etc/EtcImage.cpp | 685 ++ EtcLib/Etc/EtcImage.h | 249 + EtcLib/Etc/EtcMath.cpp | 64 + EtcLib/Etc/EtcMath.h | 40 + EtcLib/EtcCodec/EtcBlock4x4.cpp | 417 ++ EtcLib/EtcCodec/EtcBlock4x4.h | 172 + EtcLib/EtcCodec/EtcBlock4x4Encoding.cpp | 246 + EtcLib/EtcCodec/EtcBlock4x4Encoding.h | 148 + EtcLib/EtcCodec/EtcBlock4x4EncodingBits.h | 315 + EtcLib/EtcCodec/EtcBlock4x4Encoding_ETC1.cpp | 1280 ++++ EtcLib/EtcCodec/EtcBlock4x4Encoding_ETC1.h | 186 + EtcLib/EtcCodec/EtcBlock4x4Encoding_R11.cpp | 429 ++ EtcLib/EtcCodec/EtcBlock4x4Encoding_R11.h | 122 + EtcLib/EtcCodec/EtcBlock4x4Encoding_RG11.cpp | 447 ++ EtcLib/EtcCodec/EtcBlock4x4Encoding_RG11.h | 86 + EtcLib/EtcCodec/EtcBlock4x4Encoding_RGB8.cpp | 1727 +++++ EtcLib/EtcCodec/EtcBlock4x4Encoding_RGB8.h | 96 + .../EtcCodec/EtcBlock4x4Encoding_RGB8A1.cpp | 1819 +++++ EtcLib/EtcCodec/EtcBlock4x4Encoding_RGB8A1.h | 129 + EtcLib/EtcCodec/EtcBlock4x4Encoding_RGBA8.cpp | 474 ++ EtcLib/EtcCodec/EtcBlock4x4Encoding_RGBA8.h | 121 + EtcLib/EtcCodec/EtcDifferentialTrys.cpp | 173 + EtcLib/EtcCodec/EtcDifferentialTrys.h | 97 + EtcLib/EtcCodec/EtcErrorMetric.h | 51 + EtcLib/EtcCodec/EtcIndividualTrys.cpp | 85 + EtcLib/EtcCodec/EtcIndividualTrys.h | 95 + EtcLib/EtcCodec/EtcSortedBlockList.cpp | 228 + EtcLib/EtcCodec/EtcSortedBlockList.h | 124 + EtcTool/Args.txt | 7 + EtcTool/CMakeLists.txt | 28 + EtcTool/EtcAnalysis.cpp | 410 ++ EtcTool/EtcAnalysis.h | 71 + EtcTool/EtcComparison.cpp | 637 ++ EtcTool/EtcComparison.h | 61 + EtcTool/EtcFile.cpp | 311 + EtcTool/EtcFile.h | 99 + EtcTool/EtcFileHeader.cpp | 192 + EtcTool/EtcFileHeader.h | 146 + EtcTool/EtcMemTest.cpp | 258 + EtcTool/EtcMemTest.h | 61 + EtcTool/EtcSourceImage.cpp | 292 + EtcTool/EtcSourceImage.h | 93 + EtcTool/EtcTool.cpp | 719 ++ EtcTool/EtcTool.h | 36 + LICENSE | 404 +- README.md | 189 +- third_party/lodepng/LICENSE | 25 + third_party/lodepng/lodepng.cpp | 6168 +++++++++++++++++ third_party/lodepng/lodepng.h | 1756 +++++ 59 files changed, 22549 insertions(+), 265 deletions(-) create mode 100755 .gitignore create mode 100644 CMakeLists.txt create mode 100644 EtcLib/CMakeLists.txt create mode 100644 EtcLib/Etc/Etc.cpp create mode 100644 EtcLib/Etc/Etc.h create mode 100644 EtcLib/Etc/EtcColor.h create mode 100644 EtcLib/Etc/EtcColorFloatRGBA.h create mode 100644 EtcLib/Etc/EtcConfig.h create mode 100644 EtcLib/Etc/EtcImage.cpp create mode 100644 EtcLib/Etc/EtcImage.h create mode 100644 EtcLib/Etc/EtcMath.cpp create mode 100644 EtcLib/Etc/EtcMath.h create mode 100644 EtcLib/EtcCodec/EtcBlock4x4.cpp create mode 100644 EtcLib/EtcCodec/EtcBlock4x4.h create mode 100644 EtcLib/EtcCodec/EtcBlock4x4Encoding.cpp create mode 100644 EtcLib/EtcCodec/EtcBlock4x4Encoding.h create mode 100644 EtcLib/EtcCodec/EtcBlock4x4EncodingBits.h create mode 100644 EtcLib/EtcCodec/EtcBlock4x4Encoding_ETC1.cpp create mode 100644 EtcLib/EtcCodec/EtcBlock4x4Encoding_ETC1.h create mode 100644 EtcLib/EtcCodec/EtcBlock4x4Encoding_R11.cpp create mode 100644 EtcLib/EtcCodec/EtcBlock4x4Encoding_R11.h create mode 100644 EtcLib/EtcCodec/EtcBlock4x4Encoding_RG11.cpp create mode 100644 EtcLib/EtcCodec/EtcBlock4x4Encoding_RG11.h create mode 100644 EtcLib/EtcCodec/EtcBlock4x4Encoding_RGB8.cpp create mode 100644 EtcLib/EtcCodec/EtcBlock4x4Encoding_RGB8.h create mode 100644 EtcLib/EtcCodec/EtcBlock4x4Encoding_RGB8A1.cpp create mode 100644 EtcLib/EtcCodec/EtcBlock4x4Encoding_RGB8A1.h create mode 100644 EtcLib/EtcCodec/EtcBlock4x4Encoding_RGBA8.cpp create mode 100644 EtcLib/EtcCodec/EtcBlock4x4Encoding_RGBA8.h create mode 100644 EtcLib/EtcCodec/EtcDifferentialTrys.cpp create mode 100644 EtcLib/EtcCodec/EtcDifferentialTrys.h create mode 100644 EtcLib/EtcCodec/EtcErrorMetric.h create mode 100644 EtcLib/EtcCodec/EtcIndividualTrys.cpp create mode 100644 EtcLib/EtcCodec/EtcIndividualTrys.h create mode 100644 EtcLib/EtcCodec/EtcSortedBlockList.cpp create mode 100644 EtcLib/EtcCodec/EtcSortedBlockList.h create mode 100644 EtcTool/Args.txt create mode 100644 EtcTool/CMakeLists.txt create mode 100644 EtcTool/EtcAnalysis.cpp create mode 100644 EtcTool/EtcAnalysis.h create mode 100644 EtcTool/EtcComparison.cpp create mode 100644 EtcTool/EtcComparison.h create mode 100644 EtcTool/EtcFile.cpp create mode 100644 EtcTool/EtcFile.h create mode 100644 EtcTool/EtcFileHeader.cpp create mode 100644 EtcTool/EtcFileHeader.h create mode 100644 EtcTool/EtcMemTest.cpp create mode 100644 EtcTool/EtcMemTest.h create mode 100644 EtcTool/EtcSourceImage.cpp create mode 100644 EtcTool/EtcSourceImage.h create mode 100644 EtcTool/EtcTool.cpp create mode 100644 EtcTool/EtcTool.h create mode 100644 third_party/lodepng/LICENSE create mode 100644 third_party/lodepng/lodepng.cpp create mode 100644 third_party/lodepng/lodepng.h diff --git a/.gitignore b/.gitignore new file mode 100755 index 0000000..4100db6 --- /dev/null +++ b/.gitignore @@ -0,0 +1,32 @@ +# Copyright 2015 Etc2Comp Authors. +# +# Licensed under the Apache License, Version 2.0 (the "License"); +# you may not use this file except in compliance with the License. +# You may obtain a copy of the License at +# +# http://www.apache.org/licenses/LICENSE-2.0 +# +# Unless required by applicable law or agreed to in writing, software +# distributed under the License is distributed on an "AS IS" BASIS, +# WITHOUT WARRANTIES OR CONDITIONS OF ANY KIND, either express or implied. +# See the License for the specific language governing permissions and +# limitations under the License. +x64/ +Release/ +Debug/ +ipch/ +*.sdf +.vs/ +*Debug/ +*Release/ +*debug*/ +*release*/ +*.db +Analysis/ +CMakeFiles/ +*.vcxproj* +*.sln +*.cmake +CmakeCache.txt +build_*/* +!build_*/gen_xcode.sh diff --git a/AUTHORS b/AUTHORS index e78a7f4..32daca2 100644 --- a/AUTHORS +++ b/AUTHORS @@ -1,7 +1,7 @@ -# This is the list of Etc2Comp authors for copyright purposes. -# -# This does not necessarily list everyone who has contributed code, since in -# some cases, their employer may be the copyright holder. To see the full list -# of contributors, see the revision history in source control. -Google Inc. -Blue Shift Inc. +# This is the list of Etc2Comp authors for copyright purposes. +# +# This does not necessarily list everyone who has contributed code, since in +# some cases, their employer may be the copyright holder. To see the full list +# of contributors, see the revision history in source control. +Google Inc. +Blue Shift Inc. diff --git a/CMakeLists.txt b/CMakeLists.txt new file mode 100644 index 0000000..6a6392b --- /dev/null +++ b/CMakeLists.txt @@ -0,0 +1,29 @@ +# Copyright 2015 Etc2Comp Authors. +# +# Licensed under the Apache License, Version 2.0 (the "License"); +# you may not use this file except in compliance with the License. +# You may obtain a copy of the License at +# +# http://www.apache.org/licenses/LICENSE-2.0 +# +# Unless required by applicable law or agreed to in writing, software +# distributed under the License is distributed on an "AS IS" BASIS, +# WITHOUT WARRANTIES OR CONDITIONS OF ANY KIND, either express or implied. +# See the License for the specific language governing permissions and +# limitations under the License. + +cmake_minimum_required(VERSION 2.8.9) +project(EtcTest) + +set (CMAKE_CXX_STANDARD 11) +IF (APPLE) + set (CMAKE_CXX_FLAGS "-I/usr/include/i386-linux-gnu/c++/4.8 -I/usr/include/c++/4.8 -std=c++11 -g3 -Wall -O3") +ELSE () + IF (WIN32) + set (CMAKE_CXX_FLAGS "-I/usr/include/i386-linux-gnu/c++/4.8 -I/usr/include/c++/4.8 -W4 /EHsc") + ELSE() + set (CMAKE_CXX_FLAGS "-I/usr/include/i386-linux-gnu/c++/4.8 -I/usr/include/c++/4.8 -std=c++11 -pthread -g3 -Wall -O2") + ENDIF() +ENDIF () +ADD_SUBDIRECTORY(EtcLib) +ADD_SUBDIRECTORY(EtcTool) diff --git a/CONTRIBUTING.md b/CONTRIBUTING.md index 273adea..02979b1 100644 --- a/CONTRIBUTING.md +++ b/CONTRIBUTING.md @@ -1,48 +1,48 @@ -# Contributing | Etc2Comp - -Thank you for contributing to the Etc2Comp community! - - - [Have a usage question?](#question) - - [Think you found a bug?](#issue) - - [Have a feature request?](#feature) - - [Want to submit a pull request?](#submit) - - [Small print](#smallprint) - -## Have a usage question? - - - Review the README.md to make sure you're building the binary correctly. - - Execute the binary with -h to show the usage help. - - Search through [old issues](https://github.com/google/etc2comp/issues) - for an answer to your question. - - If you still haven't found an answer to your question, [open a new issue](https://github.com/google/etc2comp/issues/new). -Please use the provided bug report template and include a minimal repro. - -## Think you found a bug? - -The library is experimental so that's highly likely. Follow the same -procedure above for questions. If you are up to the challenge, -[submit a Pull Request](#submit) with a fix! - -## Have a feature request? - -Great! Make sure the feature request isn't already listed in -[existing issues](https://github.com/google/etc2comp/issues), -then go ahead and [open a new issue](https://github.com/google/etc2comp/issues/new). -Remove the default template information and specify what you are requesting -technically, as well as, specifying what use cases it supports. - -## Want to submit a pull request? - -Sweet, we'd love to accept your contribution! [Open a new pull request](https://github.com/google/etc2comp/compare). - -If you want to implement a new feature, please open an issue with a -proposal first to discuss the change. - -You will need to sign our [Contributor License Agreement](https://cla.developers.google.com/about/google-individual) -before we can accept your pull request. - -## The small print -Contributions made by corporations are covered by a different -agreement than the one above, the -[Software Grant and Corporate Contributor License Agreement] +# Contributing | Etc2Comp + +Thank you for contributing to the Etc2Comp community! + + - [Have a usage question?](#question) + - [Think you found a bug?](#issue) + - [Have a feature request?](#feature) + - [Want to submit a pull request?](#submit) + - [Small print](#smallprint) + +## Have a usage question? + + - Review the README.md to make sure you're building the binary correctly. + - Execute the binary with -h to show the usage help. + - Search through [old issues](https://github.com/google/etc2comp/issues) + for an answer to your question. + - If you still haven't found an answer to your question, [open a new issue](https://github.com/google/etc2comp/issues/new). +Please use the provided bug report template and include a minimal repro. + +## Think you found a bug? + +The library is experimental so that's highly likely. Follow the same +procedure above for questions. If you are up to the challenge, +[submit a Pull Request](#submit) with a fix! + +## Have a feature request? + +Great! Make sure the feature request isn't already listed in +[existing issues](https://github.com/google/etc2comp/issues), +then go ahead and [open a new issue](https://github.com/google/etc2comp/issues/new). +Remove the default template information and specify what you are requesting +technically, as well as, specifying what use cases it supports. + +## Want to submit a pull request? + +Sweet, we'd love to accept your contribution! [Open a new pull request](https://github.com/google/etc2comp/compare). + +If you want to implement a new feature, please open an issue with a +proposal first to discuss the change. + +You will need to sign our [Contributor License Agreement](https://cla.developers.google.com/about/google-individual) +before we can accept your pull request. + +## The small print +Contributions made by corporations are covered by a different +agreement than the one above, the +[Software Grant and Corporate Contributor License Agreement] (https://cla.developers.google.com/about/google-corporate). \ No newline at end of file diff --git a/EtcLib/CMakeLists.txt b/EtcLib/CMakeLists.txt new file mode 100644 index 0000000..b584b88 --- /dev/null +++ b/EtcLib/CMakeLists.txt @@ -0,0 +1,24 @@ +# Copyright 2015 The Etc2Comp Authors. +# +# Licensed under the Apache License, Version 2.0 (the "License"); +# you may not use this file except in compliance with the License. +# You may obtain a copy of the License at +# +# http://www.apache.org/licenses/LICENSE-2.0 +# +# Unless required by applicable law or agreed to in writing, software +# distributed under the License is distributed on an "AS IS" BASIS, +# WITHOUT WARRANTIES OR CONDITIONS OF ANY KIND, either express or implied. +# See the License for the specific language governing permissions and +# limitations under the License. + +project(EtcLib) +include_directories(./Etc) +include_directories(./EtcCodec) + +file(GLOB SOURCES + ${PROJECT_SOURCE_DIR}/Etc/*.h + ${PROJECT_SOURCE_DIR}/EtcCodec/*.h + ${PROJECT_SOURCE_DIR}/Etc/*.cpp + ${PROJECT_SOURCE_DIR}/EtcCodec/*.cpp) +ADD_LIBRARY(EtcLib ${SOURCES}) diff --git a/EtcLib/Etc/Etc.cpp b/EtcLib/Etc/Etc.cpp new file mode 100644 index 0000000..87e1d9b --- /dev/null +++ b/EtcLib/Etc/Etc.cpp @@ -0,0 +1,58 @@ +/* + * Copyright 2015 The Etc2Comp Authors. + * + * Licensed under the Apache License, Version 2.0 (the "License"); + * you may not use this file except in compliance with the License. + * You may obtain a copy of the License at + * + * http://www.apache.org/licenses/LICENSE-2.0 + * + * Unless required by applicable law or agreed to in writing, software + * distributed under the License is distributed on an "AS IS" BASIS, + * WITHOUT WARRANTIES OR CONDITIONS OF ANY KIND, either express or implied. + * See the License for the specific language governing permissions and + * limitations under the License. + */ + +#include "EtcConfig.h" +#include "Etc.h" + +#include + +namespace Etc +{ + // ---------------------------------------------------------------------------------------------------- + // C-style inteface to the encoder + // + void Encode(float *a_pafSourceRGBA, + unsigned int a_uiSourceWidth, + unsigned int a_uiSourceHeight, + Image::Format a_format, + ErrorMetric a_eErrMetric, + float a_fEffort, + unsigned int a_uiJobs, + unsigned int a_uiMaxJobs, + unsigned char **a_ppaucEncodingBits, + unsigned int *a_puiEncodingBitsBytes, + unsigned int *a_puiExtendedWidth, + unsigned int *a_puiExtendedHeight, + int *a_piEncodingTime_ms, bool a_bVerboseOutput) + { + + Image image(a_pafSourceRGBA, a_uiSourceWidth, + a_uiSourceHeight, + a_eErrMetric); + image.m_bVerboseOutput = a_bVerboseOutput; + image.Encode(a_format, a_eErrMetric, a_fEffort, a_uiJobs, a_uiMaxJobs); + + *a_ppaucEncodingBits = image.GetEncodingBits(); + *a_puiEncodingBitsBytes = image.GetEncodingBitsBytes(); + *a_puiExtendedWidth = image.GetExtendedWidth(); + *a_puiExtendedHeight = image.GetExtendedHeight(); + *a_piEncodingTime_ms = image.GetEncodingTimeMs(); + } + + // ---------------------------------------------------------------------------------------------------- + // + +} diff --git a/EtcLib/Etc/Etc.h b/EtcLib/Etc/Etc.h new file mode 100644 index 0000000..6fa6d5c --- /dev/null +++ b/EtcLib/Etc/Etc.h @@ -0,0 +1,43 @@ +/* + * Copyright 2015 The Etc2Comp Authors. + * + * Licensed under the Apache License, Version 2.0 (the "License"); + * you may not use this file except in compliance with the License. + * You may obtain a copy of the License at + * + * http://www.apache.org/licenses/LICENSE-2.0 + * + * Unless required by applicable law or agreed to in writing, software + * distributed under the License is distributed on an "AS IS" BASIS, + * WITHOUT WARRANTIES OR CONDITIONS OF ANY KIND, either express or implied. + * See the License for the specific language governing permissions and + * limitations under the License. + */ + +#pragma once + +#include "EtcConfig.h" +#include "EtcImage.h" +#include "EtcColor.h" +#include "EtcErrorMetric.h" + +namespace Etc +{ + class Block4x4EncodingBits; + + // C-style inteface to the encoder + void Encode(float *a_pafSourceRGBA, + unsigned int a_uiSourceWidth, + unsigned int a_uiSourceHeight, + Image::Format a_format, + ErrorMetric a_eErrMetric, + float a_fEffort, + unsigned int a_uiJobs, + unsigned int a_uimaxJobs, + unsigned char **a_ppaucEncodingBits, + unsigned int *a_puiEncodingBitsBytes, + unsigned int *a_puiExtendedWidth, + unsigned int *a_puiExtendedHeight, + int *a_piEncodingTime_ms, bool a_bVerboseOutput = false); + +} diff --git a/EtcLib/Etc/EtcColor.h b/EtcLib/Etc/EtcColor.h new file mode 100644 index 0000000..0bd86e8 --- /dev/null +++ b/EtcLib/Etc/EtcColor.h @@ -0,0 +1,64 @@ +/* + * Copyright 2015 The Etc2Comp Authors. + * + * Licensed under the Apache License, Version 2.0 (the "License"); + * you may not use this file except in compliance with the License. + * You may obtain a copy of the License at + * + * http://www.apache.org/licenses/LICENSE-2.0 + * + * Unless required by applicable law or agreed to in writing, software + * distributed under the License is distributed on an "AS IS" BASIS, + * WITHOUT WARRANTIES OR CONDITIONS OF ANY KIND, either express or implied. + * See the License for the specific language governing permissions and + * limitations under the License. + */ + +#pragma once + +#include + +namespace Etc +{ + + inline float LogToLinear(float a_fLog) + { + static const float ALPHA = 0.055f; + static const float ONE_PLUS_ALPHA = 1.0f + ALPHA; + + if (a_fLog <= 0.04045f) + { + return a_fLog / 12.92f; + } + else + { + return powf((a_fLog + ALPHA) / ONE_PLUS_ALPHA, 2.4f); + } + } + + inline float LinearToLog(float &a_fLinear) + { + static const float ALPHA = 0.055f; + static const float ONE_PLUS_ALPHA = 1.0f + ALPHA; + + if (a_fLinear <= 0.0031308f) + { + return 12.92f * a_fLinear; + } + else + { + return ONE_PLUS_ALPHA * powf(a_fLinear, (1.0f/2.4f)) - ALPHA; + } + } + + class ColorR8G8B8A8 + { + public: + + unsigned char ucR; + unsigned char ucG; + unsigned char ucB; + unsigned char ucA; + + }; +} diff --git a/EtcLib/Etc/EtcColorFloatRGBA.h b/EtcLib/Etc/EtcColorFloatRGBA.h new file mode 100644 index 0000000..5afd6ef --- /dev/null +++ b/EtcLib/Etc/EtcColorFloatRGBA.h @@ -0,0 +1,321 @@ +/* + * Copyright 2015 The Etc2Comp Authors. + * + * Licensed under the Apache License, Version 2.0 (the "License"); + * you may not use this file except in compliance with the License. + * You may obtain a copy of the License at + * + * http://www.apache.org/licenses/LICENSE-2.0 + * + * Unless required by applicable law or agreed to in writing, software + * distributed under the License is distributed on an "AS IS" BASIS, + * WITHOUT WARRANTIES OR CONDITIONS OF ANY KIND, either express or implied. + * See the License for the specific language governing permissions and + * limitations under the License. + */ + +#pragma once + +#include "EtcConfig.h" +#include "EtcColor.h" + +#include + +namespace Etc +{ + + class ColorFloatRGBA + { + public: + + ColorFloatRGBA(void) + { + fR = fG = fB = fA = 0.0f; + } + + ColorFloatRGBA(float a_fR, float a_fG, float a_fB, float a_fA) + { + fR = a_fR; + fG = a_fG; + fB = a_fB; + fA = a_fA; + } + + inline ColorFloatRGBA operator+(ColorFloatRGBA& a_rfrgba) + { + ColorFloatRGBA frgba; + frgba.fR = fR + a_rfrgba.fR; + frgba.fG = fG + a_rfrgba.fG; + frgba.fB = fB + a_rfrgba.fB; + frgba.fA = fA + a_rfrgba.fA; + return frgba; + } + + inline ColorFloatRGBA operator+(float a_f) + { + ColorFloatRGBA frgba; + frgba.fR = fR + a_f; + frgba.fG = fG + a_f; + frgba.fB = fB + a_f; + frgba.fA = fA; + return frgba; + } + + inline ColorFloatRGBA operator-(float a_f) + { + ColorFloatRGBA frgba; + frgba.fR = fR - a_f; + frgba.fG = fG - a_f; + frgba.fB = fB - a_f; + frgba.fA = fA; + return frgba; + } + + inline ColorFloatRGBA operator-(ColorFloatRGBA& a_rfrgba) + { + ColorFloatRGBA frgba; + frgba.fR = fR - a_rfrgba.fR; + frgba.fG = fG - a_rfrgba.fG; + frgba.fB = fB - a_rfrgba.fB; + frgba.fA = fA - a_rfrgba.fA; + return frgba; + } + + inline ColorFloatRGBA operator*(float a_f) + { + ColorFloatRGBA frgba; + frgba.fR = fR * a_f; + frgba.fG = fG * a_f; + frgba.fB = fB * a_f; + frgba.fA = fA; + + return frgba; + } + + inline ColorFloatRGBA ScaleRGB(float a_f) + { + ColorFloatRGBA frgba; + frgba.fR = a_f * fR; + frgba.fG = a_f * fG; + frgba.fB = a_f * fB; + frgba.fA = fA; + + return frgba; + } + + inline ColorFloatRGBA RoundRGB(void) + { + ColorFloatRGBA frgba; + frgba.fR = roundf(fR); + frgba.fG = roundf(fG); + frgba.fB = roundf(fB); + + return frgba; + } + + inline ColorFloatRGBA ToLinear() + { + ColorFloatRGBA frgbaLinear; + frgbaLinear.fR = LogToLinear(fR); + frgbaLinear.fG = LogToLinear(fG); + frgbaLinear.fB = LogToLinear(fB); + frgbaLinear.fA = fA; + + return frgbaLinear; + } + + inline ColorFloatRGBA ToLog(void) + { + ColorFloatRGBA frgbaLog; + frgbaLog.fR = LinearToLog(fR); + frgbaLog.fG = LinearToLog(fG); + frgbaLog.fB = LinearToLog(fB); + frgbaLog.fA = fA; + + return frgbaLog; + } + + inline static ColorFloatRGBA ConvertFromRGBA8(unsigned char a_ucR, + unsigned char a_ucG, unsigned char a_ucB, unsigned char a_ucA) + { + ColorFloatRGBA frgba; + + frgba.fR = (float)a_ucR / 255.0f; + frgba.fG = (float)a_ucG / 255.0f; + frgba.fB = (float)a_ucB / 255.0f; + frgba.fA = (float)a_ucA / 255.0f; + + return frgba; + } + + inline static ColorFloatRGBA ConvertFromRGB4(unsigned char a_ucR4, + unsigned char a_ucG4, + unsigned char a_ucB4) + { + ColorFloatRGBA frgba; + + unsigned char ucR8 = (unsigned char)((a_ucR4 << 4) + a_ucR4); + unsigned char ucG8 = (unsigned char)((a_ucG4 << 4) + a_ucG4); + unsigned char ucB8 = (unsigned char)((a_ucB4 << 4) + a_ucB4); + + frgba.fR = (float)ucR8 / 255.0f; + frgba.fG = (float)ucG8 / 255.0f; + frgba.fB = (float)ucB8 / 255.0f; + frgba.fA = 1.0f; + + return frgba; + } + + inline static ColorFloatRGBA ConvertFromRGB5(unsigned char a_ucR5, + unsigned char a_ucG5, + unsigned char a_ucB5) + { + ColorFloatRGBA frgba; + + unsigned char ucR8 = (unsigned char)((a_ucR5 << 3) + (a_ucR5 >> 2)); + unsigned char ucG8 = (unsigned char)((a_ucG5 << 3) + (a_ucG5 >> 2)); + unsigned char ucB8 = (unsigned char)((a_ucB5 << 3) + (a_ucB5 >> 2)); + + frgba.fR = (float)ucR8 / 255.0f; + frgba.fG = (float)ucG8 / 255.0f; + frgba.fB = (float)ucB8 / 255.0f; + frgba.fA = 1.0f; + + return frgba; + } + + inline static ColorFloatRGBA ConvertFromR6G7B6(unsigned char a_ucR6, + unsigned char a_ucG7, + unsigned char a_ucB6) + { + ColorFloatRGBA frgba; + + unsigned char ucR8 = (unsigned char)((a_ucR6 << 2) + (a_ucR6 >> 4)); + unsigned char ucG8 = (unsigned char)((a_ucG7 << 1) + (a_ucG7 >> 6)); + unsigned char ucB8 = (unsigned char)((a_ucB6 << 2) + (a_ucB6 >> 4)); + + frgba.fR = (float)ucR8 / 255.0f; + frgba.fG = (float)ucG8 / 255.0f; + frgba.fB = (float)ucB8 / 255.0f; + frgba.fA = 1.0f; + + return frgba; + } + + // quantize to 4 bits, expand to 8 bits + inline ColorFloatRGBA QuantizeR4G4B4(void) const + { + ColorFloatRGBA frgba = *this; + + // quantize to 4 bits + frgba = frgba.ClampRGB().ScaleRGB(15.0f).RoundRGB(); + unsigned int uiR4 = (unsigned int)frgba.fR; + unsigned int uiG4 = (unsigned int)frgba.fG; + unsigned int uiB4 = (unsigned int)frgba.fB; + + // expand to 8 bits + frgba.fR = (float) ((uiR4 << 4) + uiR4); + frgba.fG = (float) ((uiG4 << 4) + uiG4); + frgba.fB = (float) ((uiB4 << 4) + uiB4); + + frgba = frgba.ScaleRGB(1.0f/255.0f); + + return frgba; + } + + // quantize to 5 bits, expand to 8 bits + inline ColorFloatRGBA QuantizeR5G5B5(void) const + { + ColorFloatRGBA frgba = *this; + + // quantize to 5 bits + frgba = frgba.ClampRGB().ScaleRGB(31.0f).RoundRGB(); + unsigned int uiR5 = (unsigned int)frgba.fR; + unsigned int uiG5 = (unsigned int)frgba.fG; + unsigned int uiB5 = (unsigned int)frgba.fB; + + // expand to 8 bits + frgba.fR = (float)((uiR5 << 3) + (uiR5 >> 2)); + frgba.fG = (float)((uiG5 << 3) + (uiG5 >> 2)); + frgba.fB = (float)((uiB5 << 3) + (uiB5 >> 2)); + + frgba = frgba.ScaleRGB(1.0f / 255.0f); + + return frgba; + } + + // quantize to 6/7/6 bits, expand to 8 bits + inline ColorFloatRGBA QuantizeR6G7B6(void) const + { + ColorFloatRGBA frgba = *this; + + // quantize to 6/7/6 bits + ColorFloatRGBA frgba6 = frgba.ClampRGB().ScaleRGB(63.0f).RoundRGB(); + ColorFloatRGBA frgba7 = frgba.ClampRGB().ScaleRGB(127.0f).RoundRGB(); + unsigned int uiR6 = (unsigned int)frgba6.fR; + unsigned int uiG7 = (unsigned int)frgba7.fG; + unsigned int uiB6 = (unsigned int)frgba6.fB; + + // expand to 8 bits + frgba.fR = (float)((uiR6 << 2) + (uiR6 >> 4)); + frgba.fG = (float)((uiG7 << 1) + (uiG7 >> 6)); + frgba.fB = (float)((uiB6 << 2) + (uiB6 >> 4)); + + frgba = frgba.ScaleRGB(1.0f / 255.0f); + + return frgba; + } + + inline ColorFloatRGBA ClampRGB(void) + { + ColorFloatRGBA frgba = *this; + if (frgba.fR < 0.0f) { frgba.fR = 0.0f; } + if (frgba.fR > 1.0f) { frgba.fR = 1.0f; } + if (frgba.fG < 0.0f) { frgba.fG = 0.0f; } + if (frgba.fG > 1.0f) { frgba.fG = 1.0f; } + if (frgba.fB < 0.0f) { frgba.fB = 0.0f; } + if (frgba.fB > 1.0f) { frgba.fB = 1.0f; } + + return frgba; + } + + inline ColorFloatRGBA ClampRGBA(void) + { + ColorFloatRGBA frgba = *this; + if (frgba.fR < 0.0f) { frgba.fR = 0.0f; } + if (frgba.fR > 1.0f) { frgba.fR = 1.0f; } + if (frgba.fG < 0.0f) { frgba.fG = 0.0f; } + if (frgba.fG > 1.0f) { frgba.fG = 1.0f; } + if (frgba.fB < 0.0f) { frgba.fB = 0.0f; } + if (frgba.fB > 1.0f) { frgba.fB = 1.0f; } + if (frgba.fA < 0.0f) { frgba.fA = 0.0f; } + if (frgba.fA > 1.0f) { frgba.fA = 1.0f; } + + return frgba; + } + + inline int IntRed(float a_fScale) + { + return (int)roundf(fR * a_fScale); + } + + inline int IntGreen(float a_fScale) + { + return (int)roundf(fG * a_fScale); + } + + inline int IntBlue(float a_fScale) + { + return (int)roundf(fB * a_fScale); + } + + inline int IntAlpha(float a_fScale) + { + return (int)roundf(fA * a_fScale); + } + + float fR, fG, fB, fA; + }; + +} + diff --git a/EtcLib/Etc/EtcConfig.h b/EtcLib/Etc/EtcConfig.h new file mode 100644 index 0000000..6ce1b8c --- /dev/null +++ b/EtcLib/Etc/EtcConfig.h @@ -0,0 +1,67 @@ +/* + * Copyright 2015 The Etc2Comp Authors. + * + * Licensed under the Apache License, Version 2.0 (the "License"); + * you may not use this file except in compliance with the License. + * You may obtain a copy of the License at + * + * http://www.apache.org/licenses/LICENSE-2.0 + * + * Unless required by applicable law or agreed to in writing, software + * distributed under the License is distributed on an "AS IS" BASIS, + * WITHOUT WARRANTIES OR CONDITIONS OF ANY KIND, either express or implied. + * See the License for the specific language governing permissions and + * limitations under the License. + */ + +#pragma once + +#ifdef _WIN32 +#define ETC_WINDOWS (1) +#else +#define ETC_WINDOWS (0) +#endif + +#if __APPLE__ +#define ETC_OSX (1) +#else +#define ETC_OSX (0) +#endif + +#if __unix__ +#define ETC_UNIX (1) +#else +#define ETC_UNIX (0) +#endif + + +// short names for common types +#include +typedef int8_t i8; +typedef int16_t i16; +typedef int32_t i32; +typedef int64_t i64; + +typedef uint8_t u8; +typedef uint16_t u16; +typedef uint32_t u32; +typedef uint64_t u64; + +typedef float f32; +typedef double f64; + +// Keep asserts enabled in release builds during development +#undef NDEBUG + +// 0=disable. stb_image can be used if you need to compress +//other image formats like jpg +#define USE_STB_IMAGE_LOAD 0 + +#if ETC_WINDOWS +#include +#define _CRT_SECURE_NO_WARNINGS (1) +#include +#endif + +#include + diff --git a/EtcLib/Etc/EtcImage.cpp b/EtcLib/Etc/EtcImage.cpp new file mode 100644 index 0000000..94a0e7b --- /dev/null +++ b/EtcLib/Etc/EtcImage.cpp @@ -0,0 +1,685 @@ +/* + * Copyright 2015 The Etc2Comp Authors. + * + * Licensed under the Apache License, Version 2.0 (the "License"); + * you may not use this file except in compliance with the License. + * You may obtain a copy of the License at + * + * http://www.apache.org/licenses/LICENSE-2.0 + * + * Unless required by applicable law or agreed to in writing, software + * distributed under the License is distributed on an "AS IS" BASIS, + * WITHOUT WARRANTIES OR CONDITIONS OF ANY KIND, either express or implied. + * See the License for the specific language governing permissions and + * limitations under the License. + */ + +/* +EtcImage.cpp + +Image is an array of 4x4 blocks that represent the encoding of the source image + +*/ + +#include "EtcConfig.h" + +#include + +#include "EtcImage.h" + +#include "Etc.h" +#include "EtcBlock4x4.h" +#include "EtcBlock4x4EncodingBits.h" +#include "EtcSortedBlockList.h" + +#if ETC_WINDOWS +#include +#endif +#include +#include +#include +#include +#include +#include + +// fix conflict with Block4x4::AlphaMix +#ifdef OPAQUE +#undef OPAQUE +#endif +#ifdef TRANSPARENT +#undef TRANSPARENT +#endif + +namespace Etc +{ + + // ---------------------------------------------------------------------------------------------------- + // + Image::Image(void) + { + m_encodingStatus = EncodingStatus::SUCCESS; + m_warningsToCapture = EncodingStatus::SUCCESS; + m_pafrgbaSource = nullptr; + + m_pablock = nullptr; + + m_encodingbitsformat = Block4x4EncodingBits::Format::UNKNOWN; + m_uiEncodingBitsBytes = 0; + m_paucEncodingBits = nullptr; + + m_format = Format::UNKNOWN; + m_iNumOpaquePixels = 0; + m_iNumTranslucentPixels = 0; + m_iNumTransparentPixels = 0; + } + + // ---------------------------------------------------------------------------------------------------- + // constructor using source image + // used to set state before Encode() is called + // + Image::Image(float *a_pafSourceRGBA, unsigned int a_uiSourceWidth, + unsigned int a_uiSourceHeight, + ErrorMetric a_errormetric) + { + m_encodingStatus = EncodingStatus::SUCCESS; + m_warningsToCapture = EncodingStatus::SUCCESS; + m_pafrgbaSource = (ColorFloatRGBA *) a_pafSourceRGBA; + m_uiSourceWidth = a_uiSourceWidth; + m_uiSourceHeight = a_uiSourceHeight; + + m_uiExtendedWidth = CalcExtendedDimension((unsigned short)m_uiSourceWidth); + m_uiExtendedHeight = CalcExtendedDimension((unsigned short)m_uiSourceHeight); + + m_uiBlockColumns = m_uiExtendedWidth >> 2; + m_uiBlockRows = m_uiExtendedHeight >> 2; + + m_pablock = new Block4x4[GetNumberOfBlocks()]; + assert(m_pablock); + + m_format = Format::UNKNOWN; + + m_encodingbitsformat = Block4x4EncodingBits::Format::UNKNOWN; + m_uiEncodingBitsBytes = 0; + m_paucEncodingBits = nullptr; + + m_errormetric = a_errormetric; + m_fEffort = 0.0f; + + m_iEncodeTime_ms = -1; + + m_iNumOpaquePixels = 0; + m_iNumTranslucentPixels = 0; + m_iNumTransparentPixels = 0; + m_bVerboseOutput = false; + + } + + // ---------------------------------------------------------------------------------------------------- + // constructor using encoding bits + // recreates encoding state using a previously encoded image + // + Image::Image(Format a_format, + unsigned int a_uiSourceWidth, unsigned int a_uiSourceHeight, + unsigned char *a_paucEncidingBits, unsigned int a_uiEncodingBitsBytes, + Image *a_pimageSource, ErrorMetric a_errormetric) + { + m_encodingStatus = EncodingStatus::SUCCESS; + m_pafrgbaSource = nullptr; + m_uiSourceWidth = a_uiSourceWidth; + m_uiSourceHeight = a_uiSourceHeight; + + m_uiExtendedWidth = CalcExtendedDimension((unsigned short)m_uiSourceWidth); + m_uiExtendedHeight = CalcExtendedDimension((unsigned short)m_uiSourceHeight); + + m_uiBlockColumns = m_uiExtendedWidth >> 2; + m_uiBlockRows = m_uiExtendedHeight >> 2; + + unsigned int uiBlocks = GetNumberOfBlocks(); + + m_pablock = new Block4x4[uiBlocks]; + assert(m_pablock); + + m_format = a_format; + + m_iNumOpaquePixels = 0; + m_iNumTranslucentPixels = 0; + m_iNumTransparentPixels = 0; + + m_encodingbitsformat = DetermineEncodingBitsFormat(m_format); + if (m_encodingbitsformat == Block4x4EncodingBits::Format::UNKNOWN) + { + AddToEncodingStatus(ERROR_UNKNOWN_FORMAT); + return; + } + m_uiEncodingBitsBytes = a_uiEncodingBitsBytes; + m_paucEncodingBits = a_paucEncidingBits; + + m_errormetric = a_errormetric; + m_fEffort = 0.0f; + m_bVerboseOutput = false; + m_iEncodeTime_ms = -1; + + unsigned char *paucEncodingBits = m_paucEncodingBits; + unsigned int uiEncodingBitsBytesPerBlock = Block4x4EncodingBits::GetBytesPerBlock(m_encodingbitsformat); + + unsigned int uiH = 0; + unsigned int uiV = 0; + for (unsigned int uiBlock = 0; uiBlock < uiBlocks; uiBlock++) + { + m_pablock[uiBlock].InitFromEtcEncodingBits(a_format, uiH, uiV, paucEncodingBits, + a_pimageSource, a_errormetric); + paucEncodingBits += uiEncodingBitsBytesPerBlock; + uiH += 4; + if (uiH >= m_uiSourceWidth) + { + uiH = 0; + uiV += 4; + } + } + + } + + // ---------------------------------------------------------------------------------------------------- + // + Image::~Image(void) + { + if (m_pablock != nullptr) + { + delete[] m_pablock; + m_pablock = nullptr; + } + + /*if (m_paucEncodingBits != nullptr) + { + delete[] m_paucEncodingBits; + m_paucEncodingBits = nullptr; + }*/ + } + + // ---------------------------------------------------------------------------------------------------- + // encode an image + // create a set of encoding bits that conforms to a_format + // find best fit using a_errormetric + // explore a range of possible encodings based on a_fEffort (range = [0:100]) + // speed up process using a_uiJobs as the number of process threads (a_uiJobs must not excede a_uiMaxJobs) + // + Image::EncodingStatus Image::Encode(Format a_format, ErrorMetric a_errormetric, float a_fEffort, unsigned int a_uiJobs, unsigned int a_uiMaxJobs) + { + + std::chrono::time_point start = std::chrono::steady_clock::now(); + + m_encodingStatus = EncodingStatus::SUCCESS; + + m_format = a_format; + m_errormetric = a_errormetric; + m_fEffort = a_fEffort; + + if (m_errormetric < 0 || m_errormetric > ERROR_METRICS) + { + AddToEncodingStatus(ERROR_UNKNOWN_ERROR_METRIC); + return m_encodingStatus; + } + + if (m_fEffort < 0.0f) + { + AddToEncodingStatus(WARNING_EFFORT_OUT_OF_RANGE); + m_fEffort = 0.0f; + } + else if (m_fEffort > 100.0f) + { + AddToEncodingStatus(WARNING_EFFORT_OUT_OF_RANGE); + m_fEffort = 100.0f; + } + if (a_uiJobs < 1) + { + a_uiJobs = 1; + AddToEncodingStatus(WARNING_JOBS_OUT_OF_RANGE); + } + else if (a_uiJobs > a_uiMaxJobs) + { + a_uiJobs = a_uiMaxJobs; + AddToEncodingStatus(WARNING_JOBS_OUT_OF_RANGE); + } + + m_encodingbitsformat = DetermineEncodingBitsFormat(m_format); + + if (m_encodingbitsformat == Block4x4EncodingBits::Format::UNKNOWN) + { + AddToEncodingStatus(ERROR_UNKNOWN_FORMAT); + return m_encodingStatus; + } + + assert(m_paucEncodingBits == nullptr); + m_uiEncodingBitsBytes = GetNumberOfBlocks() * Block4x4EncodingBits::GetBytesPerBlock(m_encodingbitsformat); + m_paucEncodingBits = new unsigned char[m_uiEncodingBitsBytes]; + + InitBlocksAndBlockSorter(); + + + std::future *handle = new std::future[a_uiMaxJobs]; + + unsigned int uiNumThreadsNeeded = 0; + unsigned int uiUnfinishedBlocks = GetNumberOfBlocks(); + + uiNumThreadsNeeded = (uiUnfinishedBlocks < a_uiJobs) ? uiUnfinishedBlocks : a_uiJobs; + + for (int i = 0; i < (int)uiNumThreadsNeeded - 1; i++) + { + handle[i] = async(std::launch::async, &Image::RunFirstPass, this, i, uiNumThreadsNeeded); + } + + RunFirstPass(uiNumThreadsNeeded - 1, uiNumThreadsNeeded); + + for (int i = 0; i < (int)uiNumThreadsNeeded - 1; i++) + { + handle[i].get(); + } + + // perform effort-based encoding + if (m_fEffort > 0.0f) + { + unsigned int uiFinishedBlocks = 0; + unsigned int uiTotalEffortBlocks = static_cast(roundf(0.01f * m_fEffort * GetNumberOfBlocks())); + + if (m_bVerboseOutput) + { + printf("effortblocks = %d\n", uiTotalEffortBlocks); + } + unsigned int uiPass = 0; + while (1) + { + if (m_bVerboseOutput) + { + uiPass++; + printf("pass %u\n", uiPass); + } + m_psortedblocklist->Sort(); + uiUnfinishedBlocks = m_psortedblocklist->GetNumberOfSortedBlocks(); + uiFinishedBlocks = GetNumberOfBlocks() - uiUnfinishedBlocks; + if (m_bVerboseOutput) + { + printf(" %u unfinished blocks\n", uiUnfinishedBlocks); + // m_psortedblocklist->Print(); + } + + + + //stop enocding when we did enough to satify the effort percentage + if (uiFinishedBlocks >= uiTotalEffortBlocks) + { + if (m_bVerboseOutput) + { + printf("Finished %d Blocks out of %d\n", uiFinishedBlocks, uiTotalEffortBlocks); + } + break; + } + + unsigned int uiIteratedBlocks = 0; + unsigned int blocksToIterateThisPass = (uiTotalEffortBlocks - uiFinishedBlocks); + uiNumThreadsNeeded = (uiUnfinishedBlocks < a_uiJobs) ? uiUnfinishedBlocks : a_uiJobs; + + if (uiNumThreadsNeeded <= 1) + { + //since we already how many blocks each thread will process + //cap the thread limit to do the proper amount of work, and not more + uiIteratedBlocks = IterateThroughWorstBlocks(blocksToIterateThisPass, 0, 1); + } + else + { + //we have a lot of work to do, so lets multi thread it + std::future *handleToBlockEncoders = new std::future[uiNumThreadsNeeded-1]; + + for (int i = 0; i < (int)uiNumThreadsNeeded - 1; i++) + { + handleToBlockEncoders[i] = async(std::launch::async, &Image::IterateThroughWorstBlocks, this, blocksToIterateThisPass, i, uiNumThreadsNeeded); + } + uiIteratedBlocks = IterateThroughWorstBlocks(blocksToIterateThisPass, uiNumThreadsNeeded - 1, uiNumThreadsNeeded); + + for (int i = 0; i < (int)uiNumThreadsNeeded - 1; i++) + { + uiIteratedBlocks += handleToBlockEncoders[i].get(); + } + + delete[] handleToBlockEncoders; + } + + if (m_bVerboseOutput) + { + printf(" %u iterated blocks\n", uiIteratedBlocks); + } + } + } + + // generate Etc2-compatible bit-format 4x4 blocks + for (int i = 0; i < (int)a_uiJobs - 1; i++) + { + handle[i] = async(std::launch::async, &Image::SetEncodingBits, this, i, a_uiJobs); + } + SetEncodingBits(a_uiJobs - 1, a_uiJobs); + + for (int i = 0; i < (int)a_uiJobs - 1; i++) + { + handle[i].get(); + } + + std::chrono::time_point end = std::chrono::steady_clock::now(); + std::chrono::milliseconds elapsed = std::chrono::duration_cast(end - start); + m_iEncodeTime_ms = (int)elapsed.count(); + + delete[] handle; + delete m_psortedblocklist; + return m_encodingStatus; + } + + // ---------------------------------------------------------------------------------------------------- + // iterate the encoding thru the blocks with the worst error + // stop when a_uiMaxBlocks blocks have been iterated + // split the blocks between the process threads using a_uiMultithreadingOffset and a_uiMultithreadingStride + // + unsigned int Image::IterateThroughWorstBlocks(unsigned int a_uiMaxBlocks, + unsigned int a_uiMultithreadingOffset, + unsigned int a_uiMultithreadingStride) + { + assert(a_uiMultithreadingStride > 0); + unsigned int uiIteratedBlocks = a_uiMultithreadingOffset; + + SortedBlockList::Link *plink = m_psortedblocklist->GetLinkToFirstBlock(); + for (plink = plink->Advance(a_uiMultithreadingOffset); + plink != nullptr; + plink = plink->Advance(a_uiMultithreadingStride) ) + { + if (uiIteratedBlocks >= a_uiMaxBlocks) + { + break; + } + + plink->GetBlock()->PerformEncodingIteration(m_fEffort); + + uiIteratedBlocks += a_uiMultithreadingStride; + } + + return uiIteratedBlocks; + } + + // ---------------------------------------------------------------------------------------------------- + // determine which warnings to check for during Encode() based on encoding format + // + void Image::FindEncodingWarningTypesForCurFormat() + { + TrackEncodingWarning(WARNING_ALL_TRANSPARENT_PIXELS); + TrackEncodingWarning(WARNING_SOME_RGBA_NOT_0_TO_1); + switch (m_format) + { + case Image::Format::ETC1: + case Image::Format::RGB8: + case Image::Format::SRGB8: + TrackEncodingWarning(WARNING_SOME_NON_OPAQUE_PIXELS); + TrackEncodingWarning(WARNING_SOME_TRANSLUCENT_PIXELS); + break; + + case Image::Format::RGB8A1: + case Image::Format::SRGB8A1: + TrackEncodingWarning(WARNING_SOME_TRANSLUCENT_PIXELS); + TrackEncodingWarning(WARNING_ALL_OPAQUE_PIXELS); + break; + case Image::Format::RGBA8: + case Image::Format::SRGBA8: + TrackEncodingWarning(WARNING_ALL_OPAQUE_PIXELS); + break; + + case Image::Format::R11: + case Image::Format::SIGNED_R11: + TrackEncodingWarning(WARNING_SOME_NON_OPAQUE_PIXELS); + TrackEncodingWarning(WARNING_SOME_TRANSLUCENT_PIXELS); + TrackEncodingWarning(WARNING_SOME_GREEN_VALUES_ARE_NOT_ZERO); + TrackEncodingWarning(WARNING_SOME_BLUE_VALUES_ARE_NOT_ZERO); + break; + + case Image::Format::RG11: + case Image::Format::SIGNED_RG11: + TrackEncodingWarning(WARNING_SOME_NON_OPAQUE_PIXELS); + TrackEncodingWarning(WARNING_SOME_TRANSLUCENT_PIXELS); + TrackEncodingWarning(WARNING_SOME_BLUE_VALUES_ARE_NOT_ZERO); + break; + case Image::Format::FORMATS: + case Image::Format::UNKNOWN: + default: + assert(0); + break; + } + } + + // ---------------------------------------------------------------------------------------------------- + // examine source pixels to check for warnings + // + void Image::FindAndSetEncodingWarnings() + { + int numPixels = (m_uiBlockRows * 4) * (m_uiBlockColumns * 4); + if (m_iNumOpaquePixels == numPixels) + { + AddToEncodingStatusIfSignfigant(Image::EncodingStatus::WARNING_ALL_OPAQUE_PIXELS); + } + if (m_iNumOpaquePixels < numPixels) + { + AddToEncodingStatusIfSignfigant(Image::EncodingStatus::WARNING_SOME_NON_OPAQUE_PIXELS); + } + if (m_iNumTranslucentPixels > 0) + { + AddToEncodingStatusIfSignfigant(Image::EncodingStatus::WARNING_SOME_TRANSLUCENT_PIXELS); + } + if (m_iNumTransparentPixels == numPixels) + { + AddToEncodingStatusIfSignfigant(Image::EncodingStatus::WARNING_ALL_TRANSPARENT_PIXELS); + } + if (m_numColorValues.fB > 0.0f) + { + AddToEncodingStatusIfSignfigant(Image::EncodingStatus::WARNING_SOME_BLUE_VALUES_ARE_NOT_ZERO); + } + if (m_numColorValues.fG > 0.0f) + { + AddToEncodingStatusIfSignfigant(Image::EncodingStatus::WARNING_SOME_GREEN_VALUES_ARE_NOT_ZERO); + } + + if (m_numOutOfRangeValues.fR > 0.0f || m_numOutOfRangeValues.fG > 0.0f) + { + AddToEncodingStatusIfSignfigant(Image::EncodingStatus::WARNING_SOME_RGBA_NOT_0_TO_1); + } + if (m_numOutOfRangeValues.fB > 0.0f || m_numOutOfRangeValues.fA > 0.0f) + { + AddToEncodingStatusIfSignfigant(Image::EncodingStatus::WARNING_SOME_RGBA_NOT_0_TO_1); + } + } + + // ---------------------------------------------------------------------------------------------------- + // return a string name for a given image format + // + const char * Image::EncodingFormatToString(Image::Format a_format) + { + switch (a_format) + { + case Image::Format::ETC1: + return "ETC1"; + case Image::Format::RGB8: + return "RGB8"; + case Image::Format::SRGB8: + return "SRGB8"; + + case Image::Format::RGB8A1: + return "RGB8A1"; + case Image::Format::SRGB8A1: + return "SRGB8A1"; + case Image::Format::RGBA8: + return "RGBA8"; + case Image::Format::SRGBA8: + return "SRGBA8"; + + case Image::Format::R11: + return "R11"; + case Image::Format::SIGNED_R11: + return "SIGNED_R11"; + + case Image::Format::RG11: + return "RG11"; + case Image::Format::SIGNED_RG11: + return "SIGNED_RG11"; + case Image::Format::FORMATS: + case Image::Format::UNKNOWN: + default: + return "UNKNOWN"; + } + } + + // ---------------------------------------------------------------------------------------------------- + // return a string name for the image's format + // + const char * Image::EncodingFormatToString(void) + { + return EncodingFormatToString(m_format); + } + + // ---------------------------------------------------------------------------------------------------- + // init image blocks prior to encoding + // init block sorter for subsequent sortings + // check for encoding warnings + // + void Image::InitBlocksAndBlockSorter(void) + { + + FindEncodingWarningTypesForCurFormat(); + + // init each block + Block4x4 *pblock = m_pablock; + unsigned char *paucEncodingBits = m_paucEncodingBits; + for (unsigned int uiBlockRow = 0; uiBlockRow < m_uiBlockRows; uiBlockRow++) + { + unsigned int uiBlockV = uiBlockRow * 4; + + for (unsigned int uiBlockColumn = 0; uiBlockColumn < m_uiBlockColumns; uiBlockColumn++) + { + unsigned int uiBlockH = uiBlockColumn * 4; + + pblock->InitFromSource(this, uiBlockH, uiBlockV, paucEncodingBits, m_errormetric); + + paucEncodingBits += Block4x4EncodingBits::GetBytesPerBlock(m_encodingbitsformat); + + pblock++; + } + } + + FindAndSetEncodingWarnings(); + + // init block sorter + { + m_psortedblocklist = new SortedBlockList(GetNumberOfBlocks(), 100); + + for (unsigned int uiBlock = 0; uiBlock < GetNumberOfBlocks(); uiBlock++) + { + pblock = &m_pablock[uiBlock]; + m_psortedblocklist->AddBlock(pblock); + } + } + + } + + // ---------------------------------------------------------------------------------------------------- + // run the first pass of the encoder + // the encoder generally finds a reasonable, fast encoding + // this is run on all blocks regardless of effort to ensure that all blocks have a valid encoding + // + void Image::RunFirstPass(unsigned int a_uiMultithreadingOffset, unsigned int a_uiMultithreadingStride) + { + assert(a_uiMultithreadingStride > 0); + + for (unsigned int uiBlock = a_uiMultithreadingOffset; + uiBlock < GetNumberOfBlocks(); + uiBlock += a_uiMultithreadingStride) + { + Block4x4 *pblock = &m_pablock[uiBlock]; + pblock->PerformEncodingIteration(m_fEffort); + } + } + + // ---------------------------------------------------------------------------------------------------- + // set the encoding bits (for the output file) based on the best encoding for each block + // + void Image::SetEncodingBits(unsigned int a_uiMultithreadingOffset, + unsigned int a_uiMultithreadingStride) + { + assert(a_uiMultithreadingStride > 0); + + for (unsigned int uiBlock = a_uiMultithreadingOffset; + uiBlock < GetNumberOfBlocks(); + uiBlock += a_uiMultithreadingStride) + { + Block4x4 *pblock = &m_pablock[uiBlock]; + pblock->SetEncodingBitsFromEncoding(); + } + + } + + // ---------------------------------------------------------------------------------------------------- + // return the image error + // image error is the sum of all block errors + // + float Image::GetError(void) + { + float fError = 0.0f; + + for (unsigned int uiBlock = 0; uiBlock < GetNumberOfBlocks(); uiBlock++) + { + Block4x4 *pblock = &m_pablock[uiBlock]; + fError += pblock->GetError(); + } + + return fError; + } + + // ---------------------------------------------------------------------------------------------------- + // determine the encoding bits format based on the encoding format + // the encoding bits format is a family of bit encodings that are shared across various encoding formats + // + Block4x4EncodingBits::Format Image::DetermineEncodingBitsFormat(Format a_format) + { + Block4x4EncodingBits::Format encodingbitsformat; + + // determine encoding bits format from image format + switch (a_format) + { + case Format::ETC1: + case Format::RGB8: + case Format::SRGB8: + encodingbitsformat = Block4x4EncodingBits::Format::RGB8; + break; + + case Format::RGBA8: + case Format::SRGBA8: + encodingbitsformat = Block4x4EncodingBits::Format::RGBA8; + break; + + case Format::R11: + case Format::SIGNED_R11: + encodingbitsformat = Block4x4EncodingBits::Format::R11; + break; + + case Format::RG11: + case Format::SIGNED_RG11: + encodingbitsformat = Block4x4EncodingBits::Format::RG11; + break; + + case Format::RGB8A1: + case Format::SRGB8A1: + encodingbitsformat = Block4x4EncodingBits::Format::RGB8A1; + break; + + default: + encodingbitsformat = Block4x4EncodingBits::Format::UNKNOWN; + break; + } + + return encodingbitsformat; + } + + // ---------------------------------------------------------------------------------------------------- + // + +} // namespace Etc diff --git a/EtcLib/Etc/EtcImage.h b/EtcLib/Etc/EtcImage.h new file mode 100644 index 0000000..fb67131 --- /dev/null +++ b/EtcLib/Etc/EtcImage.h @@ -0,0 +1,249 @@ +/* + * Copyright 2015 The Etc2Comp Authors. + * + * Licensed under the Apache License, Version 2.0 (the "License"); + * you may not use this file except in compliance with the License. + * You may obtain a copy of the License at + * + * http://www.apache.org/licenses/LICENSE-2.0 + * + * Unless required by applicable law or agreed to in writing, software + * distributed under the License is distributed on an "AS IS" BASIS, + * WITHOUT WARRANTIES OR CONDITIONS OF ANY KIND, either express or implied. + * See the License for the specific language governing permissions and + * limitations under the License. + */ + +#pragma once + +//#include "Etc.h" +#include "EtcColorFloatRGBA.h" +#include "EtcBlock4x4EncodingBits.h" +#include "EtcErrorMetric.h" + + +namespace Etc +{ + class Block4x4; + class EncoderSpec; + class SortedBlockList; + + class Image + { + public: + + //the differnt warning and errors that can come up during encoding + enum EncodingStatus + { + SUCCESS = 0, + // + WARNING_THRESHOLD = 1 << 0, + // + WARNING_EFFORT_OUT_OF_RANGE = 1 << 1, + WARNING_JOBS_OUT_OF_RANGE = 1 << 2, + WARNING_SOME_NON_OPAQUE_PIXELS = 1 << 3,//just for opaque formats, etc1, rgb8, r11, rg11 + WARNING_ALL_OPAQUE_PIXELS = 1 << 4, + WARNING_ALL_TRANSPARENT_PIXELS = 1 << 5, + WARNING_SOME_TRANSLUCENT_PIXELS = 1 << 6,//just for rgb8A1 + WARNING_SOME_RGBA_NOT_0_TO_1 = 1 << 7, + WARNING_SOME_BLUE_VALUES_ARE_NOT_ZERO = 1 << 8, + WARNING_SOME_GREEN_VALUES_ARE_NOT_ZERO = 1 << 9, + // + ERROR_THRESHOLD = 1 << 16, + // + ERROR_UNKNOWN_FORMAT = 1 << 17, + ERROR_UNKNOWN_ERROR_METRIC = 1 << 18, + ERROR_ZERO_WIDTH_OR_HEIGHT = 1 << 19, + // + }; + + enum class Format + { + UNKNOWN, + // + ETC1, + // + // ETC2 formats + RGB8, + SRGB8, + RGBA8, + SRGBA8, + R11, + SIGNED_R11, + RG11, + SIGNED_RG11, + RGB8A1, + SRGB8A1, + // + FORMATS, + // + DEFAULT = SRGB8 + }; + + // constructor using source image + Image(float *a_pafSourceRGBA, unsigned int a_uiSourceWidth, + unsigned int a_uiSourceHeight, + ErrorMetric a_errormetric); + + // constructor using encoding bits + Image(Format a_format, + unsigned int a_uiSourceWidth, unsigned int a_uiSourceHeight, + unsigned char *a_paucEncidingBits, unsigned int a_uiEncodingBitsBytes, + Image *a_pimageSource, + ErrorMetric a_errormetric); + + ~Image(void); + + EncodingStatus Encode(Format a_format, ErrorMetric a_errormetric, float a_fEffort, + unsigned int a_uiJobs, unsigned int a_uiMaxJobs); + + inline void AddToEncodingStatus(EncodingStatus a_encStatus) + { + m_encodingStatus = (EncodingStatus)((unsigned int)m_encodingStatus | (unsigned int)a_encStatus); + } + + inline unsigned int GetSourceWidth(void) + { + return m_uiSourceWidth; + } + + inline unsigned int GetSourceHeight(void) + { + return m_uiSourceHeight; + } + + inline unsigned int GetExtendedWidth(void) + { + return m_uiExtendedWidth; + } + + inline unsigned int GetExtendedHeight(void) + { + return m_uiExtendedHeight; + } + + inline unsigned int GetNumberOfBlocks() + { + return m_uiBlockColumns * m_uiBlockRows; + } + + inline Block4x4 * GetBlocks() + { + return m_pablock; + } + + inline unsigned char * GetEncodingBits(void) + { + return m_paucEncodingBits; + } + + inline unsigned int GetEncodingBitsBytes(void) + { + return m_uiEncodingBitsBytes; + } + + inline int GetEncodingTimeMs(void) + { + return m_iEncodeTime_ms; + } + + float GetError(void); + + inline ColorFloatRGBA * GetSourcePixel(unsigned int a_uiH, unsigned int a_uiV) + { + if (a_uiH >= m_uiSourceWidth || a_uiV >= m_uiSourceHeight) + { + return nullptr; + } + + return &m_pafrgbaSource[a_uiV*m_uiSourceWidth + a_uiH]; + } + + inline Format GetFormat(void) + { + return m_format; + } + + static Block4x4EncodingBits::Format DetermineEncodingBitsFormat(Format a_format); + + inline static unsigned short CalcExtendedDimension(unsigned short a_ushOriginalDimension) + { + return (unsigned short)((a_ushOriginalDimension + 3) & ~3); + } + + inline ErrorMetric GetErrorMetric(void) + { + return m_errormetric; + } + + static const char * EncodingFormatToString(Image::Format a_format); + const char * EncodingFormatToString(void); + //used to get basic information about the image data + int m_iNumOpaquePixels; + int m_iNumTranslucentPixels; + int m_iNumTransparentPixels; + + ColorFloatRGBA m_numColorValues; + ColorFloatRGBA m_numOutOfRangeValues; + + bool m_bVerboseOutput; + private: + //add a warning or error to check for while encoding + inline void TrackEncodingWarning(EncodingStatus a_encStatus) + { + m_warningsToCapture = (EncodingStatus)((unsigned int)m_warningsToCapture | (unsigned int)a_encStatus); + } + + //report the warning if it is something we care about for this encoding + inline void AddToEncodingStatusIfSignfigant(EncodingStatus a_encStatus) + { + if ((EncodingStatus)((unsigned int)m_warningsToCapture & (unsigned int)a_encStatus) == a_encStatus) + { + AddToEncodingStatus(a_encStatus); + } + } + + Image(void); + void FindEncodingWarningTypesForCurFormat(); + void FindAndSetEncodingWarnings(); + + void InitBlocksAndBlockSorter(void); + + void RunFirstPass(unsigned int a_uiMultithreadingOffset, + unsigned int a_uiMultithreadingStride); + + void SetEncodingBits(unsigned int a_uiMultithreadingOffset, + unsigned int a_uiMultithreadingStride); + + unsigned int IterateThroughWorstBlocks(unsigned int a_uiMaxBlocks, + unsigned int a_uiMultithreadingOffset, + unsigned int a_uiMultithreadingStride); + + // inputs + ColorFloatRGBA *m_pafrgbaSource; + unsigned int m_uiSourceWidth; + unsigned int m_uiSourceHeight; + unsigned int m_uiExtendedWidth; + unsigned int m_uiExtendedHeight; + unsigned int m_uiBlockColumns; + unsigned int m_uiBlockRows; + // intermediate data + Block4x4 *m_pablock; + // encoding + Format m_format; + Block4x4EncodingBits::Format m_encodingbitsformat; + unsigned int m_uiEncodingBitsBytes; // for entire image + unsigned char *m_paucEncodingBits; + ErrorMetric m_errormetric; + float m_fEffort; + // stats + int m_iEncodeTime_ms; + + SortedBlockList *m_psortedblocklist; + //this will hold any warning or errors that happen during encoding + EncodingStatus m_encodingStatus; + //these will be the warnings we are tracking + EncodingStatus m_warningsToCapture; + }; + +} // namespace Etc diff --git a/EtcLib/Etc/EtcMath.cpp b/EtcLib/Etc/EtcMath.cpp new file mode 100644 index 0000000..cd70a9a --- /dev/null +++ b/EtcLib/Etc/EtcMath.cpp @@ -0,0 +1,64 @@ +/* + * Copyright 2015 The Etc2Comp Authors. + * + * Licensed under the Apache License, Version 2.0 (the "License"); + * you may not use this file except in compliance with the License. + * You may obtain a copy of the License at + * + * http://www.apache.org/licenses/LICENSE-2.0 + * + * Unless required by applicable law or agreed to in writing, software + * distributed under the License is distributed on an "AS IS" BASIS, + * WITHOUT WARRANTIES OR CONDITIONS OF ANY KIND, either express or implied. + * See the License for the specific language governing permissions and + * limitations under the License. + */ + +#include "EtcConfig.h" +#include "EtcMath.h" + +namespace Etc +{ + + // ---------------------------------------------------------------------------------------------------- + // calculate the line that best fits the set of XY points contained in a_afX[] and a_afY[] + // use a_fSlope and a_fOffset to define that line + // + bool Regression(float a_afX[], float a_afY[], unsigned int a_Points, + float *a_fSlope, float *a_fOffset) + { + float fPoints = (float)a_Points; + + float fSumX = 0.0f; + float fSumY = 0.0f; + float fSumXY = 0.0f; + float fSumX2 = 0.0f; + + for (unsigned int uiPoint = 0; uiPoint < a_Points; uiPoint++) + { + fSumX += a_afX[uiPoint]; + fSumY += a_afY[uiPoint]; + fSumXY += a_afX[uiPoint] * a_afY[uiPoint]; + fSumX2 += a_afX[uiPoint] * a_afX[uiPoint]; + } + + float fDivisor = fPoints*fSumX2 - fSumX*fSumX; + + // if vertical line + if (fDivisor == 0.0f) + { + *a_fSlope = 0.0f; + *a_fOffset = 0.0f; + return true; + } + + *a_fSlope = (fPoints*fSumXY - fSumX*fSumY) / fDivisor; + *a_fOffset = (fSumY - (*a_fSlope)*fSumX) / fPoints; + + return false; + } + + // ---------------------------------------------------------------------------------------------------- + // + +} // namespace Etc diff --git a/EtcLib/Etc/EtcMath.h b/EtcLib/Etc/EtcMath.h new file mode 100644 index 0000000..3d951fe --- /dev/null +++ b/EtcLib/Etc/EtcMath.h @@ -0,0 +1,40 @@ +/* + * Copyright 2015 The Etc2Comp Authors. + * + * Licensed under the Apache License, Version 2.0 (the "License"); + * you may not use this file except in compliance with the License. + * You may obtain a copy of the License at + * + * http://www.apache.org/licenses/LICENSE-2.0 + * + * Unless required by applicable law or agreed to in writing, software + * distributed under the License is distributed on an "AS IS" BASIS, + * WITHOUT WARRANTIES OR CONDITIONS OF ANY KIND, either express or implied. + * See the License for the specific language governing permissions and + * limitations under the License. + */ + +#pragma once + +#include + +namespace Etc +{ + + // ---------------------------------------------------------------------------------------------------- + // return true if vertical line + bool Regression(float a_afX[], float a_afY[], unsigned int a_Points, + float *a_fSlope, float *a_fOffset); + + inline float ConvertMSEToPSNR(float a_fMSE) + { + if (a_fMSE == 0.0f) + { + return INFINITY; + } + + return 10.0f * log10f(1.0f / a_fMSE); + } + + +} diff --git a/EtcLib/EtcCodec/EtcBlock4x4.cpp b/EtcLib/EtcCodec/EtcBlock4x4.cpp new file mode 100644 index 0000000..64a420e --- /dev/null +++ b/EtcLib/EtcCodec/EtcBlock4x4.cpp @@ -0,0 +1,417 @@ +/* + * Copyright 2015 The Etc2Comp Authors. + * + * Licensed under the Apache License, Version 2.0 (the "License"); + * you may not use this file except in compliance with the License. + * You may obtain a copy of the License at + * + * http://www.apache.org/licenses/LICENSE-2.0 + * + * Unless required by applicable law or agreed to in writing, software + * distributed under the License is distributed on an "AS IS" BASIS, + * WITHOUT WARRANTIES OR CONDITIONS OF ANY KIND, either express or implied. + * See the License for the specific language governing permissions and + * limitations under the License. + */ + +/* +EtcBlock4x4.cpp + +Implements the state associated with each 4x4 block of pixels in an image + +Source images that are not a multiple of 4x4 are extended to fill the Block4x4 using pixels with an +alpha of NAN + +*/ + +#include "EtcConfig.h" +#include "EtcBlock4x4.h" + +#include "EtcBlock4x4EncodingBits.h" +#include "EtcColor.h" +#include "EtcImage.h" +#include "EtcColorFloatRGBA.h" +#include "EtcBlock4x4Encoding_RGB8.h" +#include "EtcBlock4x4Encoding_RGBA8.h" +#include "EtcBlock4x4Encoding_RGB8A1.h" +#include "EtcBlock4x4Encoding_R11.h" +#include "EtcBlock4x4Encoding_RG11.h" + +#include +#include +#include + +namespace Etc +{ + // ETC pixels are scanned vertically. + // this mapping is for when someone wants to scan the ETC pixels horizontally + const unsigned int Block4x4::s_auiPixelOrderHScan[PIXELS] = { 0, 4, 8, 12, 1, 5, 9, 13, 2, 6, 10, 14, 3, 7, 11, 15 }; + + // ---------------------------------------------------------------------------------------------------- + // + Block4x4::Block4x4(void) + { + m_pimageSource = nullptr; + m_uiSourceH = 0; + m_uiSourceV = 0; + + m_sourcealphamix = SourceAlphaMix::UNKNOWN; + m_boolBorderPixels = false; + m_boolPunchThroughPixels = false; + + m_pencoding = nullptr; + + m_errormetric = ErrorMetric::NUMERIC; + + } + Block4x4::~Block4x4() + { + m_pimageSource = nullptr; + if (m_pencoding) + { + delete m_pencoding; + m_pencoding = nullptr; + } + } + // ---------------------------------------------------------------------------------------------------- + // initialization prior to encoding from a source image + // [a_uiSourceH,a_uiSourceV] is the location of the block in a_pimageSource + // a_paucEncodingBits is the place to store the final encoding + // a_errormetric is used for finding the best encoding + // + void Block4x4::InitFromSource(Image *a_pimageSource, + unsigned int a_uiSourceH, unsigned int a_uiSourceV, + unsigned char *a_paucEncodingBits, + ErrorMetric a_errormetric) + { + + Block4x4(); + + m_pimageSource = a_pimageSource; + m_uiSourceH = a_uiSourceH; + m_uiSourceV = a_uiSourceV; + m_errormetric = a_errormetric; + + SetSourcePixels(); + + // set block encoder function + switch (m_pimageSource->GetFormat()) + { + case Image::Format::ETC1: + m_pencoding = new Block4x4Encoding_ETC1; + break; + + case Image::Format::RGB8: + case Image::Format::SRGB8: + m_pencoding = new Block4x4Encoding_RGB8; + break; + + case Image::Format::RGBA8: + case Image::Format::SRGBA8: + switch (m_sourcealphamix) + { + case SourceAlphaMix::OPAQUE: + m_pencoding = new Block4x4Encoding_RGBA8_Opaque; + break; + + case SourceAlphaMix::TRANSPARENT: + m_pencoding = new Block4x4Encoding_RGBA8_Transparent; + break; + + case SourceAlphaMix::TRANSLUCENT: + m_pencoding = new Block4x4Encoding_RGBA8; + break; + + default: + assert(0); + break; + } + break; + + case Image::Format::RGB8A1: + case Image::Format::SRGB8A1: + switch (m_sourcealphamix) + { + case SourceAlphaMix::OPAQUE: + m_pencoding = new Block4x4Encoding_RGB8A1_Opaque; + break; + + case SourceAlphaMix::TRANSPARENT: + m_pencoding = new Block4x4Encoding_RGB8A1_Transparent; + break; + + case SourceAlphaMix::TRANSLUCENT: + if (m_boolPunchThroughPixels) + { + m_pencoding = new Block4x4Encoding_RGB8A1; + } + else + { + m_pencoding = new Block4x4Encoding_RGB8A1_Opaque; + } + break; + + default: + assert(0); + break; + } + break; + + case Image::Format::R11: + case Image::Format::SIGNED_R11: + m_pencoding = new Block4x4Encoding_R11; + break; + case Image::Format::RG11: + case Image::Format::SIGNED_RG11: + m_pencoding = new Block4x4Encoding_RG11; + break; + default: + assert(0); + break; + } + + m_pencoding->InitFromSource(this, m_afrgbaSource, + a_paucEncodingBits, a_errormetric); + + } + + // ---------------------------------------------------------------------------------------------------- + // initialization of encoding state from a prior encoding using encoding bits + // [a_uiSourceH,a_uiSourceV] is the location of the block in a_pimageSource + // a_paucEncodingBits is the place to read the prior encoding + // a_imageformat is used to determine how to interpret a_paucEncodingBits + // a_errormetric was used for the prior encoding + // + void Block4x4::InitFromEtcEncodingBits(Image::Format a_imageformat, + unsigned int a_uiSourceH, unsigned int a_uiSourceV, + unsigned char *a_paucEncodingBits, + Image *a_pimageSource, + ErrorMetric a_errormetric) + { + Block4x4(); + + m_pimageSource = a_pimageSource; + m_uiSourceH = a_uiSourceH; + m_uiSourceV = a_uiSourceV; + m_errormetric = a_errormetric; + + SetSourcePixels(); + + // set block encoder function + switch (a_imageformat) + { + case Image::Format::ETC1: + m_pencoding = new Block4x4Encoding_ETC1; + break; + + case Image::Format::RGB8: + case Image::Format::SRGB8: + m_pencoding = new Block4x4Encoding_RGB8; + break; + + case Image::Format::RGBA8: + case Image::Format::SRGBA8: + m_pencoding = new Block4x4Encoding_RGBA8; + break; + + case Image::Format::RGB8A1: + case Image::Format::SRGB8A1: + m_pencoding = new Block4x4Encoding_RGB8A1; + break; + + case Image::Format::R11: + case Image::Format::SIGNED_R11: + m_pencoding = new Block4x4Encoding_R11; + break; + case Image::Format::RG11: + case Image::Format::SIGNED_RG11: + m_pencoding = new Block4x4Encoding_RG11; + break; + default: + assert(0); + break; + } + + m_pencoding->InitFromEncodingBits(this, a_paucEncodingBits, m_afrgbaSource, + m_pimageSource->GetErrorMetric()); + + } + + // ---------------------------------------------------------------------------------------------------- + // set source pixels from m_pimageSource + // set m_alphamix + // + void Block4x4::SetSourcePixels(void) + { + + Image::Format imageformat = m_pimageSource->GetFormat(); + + // alpha census + unsigned int uiTransparentSourcePixels = 0; + unsigned int uiOpaqueSourcePixels = 0; + + // copy source to consecutive memory locations + // convert from image horizontal scan to block vertical scan + unsigned int uiPixel = 0; + for (unsigned int uiBlockPixelH = 0; uiBlockPixelH < Block4x4::COLUMNS; uiBlockPixelH++) + { + unsigned int uiSourcePixelH = m_uiSourceH + uiBlockPixelH; + + for (unsigned int uiBlockPixelV = 0; uiBlockPixelV < Block4x4::ROWS; uiBlockPixelV++) + { + unsigned int uiSourcePixelV = m_uiSourceV + uiBlockPixelV; + + ColorFloatRGBA *pfrgbaSource = m_pimageSource->GetSourcePixel(uiSourcePixelH, uiSourcePixelV); + + // if pixel extends beyond source image because of block padding + if (pfrgbaSource == nullptr) + { + m_afrgbaSource[uiPixel] = ColorFloatRGBA(0.0f, 0.0f, 0.0f, NAN); // denotes border pixel + m_boolBorderPixels = true; + uiTransparentSourcePixels++; + } + else + { + //get teh current pixel data, and store some of the attributes + //before capping values to fit the encoder type + + m_afrgbaSource[uiPixel] = (*pfrgbaSource).ClampRGBA(); + + if (m_afrgbaSource[uiPixel].fA == 1.0f) + { + m_pimageSource->m_iNumOpaquePixels++; + } + else if (m_afrgbaSource[uiPixel].fA == 0.0f) + { + m_pimageSource->m_iNumTransparentPixels++; + } + else if(m_afrgbaSource[uiPixel].fA > 0.0f && m_afrgbaSource[uiPixel].fA < 1.0f) + { + m_pimageSource->m_iNumTranslucentPixels++; + } + else + { + m_pimageSource->m_numOutOfRangeValues.fA++; + } + + if (m_afrgbaSource[uiPixel].fR != 0.0f) + { + m_pimageSource->m_numColorValues.fR++; + //make sure we are getting a float between 0-1 + if (m_afrgbaSource[uiPixel].fR - 1.0f > 0.0f) + { + m_pimageSource->m_numOutOfRangeValues.fR++; + } + } + + if (m_afrgbaSource[uiPixel].fG != 0.0f) + { + m_pimageSource->m_numColorValues.fG++; + if (m_afrgbaSource[uiPixel].fG - 1.0f > 0.0f) + { + m_pimageSource->m_numOutOfRangeValues.fG++; + } + } + if (m_afrgbaSource[uiPixel].fB != 0.0f) + { + m_pimageSource->m_numColorValues.fB++; + if (m_afrgbaSource[uiPixel].fB - 1.0f > 0.0f) + { + m_pimageSource->m_numOutOfRangeValues.fB++; + } + } + // for formats with no alpha, set source alpha to 1 + if (imageformat == Image::Format::ETC1 || + imageformat == Image::Format::RGB8 || + imageformat == Image::Format::SRGB8) + { + m_afrgbaSource[uiPixel].fA = 1.0f; + } + + if (imageformat == Image::Format::R11 || + imageformat == Image::Format::SIGNED_R11) + { + m_afrgbaSource[uiPixel].fA = 1.0f; + m_afrgbaSource[uiPixel].fG = 0.0f; + m_afrgbaSource[uiPixel].fB = 0.0f; + } + + if (imageformat == Image::Format::RG11 || + imageformat == Image::Format::SIGNED_RG11) + { + m_afrgbaSource[uiPixel].fA = 1.0f; + m_afrgbaSource[uiPixel].fB = 0.0f; + } + + + // for RGB8A1, set source alpha to 0.0 or 1.0 + // set punch through flag + if (imageformat == Image::Format::RGB8A1 || + imageformat == Image::Format::SRGB8A1) + { + if (m_afrgbaSource[uiPixel].fA >= 0.5f) + { + m_afrgbaSource[uiPixel].fA = 1.0f; + } + else + { + m_afrgbaSource[uiPixel].fA = 0.0f; + m_boolPunchThroughPixels = true; + } + } + + if (m_afrgbaSource[uiPixel].fA == 1.0f) + { + uiOpaqueSourcePixels++; + } + else if (m_afrgbaSource[uiPixel].fA == 0.0f) + { + uiTransparentSourcePixels++; + } + + } + + uiPixel += 1; + } + } + + if (uiOpaqueSourcePixels == PIXELS) + { + m_sourcealphamix = SourceAlphaMix::OPAQUE; + } + else if (uiTransparentSourcePixels == PIXELS) + { + m_sourcealphamix = SourceAlphaMix::TRANSPARENT; + } + else + { + m_sourcealphamix = SourceAlphaMix::TRANSLUCENT; + } + + } + + // ---------------------------------------------------------------------------------------------------- + // return a name for the encoding mode + // + const char * Block4x4::GetEncodingModeName(void) + { + + switch (m_pencoding->GetMode()) + { + case Block4x4Encoding::MODE_ETC1: + return "ETC1"; + case Block4x4Encoding::MODE_T: + return "T"; + case Block4x4Encoding::MODE_H: + return "H"; + case Block4x4Encoding::MODE_PLANAR: + return "PLANAR"; + default: + return "???"; + } + } + + // ---------------------------------------------------------------------------------------------------- + // + +} diff --git a/EtcLib/EtcCodec/EtcBlock4x4.h b/EtcLib/EtcCodec/EtcBlock4x4.h new file mode 100644 index 0000000..7716beb --- /dev/null +++ b/EtcLib/EtcCodec/EtcBlock4x4.h @@ -0,0 +1,172 @@ +/* + * Copyright 2015 The Etc2Comp Authors. + * + * Licensed under the Apache License, Version 2.0 (the "License"); + * you may not use this file except in compliance with the License. + * You may obtain a copy of the License at + * + * http://www.apache.org/licenses/LICENSE-2.0 + * + * Unless required by applicable law or agreed to in writing, software + * distributed under the License is distributed on an "AS IS" BASIS, + * WITHOUT WARRANTIES OR CONDITIONS OF ANY KIND, either express or implied. + * See the License for the specific language governing permissions and + * limitations under the License. + */ + +#pragma once + +#include "EtcColor.h" +#include "EtcColorFloatRGBA.h" +#include "EtcErrorMetric.h" +#include "EtcImage.h" +#include "EtcBlock4x4Encoding.h" + +namespace Etc +{ + class Block4x4EncodingBits; + + class Block4x4 + { + public: + + static const unsigned int ROWS = 4; + static const unsigned int COLUMNS = 4; + static const unsigned int PIXELS = ROWS * COLUMNS; + + // the alpha mix for a 4x4 block of pixels + enum class SourceAlphaMix + { + UNKNOWN, + // + OPAQUE, // all 1.0 + TRANSPARENT, // all 0.0 or NAN + TRANSLUCENT // not all opaque or transparent + }; + + typedef void (Block4x4::*EncoderFunctionPtr)(void); + + Block4x4(void); + ~Block4x4(); + void InitFromSource(Image *a_pimageSource, + unsigned int a_uiSourceH, + unsigned int a_uiSourceV, + unsigned char *a_paucEncodingBits, + ErrorMetric a_errormetric); + + void InitFromEtcEncodingBits(Image::Format a_imageformat, + unsigned int a_uiSourceH, + unsigned int a_uiSourceV, + unsigned char *a_paucEncodingBits, + Image *a_pimageSource, + ErrorMetric a_errormetric); + + // return true if final iteration was performed + inline void PerformEncodingIteration(float a_fEffort) + { + m_pencoding->PerformIteration(a_fEffort); + } + + inline void SetEncodingBitsFromEncoding(void) + { + m_pencoding->SetEncodingBits(); + } + + inline unsigned int GetSourceH(void) + { + return m_uiSourceH; + } + + inline unsigned int GetSourceV(void) + { + return m_uiSourceV; + } + + inline float GetError(void) + { + return m_pencoding->GetError(); + } + + static const unsigned int s_auiPixelOrderHScan[PIXELS]; + + inline ColorFloatRGBA * GetDecodedColors(void) + { + return m_pencoding->GetDecodedColors(); + } + + inline float * GetDecodedAlphas(void) + { + return m_pencoding->GetDecodedAlphas(); + } + + inline Block4x4Encoding::Mode GetEncodingMode(void) + { + return m_pencoding->GetMode(); + } + + inline bool GetFlip(void) + { + return m_pencoding->GetFlip(); + } + + inline bool IsDifferential(void) + { + return m_pencoding->IsDifferential(); + } + + inline ColorFloatRGBA * GetSource() + { + return m_afrgbaSource; + } + + inline ErrorMetric GetErrorMetric() + { + return m_errormetric; + } + + const char * GetEncodingModeName(void); + + inline Block4x4Encoding * GetEncoding(void) + { + return m_pencoding; + } + + inline SourceAlphaMix GetSourceAlphaMix(void) + { + return m_sourcealphamix; + } + + inline Image * GetImageSource(void) + { + return m_pimageSource; + } + + inline bool HasBorderPixels(void) + { + return m_boolBorderPixels; + } + + inline bool HasPunchThroughPixels(void) + { + return m_boolPunchThroughPixels; + } + + private: + + void SetSourcePixels(void); + + Image *m_pimageSource; + unsigned int m_uiSourceH; + unsigned int m_uiSourceV; + ErrorMetric m_errormetric; + ColorFloatRGBA m_afrgbaSource[PIXELS]; // vertical scan + + SourceAlphaMix m_sourcealphamix; + bool m_boolBorderPixels; // marked as rgba(NAN, NAN, NAN, NAN) + bool m_boolPunchThroughPixels; // RGB8A1 or SRGB8A1 with any pixels with alpha < 0.5 + + Block4x4Encoding *m_pencoding; + + }; + +} // namespace Etc diff --git a/EtcLib/EtcCodec/EtcBlock4x4Encoding.cpp b/EtcLib/EtcCodec/EtcBlock4x4Encoding.cpp new file mode 100644 index 0000000..0ccea2a --- /dev/null +++ b/EtcLib/EtcCodec/EtcBlock4x4Encoding.cpp @@ -0,0 +1,246 @@ +/* + * Copyright 2015 The Etc2Comp Authors. + * + * Licensed under the Apache License, Version 2.0 (the "License"); + * you may not use this file except in compliance with the License. + * You may obtain a copy of the License at + * + * http://www.apache.org/licenses/LICENSE-2.0 + * + * Unless required by applicable law or agreed to in writing, software + * distributed under the License is distributed on an "AS IS" BASIS, + * WITHOUT WARRANTIES OR CONDITIONS OF ANY KIND, either express or implied. + * See the License for the specific language governing permissions and + * limitations under the License. + */ + +/* +EtcBlock4x4Encoding.cpp + +Block4x4Encoding is the abstract base class for the different encoders. Each encoder targets a +particular file format (e.g. ETC1, RGB8, RGBA8, R11) + +*/ + +#include "EtcConfig.h" +#include "EtcBlock4x4Encoding.h" + +#include "EtcBlock4x4EncodingBits.h" +#include "EtcBlock4x4.h" + +#include +#include +#include + +namespace Etc +{ + + // ---------------------------------------------------------------------------------------------------- + // + Block4x4Encoding::Block4x4Encoding(void) + { + + m_pblockParent = nullptr; + + m_pafrgbaSource = nullptr; + + m_boolBorderPixels = false; + + m_fError = -1.0f; + + m_mode = MODE_UNKNOWN; + + m_uiEncodingIterations = 0; + m_boolDone = false; + + for (unsigned int uiPixel = 0; uiPixel < PIXELS; uiPixel++) + { + m_afrgbaDecodedColors[uiPixel] = ColorFloatRGBA(-1.0f, -1.0f, -1.0f, -1.0f); + m_afDecodedAlphas[uiPixel] = -1.0f; + } + + } + + // ---------------------------------------------------------------------------------------------------- + // initialize the generic encoding for a 4x4 block + // a_pblockParent points to the block associated with this encoding + // a_errormetric is used to choose the best encoding + // init the decoded pixels to -1 to mark them as undefined + // init the error to -1 to mark it as undefined + // + void Block4x4Encoding::Init(Block4x4 *a_pblockParent, + ColorFloatRGBA *a_pafrgbaSource, + ErrorMetric a_errormetric) + { + + m_pblockParent = a_pblockParent; + + m_pafrgbaSource = a_pafrgbaSource; + + m_boolBorderPixels = m_pblockParent->HasBorderPixels(); + + m_fError = -1.0f; + + m_uiEncodingIterations = 0; + + m_errormetric = a_errormetric; + + for (unsigned int uiPixel = 0; uiPixel < PIXELS; uiPixel++) + { + m_afrgbaDecodedColors[uiPixel] = ColorFloatRGBA(-1.0f, -1.0f, -1.0f, -1.0f); + m_afDecodedAlphas[uiPixel] = -1.0f; + } + + } + + // ---------------------------------------------------------------------------------------------------- + // calculate the error for the block by summing the pixel errors + // + void Block4x4Encoding::CalcBlockError(void) + { + m_fError = 0.0f; + + for (unsigned int uiPixel = 0; uiPixel < PIXELS; uiPixel++) + { + m_fError += CalcPixelError(m_afrgbaDecodedColors[uiPixel], m_afDecodedAlphas[uiPixel], + m_pafrgbaSource[uiPixel]); + } + + } + + // ---------------------------------------------------------------------------------------------------- + // calculate the error between the source pixel and the decoded pixel + // the error amount is base on the error metric + // + float Block4x4Encoding::CalcPixelError(ColorFloatRGBA a_frgbaDecodedColor, float a_fDecodedAlpha, + ColorFloatRGBA a_frgbaSourcePixel) + { + + // if a border pixel + if (isnan(a_frgbaSourcePixel.fA)) + { + return 0.0f; + } + + if (m_errormetric == ErrorMetric::RGBA) + { + assert(a_fDecodedAlpha >= 0.0f); + + float fDRed = (a_fDecodedAlpha * a_frgbaDecodedColor.fR) - + (a_frgbaSourcePixel.fA * a_frgbaSourcePixel.fR); + float fDGreen = (a_fDecodedAlpha * a_frgbaDecodedColor.fG) - + (a_frgbaSourcePixel.fA * a_frgbaSourcePixel.fG); + float fDBlue = (a_fDecodedAlpha * a_frgbaDecodedColor.fB) - + (a_frgbaSourcePixel.fA * a_frgbaSourcePixel.fB); + + float fDAlpha = a_fDecodedAlpha - a_frgbaSourcePixel.fA; + + return fDRed*fDRed + fDGreen*fDGreen + fDBlue*fDBlue + fDAlpha*fDAlpha; + } + else if (m_errormetric == ErrorMetric::REC709) + { + assert(a_fDecodedAlpha >= 0.0f); + + float fLuma1 = a_frgbaSourcePixel.fR*0.2126f + a_frgbaSourcePixel.fG*0.7152f + a_frgbaSourcePixel.fB*0.0722f; + float fChromaR1 = 0.5f * ((a_frgbaSourcePixel.fR - fLuma1) * (1.0f / (1.0f - 0.2126f))); + float fChromaB1 = 0.5f * ((a_frgbaSourcePixel.fB - fLuma1) * (1.0f / (1.0f - 0.0722f))); + + float fLuma2 = a_frgbaDecodedColor.fR*0.2126f + + a_frgbaDecodedColor.fG*0.7152f + + a_frgbaDecodedColor.fB*0.0722f; + float fChromaR2 = 0.5f * ((a_frgbaDecodedColor.fR - fLuma2) * (1.0f / (1.0f - 0.2126f))); + float fChromaB2 = 0.5f * ((a_frgbaDecodedColor.fB - fLuma2) * (1.0f / (1.0f - 0.0722f))); + + float fDeltaL = a_frgbaSourcePixel.fA * fLuma1 - a_fDecodedAlpha * fLuma2; + float fDeltaCr = a_frgbaSourcePixel.fA * fChromaR1 - a_fDecodedAlpha * fChromaR2; + float fDeltaCb = a_frgbaSourcePixel.fA * fChromaB1 - a_fDecodedAlpha * fChromaB2; + + float fDAlpha = a_fDecodedAlpha - a_frgbaSourcePixel.fA; + + // Favor Luma accuracy over Chroma, and Red over Blue + return LUMA_WEIGHT*fDeltaL*fDeltaL + + fDeltaCr*fDeltaCr + + CHROMA_BLUE_WEIGHT*fDeltaCb*fDeltaCb + + fDAlpha*fDAlpha; + #if 0 + float fDRed = a_frgbaDecodedPixel.fR - a_frgbaSourcePixel.fR; + float fDGreen = a_frgbaDecodedPixel.fG - a_frgbaSourcePixel.fG; + float fDBlue = a_frgbaDecodedPixel.fB - a_frgbaSourcePixel.fB; + return 2.0f * 3.0f * fDeltaL * fDeltaL + fDRed*fDRed + fDGreen*fDGreen + fDBlue*fDBlue; +#endif + } + else if (m_errormetric == ErrorMetric::NORMALXYZ) + { + float fDecodedX = 2.0f * a_frgbaDecodedColor.fR - 1.0f; + float fDecodedY = 2.0f * a_frgbaDecodedColor.fG - 1.0f; + float fDecodedZ = 2.0f * a_frgbaDecodedColor.fB - 1.0f; + + float fDecodedLength = sqrtf(fDecodedX*fDecodedX + fDecodedY*fDecodedY + fDecodedZ*fDecodedZ); + + if (fDecodedLength < 0.5f) + { + return 1.0f; + } + else if (fDecodedLength == 0.0f) + { + fDecodedX = 1.0f; + fDecodedY = 0.0f; + fDecodedZ = 0.0f; + } + else + { + fDecodedX /= fDecodedLength; + fDecodedY /= fDecodedLength; + fDecodedZ /= fDecodedLength; + } + + float fSourceX = 2.0f * a_frgbaSourcePixel.fR - 1.0f; + float fSourceY = 2.0f * a_frgbaSourcePixel.fG - 1.0f; + float fSourceZ = 2.0f * a_frgbaSourcePixel.fB - 1.0f; + + float fSourceLength = sqrtf(fSourceX*fSourceX + fSourceY*fSourceY + fSourceZ*fSourceZ); + + if (fSourceLength == 0.0f) + { + fSourceX = 1.0f; + fSourceY = 0.0f; + fSourceZ = 0.0f; + } + else + { + fSourceX /= fSourceLength; + fSourceY /= fSourceLength; + fSourceZ /= fSourceLength; + } + + float fDotProduct = fSourceX*fDecodedX + fSourceY*fDecodedY + fSourceZ*fDecodedZ; + float fNormalizedDotProduct = 1.0f - 0.5f * (fDotProduct + 1.0f); + float fDotProductError = fNormalizedDotProduct * fNormalizedDotProduct; + + float fLength2 = fDecodedX*fDecodedX + fDecodedY*fDecodedY + fDecodedZ*fDecodedZ; + float fLength2Error = fabsf(1.0f - fLength2); + + float fDeltaW = a_frgbaDecodedColor.fA - a_frgbaSourcePixel.fA; + float fErrorW = fDeltaW * fDeltaW; + + return fDotProductError + fLength2Error + fErrorW; + } + else // ErrorMetric::NUMERIC + { + assert(a_fDecodedAlpha >= 0.0f); + + float fDX = a_frgbaDecodedColor.fR - a_frgbaSourcePixel.fR; + float fDY = a_frgbaDecodedColor.fG - a_frgbaSourcePixel.fG; + float fDZ = a_frgbaDecodedColor.fB - a_frgbaSourcePixel.fB; + float fDW = a_frgbaDecodedColor.fA - a_frgbaSourcePixel.fA; + + return fDX*fDX + fDY*fDY + fDZ*fDZ + fDW*fDW; + } + + } + + // ---------------------------------------------------------------------------------------------------- + // + +} // namespace Etc + diff --git a/EtcLib/EtcCodec/EtcBlock4x4Encoding.h b/EtcLib/EtcCodec/EtcBlock4x4Encoding.h new file mode 100644 index 0000000..d1fad23 --- /dev/null +++ b/EtcLib/EtcCodec/EtcBlock4x4Encoding.h @@ -0,0 +1,148 @@ +/* + * Copyright 2015 The Etc2Comp Authors. + * + * Licensed under the Apache License, Version 2.0 (the "License"); + * you may not use this file except in compliance with the License. + * You may obtain a copy of the License at + * + * http://www.apache.org/licenses/LICENSE-2.0 + * + * Unless required by applicable law or agreed to in writing, software + * distributed under the License is distributed on an "AS IS" BASIS, + * WITHOUT WARRANTIES OR CONDITIONS OF ANY KIND, either express or implied. + * See the License for the specific language governing permissions and + * limitations under the License. + */ + +#pragma once + +#include "EtcColorFloatRGBA.h" + +#include "EtcErrorMetric.h" + +#include +#include + +namespace Etc +{ + class Block4x4; + + // abstract base class for specific encodings + class Block4x4Encoding + { + public: + + static const unsigned int ROWS = 4; + static const unsigned int COLUMNS = 4; + static const unsigned int PIXELS = ROWS * COLUMNS; + static constexpr float LUMA_WEIGHT = 3.0f; + static constexpr float CHROMA_BLUE_WEIGHT = 0.5f; + + typedef enum + { + MODE_UNKNOWN, + // + MODE_ETC1, + MODE_T, + MODE_H, + MODE_PLANAR, + MODE_R11, + MODE_RG11, + // + MODES + } Mode; + + Block4x4Encoding(void); + //virtual ~Block4x4Encoding(void) =0; + virtual ~Block4x4Encoding(void) {} + virtual void InitFromSource(Block4x4 *a_pblockParent, + ColorFloatRGBA *a_pafrgbaSource, + + unsigned char *a_paucEncodingBits, ErrorMetric a_errormetric) = 0; + + virtual void InitFromEncodingBits(Block4x4 *a_pblockParent, + unsigned char *a_paucEncodingBits, + ColorFloatRGBA *a_pafrgbaSource, + + ErrorMetric a_errormetric) = 0; + + // perform an iteration of the encoding + // the first iteration must generate a complete, valid (if poor) encoding + virtual void PerformIteration(float a_fEffort) = 0; + + void CalcBlockError(void); + + inline float GetError(void) + { + assert(m_fError >= 0.0f); + + return m_fError; + } + + inline ColorFloatRGBA * GetDecodedColors(void) + { + return m_afrgbaDecodedColors; + } + + inline float * GetDecodedAlphas(void) + { + return m_afDecodedAlphas; + } + + virtual void SetEncodingBits(void) = 0; + + virtual bool GetFlip(void) = 0; + + virtual bool IsDifferential(void) = 0; + + virtual bool HasSeverelyBentDifferentialColors(void) const = 0; + + inline Mode GetMode(void) + { + return m_mode; + } + + inline bool IsDone(void) + { + return m_boolDone; + } + + inline void SetDoneIfPerfect() + { + if (GetError() == 0.0f) + { + m_boolDone = true; + } + } + + float CalcPixelError(ColorFloatRGBA a_frgbaDecodedColor, float a_fDecodedAlpha, + ColorFloatRGBA a_frgbaSourcePixel); + + protected: + + void Init(Block4x4 *a_pblockParent, + ColorFloatRGBA *a_pafrgbaSource, + + ErrorMetric a_errormetric); + + Block4x4 *m_pblockParent; + ColorFloatRGBA *m_pafrgbaSource; + + bool m_boolBorderPixels; // if block has any border pixels + + ColorFloatRGBA m_afrgbaDecodedColors[PIXELS]; // decoded RGB components, ignore Alpha + float m_afDecodedAlphas[PIXELS]; // decoded alpha component + float m_fError; // error for RGBA relative to m_pafrgbaSource + + // intermediate encoding + Mode m_mode; + + unsigned int m_uiEncodingIterations; + bool m_boolDone; // all iterations have been done + ErrorMetric m_errormetric; + + private: + + }; + +} // namespace Etc diff --git a/EtcLib/EtcCodec/EtcBlock4x4EncodingBits.h b/EtcLib/EtcCodec/EtcBlock4x4EncodingBits.h new file mode 100644 index 0000000..5ba879e --- /dev/null +++ b/EtcLib/EtcCodec/EtcBlock4x4EncodingBits.h @@ -0,0 +1,315 @@ +/* + * Copyright 2015 The Etc2Comp Authors. + * + * Licensed under the Apache License, Version 2.0 (the "License"); + * you may not use this file except in compliance with the License. + * You may obtain a copy of the License at + * + * http://www.apache.org/licenses/LICENSE-2.0 + * + * Unless required by applicable law or agreed to in writing, software + * distributed under the License is distributed on an "AS IS" BASIS, + * WITHOUT WARRANTIES OR CONDITIONS OF ANY KIND, either express or implied. + * See the License for the specific language governing permissions and + * limitations under the License. + */ + +#pragma once + +#include + +namespace Etc +{ + + // ################################################################################ + // Block4x4EncodingBits + // Base class for Block4x4EncodingBits_XXXX + // ################################################################################ + + class Block4x4EncodingBits + { + public: + + enum class Format + { + UNKNOWN, + // + RGB8, + RGBA8, + R11, + RG11, + RGB8A1, + // + FORMATS + }; + + static unsigned int GetBytesPerBlock(Format a_format) + { + switch (a_format) + { + case Format::RGB8: + case Format::R11: + case Format::RGB8A1: + return 8; + break; + + case Format::RGBA8: + case Format::RG11: + return 16; + break; + + default: + return 0; + break; + } + + } + + }; + + // ################################################################################ + // Block4x4EncodingBits_RGB8 + // Encoding bits for the RGB portion of ETC1, RGB8, RGB8A1 and RGBA8 + // ################################################################################ + + class Block4x4EncodingBits_RGB8 + { + public: + + static const unsigned int BYTES_PER_BLOCK = 8; + + inline Block4x4EncodingBits_RGB8(void) + { + assert(sizeof(Block4x4EncodingBits_RGB8) == BYTES_PER_BLOCK); + + for (unsigned int uiByte = 0; uiByte < BYTES_PER_BLOCK; uiByte++) + { + auc[uiByte] = 0; + } + + } + + typedef struct + { + unsigned red2 : 4; + unsigned red1 : 4; + // + unsigned green2 : 4; + unsigned green1 : 4; + // + unsigned blue2 : 4; + unsigned blue1 : 4; + // + unsigned flip : 1; + unsigned diff : 1; + unsigned cw2 : 3; + unsigned cw1 : 3; + // + unsigned int selectors; + } Individual; + + typedef struct + { + signed dred2 : 3; + unsigned red1 : 5; + // + signed dgreen2 : 3; + unsigned green1 : 5; + // + signed dblue2 : 3; + unsigned blue1 : 5; + // + unsigned flip : 1; + unsigned diff : 1; + unsigned cw2 : 3; + unsigned cw1 : 3; + // + unsigned int selectors; + } Differential; + + typedef struct + { + unsigned red1b : 2; + unsigned detect2 : 1; + unsigned red1a : 2; + unsigned detect1 : 3; + // + unsigned blue1 : 4; + unsigned green1 : 4; + // + unsigned green2 : 4; + unsigned red2 : 4; + // + unsigned db : 1; + unsigned diff : 1; + unsigned da : 2; + unsigned blue2 : 4; + // + unsigned int selectors; + } T; + + typedef struct + { + unsigned green1a : 3; + unsigned red1 : 4; + unsigned detect1 : 1; + // + unsigned blue1b : 2; + unsigned detect3 : 1; + unsigned blue1a : 1; + unsigned green1b : 1; + unsigned detect2 : 3; + // + unsigned green2a : 3; + unsigned red2 : 4; + unsigned blue1c : 1; + // + unsigned db : 1; + unsigned diff : 1; + unsigned da : 1; + unsigned blue2 : 4; + unsigned green2b : 1; + // + unsigned int selectors; + } H; + + typedef struct + { + unsigned originGreen1 : 1; + unsigned originRed : 6; + unsigned detect1 : 1; + // + unsigned originBlue1 : 1; + unsigned originGreen2 : 6; + unsigned detect2 : 1; + // + unsigned originBlue3 : 2; + unsigned detect4 : 1; + unsigned originBlue2 : 2; + unsigned detect3 : 3; + // + unsigned horizRed2 : 1; + unsigned diff : 1; + unsigned horizRed1 : 5; + unsigned originBlue4 : 1; + // + unsigned horizBlue1: 1; + unsigned horizGreen : 7; + // + unsigned vertRed1 : 3; + unsigned horizBlue2 : 5; + // + unsigned vertGreen1 : 5; + unsigned vertRed2 : 3; + // + unsigned vertBlue : 6; + unsigned vertGreen2 : 2; + } Planar; + + union + { + unsigned char auc[BYTES_PER_BLOCK]; + unsigned long int ul; + Individual individual; + Differential differential; + T t; + H h; + Planar planar; + }; + + }; + + // ################################################################################ + // Block4x4EncodingBits_A8 + // Encoding bits for the A portion of RGBA8 + // ################################################################################ + + class Block4x4EncodingBits_A8 + { + public: + + static const unsigned int BYTES_PER_BLOCK = 8; + static const unsigned int SELECTOR_BYTES = 6; + + typedef struct + { + unsigned base : 8; + unsigned table : 4; + unsigned multiplier : 4; + unsigned selectors0 : 8; + unsigned selectors1 : 8; + unsigned selectors2 : 8; + unsigned selectors3 : 8; + unsigned selectors4 : 8; + unsigned selectors5 : 8; + } Data; + + Data data; + + }; + + // ################################################################################ + // Block4x4EncodingBits_R11 + // Encoding bits for the R portion of R11 + // ################################################################################ + + class Block4x4EncodingBits_R11 + { + public: + + static const unsigned int BYTES_PER_BLOCK = 8; + static const unsigned int SELECTOR_BYTES = 6; + + typedef struct + { + unsigned base : 8; + unsigned table : 4; + unsigned multiplier : 4; + unsigned selectors0 : 8; + unsigned selectors1 : 8; + unsigned selectors2 : 8; + unsigned selectors3 : 8; + unsigned selectors4 : 8; + unsigned selectors5 : 8; + } Data; + + Data data; + + }; + + class Block4x4EncodingBits_RG11 + { + public: + + static const unsigned int BYTES_PER_BLOCK = 16; + static const unsigned int SELECTOR_BYTES = 12; + + typedef struct + { + //Red portion + unsigned baseR : 8; + unsigned tableIndexR : 4; + unsigned multiplierR : 4; + unsigned selectorsR0 : 8; + unsigned selectorsR1 : 8; + unsigned selectorsR2 : 8; + unsigned selectorsR3 : 8; + unsigned selectorsR4 : 8; + unsigned selectorsR5 : 8; + //Green portion + unsigned baseG : 8; + unsigned tableIndexG : 4; + unsigned multiplierG : 4; + unsigned selectorsG0 : 8; + unsigned selectorsG1 : 8; + unsigned selectorsG2 : 8; + unsigned selectorsG3 : 8; + unsigned selectorsG4 : 8; + unsigned selectorsG5 : 8; + } Data; + + Data data; + + }; + +} diff --git a/EtcLib/EtcCodec/EtcBlock4x4Encoding_ETC1.cpp b/EtcLib/EtcCodec/EtcBlock4x4Encoding_ETC1.cpp new file mode 100644 index 0000000..7aa2de3 --- /dev/null +++ b/EtcLib/EtcCodec/EtcBlock4x4Encoding_ETC1.cpp @@ -0,0 +1,1280 @@ +/* + * Copyright 2015 The Etc2Comp Authors. + * + * Licensed under the Apache License, Version 2.0 (the "License"); + * you may not use this file except in compliance with the License. + * You may obtain a copy of the License at + * + * http://www.apache.org/licenses/LICENSE-2.0 + * + * Unless required by applicable law or agreed to in writing, software + * distributed under the License is distributed on an "AS IS" BASIS, + * WITHOUT WARRANTIES OR CONDITIONS OF ANY KIND, either express or implied. + * See the License for the specific language governing permissions and + * limitations under the License. + */ + +/* +EtcBlock4x4Encoding_ETC1.cpp + +Block4x4Encoding_ETC1 is the encoder to use when targetting file format ETC1. This encoder is also +used for the ETC1 subset of file format RGB8, RGBA8 and RGB8A1 + +*/ + +#include "EtcConfig.h" +#include "EtcBlock4x4Encoding_ETC1.h" + +#include "EtcBlock4x4.h" +#include "EtcBlock4x4EncodingBits.h" +#include "EtcDifferentialTrys.h" + +#include +#include +#include +#include +#include + +namespace Etc +{ + + // pixel processing order if the flip bit = 0 (horizontal split) + const unsigned int Block4x4Encoding_ETC1::s_auiPixelOrderFlip0[PIXELS] = { 0, 1, 2, 3, 4, 5, 6, 7, 8, 9, 10, 11, 12, 13, 14, 15 }; + + // pixel processing order if the flip bit = 1 (vertical split) + const unsigned int Block4x4Encoding_ETC1::s_auiPixelOrderFlip1[PIXELS] = { 0, 1, 4, 5, 8, 9, 12, 13, 2, 3, 6, 7, 10, 11, 14, 15 }; + + // pixel processing order for horizontal scan (ETC normally does a vertical scan) + const unsigned int Block4x4Encoding_ETC1::s_auiPixelOrderHScan[PIXELS] = { 0, 4, 8, 12, 1, 5, 9, 13, 2, 6, 10, 14, 3, 7, 11, 15 }; + + // pixel indices for different block halves + const unsigned int Block4x4Encoding_ETC1::s_auiLeftPixelMapping[8] = { 0, 1, 2, 3, 4, 5, 6, 7 }; + const unsigned int Block4x4Encoding_ETC1::s_auiRightPixelMapping[8] = { 8, 9, 10, 11, 12, 13, 14, 15 }; + const unsigned int Block4x4Encoding_ETC1::s_auiTopPixelMapping[8] = { 0, 1, 4, 5, 8, 9, 12, 13 }; + const unsigned int Block4x4Encoding_ETC1::s_auiBottomPixelMapping[8] = { 2, 3, 6, 7, 10, 11, 14, 15 }; + + // CW ranges that the ETC1 decoders use + // CW is basically a contrast for the different selector bits, since these values are offsets to the base color + // the first axis in the array is indexed by the CW in the encoding bits + // the second axis in the array is indexed by the selector bits + float Block4x4Encoding_ETC1::s_aafCwTable[CW_RANGES][SELECTORS] = + { + { 2.0f / 255.0f, 8.0f / 255.0f, -2.0f / 255.0f, -8.0f / 255.0f }, + { 5.0f / 255.0f, 17.0f / 255.0f, -5.0f / 255.0f, -17.0f / 255.0f }, + { 9.0f / 255.0f, 29.0f / 255.0f, -9.0f / 255.0f, -29.0f / 255.0f }, + { 13.0f / 255.0f, 42.0f / 255.0f, -13.0f / 255.0f, -42.0f / 255.0f }, + { 18.0f / 255.0f, 60.0f / 255.0f, -18.0f / 255.0f, -60.0f / 255.0f }, + { 24.0f / 255.0f, 80.0f / 255.0f, -24.0f / 255.0f, -80.0f / 255.0f }, + { 33.0f / 255.0f, 106.0f / 255.0f, -33.0f / 255.0f, -106.0f / 255.0f }, + { 47.0f / 255.0f, 183.0f / 255.0f, -47.0f / 255.0f, -183.0f / 255.0f } + }; + + // ---------------------------------------------------------------------------------------------------- + // + Block4x4Encoding_ETC1::Block4x4Encoding_ETC1(void) + { + m_mode = MODE_ETC1; + m_boolDiff = false; + m_boolFlip = false; + m_frgbaColor1 = ColorFloatRGBA(); + m_frgbaColor2 = ColorFloatRGBA(); + m_uiCW1 = 0; + m_uiCW2 = 0; + for (unsigned int uiPixel = 0; uiPixel < PIXELS; uiPixel++) + { + m_auiSelectors[uiPixel] = 0; + m_afDecodedAlphas[uiPixel] = 1.0f; + } + + m_boolMostLikelyFlip = false; + + m_fError = -1.0f; + + m_fError1 = -1.0f; + m_fError2 = -1.0f; + m_boolSeverelyBentDifferentialColors = false; + + for (unsigned int uiPixel = 0; uiPixel < PIXELS; uiPixel++) + { + m_afDecodedAlphas[uiPixel] = 1.0f; + } + + } + + Block4x4Encoding_ETC1::~Block4x4Encoding_ETC1(void) {} + + // ---------------------------------------------------------------------------------------------------- + // initialization prior to encoding + // a_pblockParent points to the block associated with this encoding + // a_errormetric is used to choose the best encoding + // a_pafrgbaSource points to a 4x4 block subset of the source image + // a_paucEncodingBits points to the final encoding bits + // + void Block4x4Encoding_ETC1::InitFromSource(Block4x4 *a_pblockParent, + ColorFloatRGBA *a_pafrgbaSource, + unsigned char *a_paucEncodingBits, ErrorMetric a_errormetric) + { + + Block4x4Encoding::Init(a_pblockParent, a_pafrgbaSource,a_errormetric); + + for (unsigned int uiPixel = 0; uiPixel < PIXELS; uiPixel++) + { + m_afDecodedAlphas[uiPixel] = 1.0f; + } + + m_fError = -1.0f; + + m_pencodingbitsRGB8 = (Block4x4EncodingBits_RGB8 *)(a_paucEncodingBits); + + } + + // ---------------------------------------------------------------------------------------------------- + // initialization from the encoding bits of a previous encoding + // a_pblockParent points to the block associated with this encoding + // a_errormetric is used to choose the best encoding + // a_pafrgbaSource points to a 4x4 block subset of the source image + // a_paucEncodingBits points to the final encoding bits of a previous encoding + // + void Block4x4Encoding_ETC1::InitFromEncodingBits(Block4x4 *a_pblockParent, + unsigned char *a_paucEncodingBits, + ColorFloatRGBA *a_pafrgbaSource, + ErrorMetric a_errormetric) + { + + Block4x4Encoding::Init(a_pblockParent, a_pafrgbaSource,a_errormetric); + m_fError = -1.0f; + + m_pencodingbitsRGB8 = (Block4x4EncodingBits_RGB8 *)a_paucEncodingBits; + + m_mode = MODE_ETC1; + m_boolDiff = m_pencodingbitsRGB8->individual.diff; + m_boolFlip = m_pencodingbitsRGB8->individual.flip; + if (m_boolDiff) + { + int iR2 = (int)(m_pencodingbitsRGB8->differential.red1 + m_pencodingbitsRGB8->differential.dred2); + if (iR2 < 0) + { + iR2 = 0; + } + else if (iR2 > 31) + { + iR2 = 31; + } + + int iG2 = (int)(m_pencodingbitsRGB8->differential.green1 + m_pencodingbitsRGB8->differential.dgreen2); + if (iG2 < 0) + { + iG2 = 0; + } + else if (iG2 > 31) + { + iG2 = 31; + } + + int iB2 = (int)(m_pencodingbitsRGB8->differential.blue1 + m_pencodingbitsRGB8->differential.dblue2); + if (iB2 < 0) + { + iB2 = 0; + } + else if (iB2 > 31) + { + iB2 = 31; + } + + m_frgbaColor1 = ColorFloatRGBA::ConvertFromRGB5(m_pencodingbitsRGB8->differential.red1, m_pencodingbitsRGB8->differential.green1, m_pencodingbitsRGB8->differential.blue1); + m_frgbaColor2 = ColorFloatRGBA::ConvertFromRGB5((unsigned char)iR2, (unsigned char)iG2, (unsigned char)iB2); + + } + else + { + m_frgbaColor1 = ColorFloatRGBA::ConvertFromRGB4(m_pencodingbitsRGB8->individual.red1, m_pencodingbitsRGB8->individual.green1, m_pencodingbitsRGB8->individual.blue1); + m_frgbaColor2 = ColorFloatRGBA::ConvertFromRGB4(m_pencodingbitsRGB8->individual.red2, m_pencodingbitsRGB8->individual.green2, m_pencodingbitsRGB8->individual.blue2); + } + + m_uiCW1 = m_pencodingbitsRGB8->individual.cw1; + m_uiCW2 = m_pencodingbitsRGB8->individual.cw2; + + InitFromEncodingBits_Selectors(); + + Decode(); + + CalcBlockError(); + } + + // ---------------------------------------------------------------------------------------------------- + // init the selectors from a prior encoding + // + void Block4x4Encoding_ETC1::InitFromEncodingBits_Selectors(void) + { + + unsigned char *paucSelectors = (unsigned char *)&m_pencodingbitsRGB8->individual.selectors; + + for (unsigned int iPixel = 0; iPixel < PIXELS; iPixel++) + { + unsigned int uiByteMSB = (unsigned int)(1 - (iPixel / 8)); + unsigned int uiByteLSB = (unsigned int)(3 - (iPixel / 8)); + unsigned int uiShift = (unsigned int)(iPixel & 7); + + unsigned int uiSelectorMSB = (unsigned int)((paucSelectors[uiByteMSB] >> uiShift) & 1); + unsigned int uiSelectorLSB = (unsigned int)((paucSelectors[uiByteLSB] >> uiShift) & 1); + + m_auiSelectors[iPixel] = (uiSelectorMSB << 1) + uiSelectorLSB; + } + + } + + // ---------------------------------------------------------------------------------------------------- + // perform a single encoding iteration + // replace the encoding if a better encoding was found + // subsequent iterations generally take longer for each iteration + // set m_boolDone if encoding is perfect or encoding is finished based on a_fEffort + // + void Block4x4Encoding_ETC1::PerformIteration(float a_fEffort) + { + assert(!m_boolDone); + + switch (m_uiEncodingIterations) + { + case 0: + PerformFirstIteration(); + break; + + case 1: + TryDifferential(m_boolMostLikelyFlip, 1, 0, 0); + break; + + case 2: + TryIndividual(m_boolMostLikelyFlip, 1); + if (a_fEffort <= 49.5f) + { + m_boolDone = true; + } + break; + + case 3: + TryDifferential(!m_boolMostLikelyFlip, 1, 0, 0); + if (a_fEffort <= 59.5f) + { + m_boolDone = true; + } + break; + + case 4: + TryIndividual(!m_boolMostLikelyFlip, 1); + if (a_fEffort <= 69.5f) + { + m_boolDone = true; + } + break; + + case 5: + TryDegenerates1(); + if (a_fEffort <= 79.5f) + { + m_boolDone = true; + } + break; + + case 6: + TryDegenerates2(); + if (a_fEffort <= 89.5f) + { + m_boolDone = true; + } + break; + + case 7: + TryDegenerates3(); + if (a_fEffort <= 99.5f) + { + m_boolDone = true; + } + break; + + case 8: + TryDegenerates4(); + m_boolDone = true; + break; + + default: + assert(0); + break; + } + + m_uiEncodingIterations++; + SetDoneIfPerfect(); + } + + // ---------------------------------------------------------------------------------------------------- + // find best initial encoding to ensure block has a valid encoding + // + void Block4x4Encoding_ETC1::PerformFirstIteration(void) + { + CalculateMostLikelyFlip(); + + m_fError = FLT_MAX; + + TryDifferential(m_boolMostLikelyFlip, 0, 0, 0); + SetDoneIfPerfect(); + if (m_boolDone) + { + return; + } + + TryIndividual(m_boolMostLikelyFlip, 0); + SetDoneIfPerfect(); + if (m_boolDone) + { + return; + } + TryDifferential(!m_boolMostLikelyFlip, 0, 0, 0); + SetDoneIfPerfect(); + if (m_boolDone) + { + return; + } + TryIndividual(!m_boolMostLikelyFlip, 0); + + } + + // ---------------------------------------------------------------------------------------------------- + // algorithm: + // create a source average color for the Left, Right, Top and Bottom halves using the 8 pixels in each half + // note: the "gray line" is the line of equal delta RGB that goes thru the average color + // for each half: + // see how close each of the 8 pixels are to the "gray line" that goes thru the source average color + // create an error value that is the sum of the distances from the gray line + // h_error is the sum of Left and Right errors + // v_error is the sum of Top and Bottom errors + // + void Block4x4Encoding_ETC1::CalculateMostLikelyFlip(void) + { + static const bool DEBUG_PRINT = false; + + CalculateSourceAverages(); + + float fLeftGrayErrorSum = 0.0f; + float fRightGrayErrorSum = 0.0f; + float fTopGrayErrorSum = 0.0f; + float fBottomGrayErrorSum = 0.0f; + + for (unsigned int uiPixel = 0; uiPixel < 8; uiPixel++) + { + ColorFloatRGBA *pfrgbaLeft = &m_pafrgbaSource[uiPixel]; + ColorFloatRGBA *pfrgbaRight = &m_pafrgbaSource[uiPixel + 8]; + ColorFloatRGBA *pfrgbaTop = &m_pafrgbaSource[s_auiTopPixelMapping[uiPixel]]; + ColorFloatRGBA *pfrgbaBottom = &m_pafrgbaSource[s_auiBottomPixelMapping[uiPixel]]; + + float fLeftGrayError = CalcGrayDistance2(*pfrgbaLeft, m_frgbaSourceAverageLeft); + float fRightGrayError = CalcGrayDistance2(*pfrgbaRight, m_frgbaSourceAverageRight); + float fTopGrayError = CalcGrayDistance2(*pfrgbaTop, m_frgbaSourceAverageTop); + float fBottomGrayError = CalcGrayDistance2(*pfrgbaBottom, m_frgbaSourceAverageBottom); + + fLeftGrayErrorSum += fLeftGrayError; + fRightGrayErrorSum += fRightGrayError; + fTopGrayErrorSum += fTopGrayError; + fBottomGrayErrorSum += fBottomGrayError; + } + + if (DEBUG_PRINT) + { + printf("\n%.2f %.2f\n", fLeftGrayErrorSum + fRightGrayErrorSum, fTopGrayErrorSum + fBottomGrayErrorSum); + } + + m_boolMostLikelyFlip = (fTopGrayErrorSum + fBottomGrayErrorSum) < (fLeftGrayErrorSum + fRightGrayErrorSum); + + } + + // ---------------------------------------------------------------------------------------------------- + // calculate source pixel averages for each 2x2 quadrant in a 4x4 block + // these are used to determine the averages for each of the 4 different halves (left, right, top, bottom) + // ignore pixels that have alpha == NAN (these are border pixels outside of the source image) + // weight the averages based on a pixel's alpha + // + void Block4x4Encoding_ETC1::CalculateSourceAverages(void) + { + static const bool DEBUG_PRINT = false; + + + if (m_pblockParent->GetSourceAlphaMix() == Block4x4::SourceAlphaMix::OPAQUE) + { + ColorFloatRGBA frgbaSumUL = m_pafrgbaSource[0] + m_pafrgbaSource[1] + m_pafrgbaSource[4] + m_pafrgbaSource[5]; + ColorFloatRGBA frgbaSumLL = m_pafrgbaSource[2] + m_pafrgbaSource[3] + m_pafrgbaSource[6] + m_pafrgbaSource[7]; + ColorFloatRGBA frgbaSumUR = m_pafrgbaSource[8] + m_pafrgbaSource[9] + m_pafrgbaSource[12] + m_pafrgbaSource[13]; + ColorFloatRGBA frgbaSumLR = m_pafrgbaSource[10] + m_pafrgbaSource[11] + m_pafrgbaSource[14] + m_pafrgbaSource[15]; + + m_frgbaSourceAverageLeft = (frgbaSumUL + frgbaSumLL) * 0.125f; + m_frgbaSourceAverageRight = (frgbaSumUR + frgbaSumLR) * 0.125f; + m_frgbaSourceAverageTop = (frgbaSumUL + frgbaSumUR) * 0.125f; + m_frgbaSourceAverageBottom = (frgbaSumLL + frgbaSumLR) * 0.125f; + } + else + { + float afSourceAlpha[PIXELS]; + + // treat alpha NAN as 0.0f + for (unsigned int uiPixel = 0; uiPixel < PIXELS; uiPixel++) + { + afSourceAlpha[uiPixel] = isnan(m_pafrgbaSource[uiPixel].fA) ? + 0.0f : + m_pafrgbaSource[uiPixel].fA; + } + + ColorFloatRGBA afrgbaAlphaWeightedSource[PIXELS]; + for (unsigned int uiPixel = 0; uiPixel < PIXELS; uiPixel++) + { + afrgbaAlphaWeightedSource[uiPixel] = m_pafrgbaSource[uiPixel] * afSourceAlpha[uiPixel]; + } + + ColorFloatRGBA frgbaSumUL = afrgbaAlphaWeightedSource[0] + + afrgbaAlphaWeightedSource[1] + + afrgbaAlphaWeightedSource[4] + + afrgbaAlphaWeightedSource[5]; + + ColorFloatRGBA frgbaSumLL = afrgbaAlphaWeightedSource[2] + + afrgbaAlphaWeightedSource[3] + + afrgbaAlphaWeightedSource[6] + + afrgbaAlphaWeightedSource[7]; + + ColorFloatRGBA frgbaSumUR = afrgbaAlphaWeightedSource[8] + + afrgbaAlphaWeightedSource[9] + + afrgbaAlphaWeightedSource[12] + + afrgbaAlphaWeightedSource[13]; + + ColorFloatRGBA frgbaSumLR = afrgbaAlphaWeightedSource[10] + + afrgbaAlphaWeightedSource[11] + + afrgbaAlphaWeightedSource[14] + + afrgbaAlphaWeightedSource[15]; + + float fWeightSumUL = afSourceAlpha[0] + + afSourceAlpha[1] + + afSourceAlpha[4] + + afSourceAlpha[5]; + + float fWeightSumLL = afSourceAlpha[2] + + afSourceAlpha[3] + + afSourceAlpha[6] + + afSourceAlpha[7]; + + float fWeightSumUR = afSourceAlpha[8] + + afSourceAlpha[9] + + afSourceAlpha[12] + + afSourceAlpha[13]; + + float fWeightSumLR = afSourceAlpha[10] + + afSourceAlpha[11] + + afSourceAlpha[14] + + afSourceAlpha[15]; + + ColorFloatRGBA frgbaSumLeft = frgbaSumUL + frgbaSumLL; + ColorFloatRGBA frgbaSumRight = frgbaSumUR + frgbaSumLR; + ColorFloatRGBA frgbaSumTop = frgbaSumUL + frgbaSumUR; + ColorFloatRGBA frgbaSumBottom = frgbaSumLL + frgbaSumLR; + + float fWeightSumLeft = fWeightSumUL + fWeightSumLL; + float fWeightSumRight = fWeightSumUR + fWeightSumLR; + float fWeightSumTop = fWeightSumUL + fWeightSumUR; + float fWeightSumBottom = fWeightSumLL + fWeightSumLR; + + // check to see if there is at least 1 pixel with non-zero alpha + // completely transparent block should not make it to this code + assert((fWeightSumLeft + fWeightSumRight) > 0.0f); + assert((fWeightSumTop + fWeightSumBottom) > 0.0f); + + if (fWeightSumLeft > 0.0f) + { + m_frgbaSourceAverageLeft = frgbaSumLeft * (1.0f/fWeightSumLeft); + } + if (fWeightSumRight > 0.0f) + { + m_frgbaSourceAverageRight = frgbaSumRight * (1.0f/fWeightSumRight); + } + if (fWeightSumTop > 0.0f) + { + m_frgbaSourceAverageTop = frgbaSumTop * (1.0f/fWeightSumTop); + } + if (fWeightSumBottom > 0.0f) + { + m_frgbaSourceAverageBottom = frgbaSumBottom * (1.0f/fWeightSumBottom); + } + + if (fWeightSumLeft == 0.0f) + { + assert(fWeightSumRight > 0.0f); + m_frgbaSourceAverageLeft = m_frgbaSourceAverageRight; + } + if (fWeightSumRight == 0.0f) + { + assert(fWeightSumLeft > 0.0f); + m_frgbaSourceAverageRight = m_frgbaSourceAverageLeft; + } + if (fWeightSumTop == 0.0f) + { + assert(fWeightSumBottom > 0.0f); + m_frgbaSourceAverageTop = m_frgbaSourceAverageBottom; + } + if (fWeightSumBottom == 0.0f) + { + assert(fWeightSumTop > 0.0f); + m_frgbaSourceAverageBottom = m_frgbaSourceAverageTop; + } + } + + + + if (DEBUG_PRINT) + { + printf("\ntarget: [%.2f,%.2f,%.2f] [%.2f,%.2f,%.2f] [%.2f,%.2f,%.2f] [%.2f,%.2f,%.2f]\n", + m_frgbaSourceAverageLeft.fR, m_frgbaSourceAverageLeft.fG, m_frgbaSourceAverageLeft.fB, + m_frgbaSourceAverageRight.fR, m_frgbaSourceAverageRight.fG, m_frgbaSourceAverageRight.fB, + m_frgbaSourceAverageTop.fR, m_frgbaSourceAverageTop.fG, m_frgbaSourceAverageTop.fB, + m_frgbaSourceAverageBottom.fR, m_frgbaSourceAverageBottom.fG, m_frgbaSourceAverageBottom.fB); + } + + } + + // ---------------------------------------------------------------------------------------------------- + // try an ETC1 differential mode encoding + // use a_boolFlip to set the encoding F bit + // use a_uiRadius to alter basecolor components in the range[-a_uiRadius:a_uiRadius] + // use a_iGrayOffset1 and a_iGrayOffset2 to offset the basecolor to search for degenerate encodings + // replace the encoding if the encoding error is less than previous encoding + // + void Block4x4Encoding_ETC1::TryDifferential(bool a_boolFlip, unsigned int a_uiRadius, + int a_iGrayOffset1, int a_iGrayOffset2) + { + + ColorFloatRGBA frgbaColor1; + ColorFloatRGBA frgbaColor2; + + const unsigned int *pauiPixelMapping1; + const unsigned int *pauiPixelMapping2; + + if (a_boolFlip) + { + frgbaColor1 = m_frgbaSourceAverageTop; + frgbaColor2 = m_frgbaSourceAverageBottom; + + pauiPixelMapping1 = s_auiTopPixelMapping; + pauiPixelMapping2 = s_auiBottomPixelMapping; + } + else + { + frgbaColor1 = m_frgbaSourceAverageLeft; + frgbaColor2 = m_frgbaSourceAverageRight; + + pauiPixelMapping1 = s_auiLeftPixelMapping; + pauiPixelMapping2 = s_auiRightPixelMapping; + } + + DifferentialTrys trys(frgbaColor1, frgbaColor2, pauiPixelMapping1, pauiPixelMapping2, + a_uiRadius, a_iGrayOffset1, a_iGrayOffset2); + + Block4x4Encoding_ETC1 encodingTry = *this; + encodingTry.m_boolFlip = a_boolFlip; + + encodingTry.TryDifferentialHalf(&trys.m_half1); + encodingTry.TryDifferentialHalf(&trys.m_half2); + + // find best halves that are within differential range + DifferentialTrys::Try *ptryBest1 = nullptr; + DifferentialTrys::Try *ptryBest2 = nullptr; + encodingTry.m_fError = FLT_MAX; + + // see if the best of each half are in differential range + int iDRed = trys.m_half2.m_ptryBest->m_iRed - trys.m_half1.m_ptryBest->m_iRed; + int iDGreen = trys.m_half2.m_ptryBest->m_iGreen - trys.m_half1.m_ptryBest->m_iGreen; + int iDBlue = trys.m_half2.m_ptryBest->m_iBlue - trys.m_half1.m_ptryBest->m_iBlue; + if (iDRed >= -4 && iDRed <= 3 && iDGreen >= -4 && iDGreen <= 3 && iDBlue >= -4 && iDBlue <= 3) + { + ptryBest1 = trys.m_half1.m_ptryBest; + ptryBest2 = trys.m_half2.m_ptryBest; + encodingTry.m_fError = trys.m_half1.m_ptryBest->m_fError + trys.m_half2.m_ptryBest->m_fError; + } + else + { + // else, find the next best halves that are in differential range + for (DifferentialTrys::Try *ptry1 = &trys.m_half1.m_atry[0]; + ptry1 < &trys.m_half1.m_atry[trys.m_half1.m_uiTrys]; + ptry1++) + { + for (DifferentialTrys::Try *ptry2 = &trys.m_half2.m_atry[0]; + ptry2 < &trys.m_half2.m_atry[trys.m_half2.m_uiTrys]; + ptry2++) + { + iDRed = ptry2->m_iRed - ptry1->m_iRed; + bool boolValidRedDelta = iDRed <= 3 && iDRed >= -4; + iDGreen = ptry2->m_iGreen - ptry1->m_iGreen; + bool boolValidGreenDelta = iDGreen <= 3 && iDGreen >= -4; + iDBlue = ptry2->m_iBlue - ptry1->m_iBlue; + bool boolValidBlueDelta = iDBlue <= 3 && iDBlue >= -4; + + if (boolValidRedDelta && boolValidGreenDelta && boolValidBlueDelta) + { + float fError = ptry1->m_fError + ptry2->m_fError; + + if (fError < encodingTry.m_fError) + { + encodingTry.m_fError = fError; + + ptryBest1 = ptry1; + ptryBest2 = ptry2; + } + } + + } + } + assert(encodingTry.m_fError < FLT_MAX); + assert(ptryBest1 != nullptr); + assert(ptryBest2 != nullptr); + } + + if (encodingTry.m_fError < m_fError) + { + m_mode = MODE_ETC1; + m_boolDiff = true; + m_boolFlip = encodingTry.m_boolFlip; + m_frgbaColor1 = ColorFloatRGBA::ConvertFromRGB5((unsigned char)ptryBest1->m_iRed, (unsigned char)ptryBest1->m_iGreen, (unsigned char)ptryBest1->m_iBlue); + m_frgbaColor2 = ColorFloatRGBA::ConvertFromRGB5((unsigned char)ptryBest2->m_iRed, (unsigned char)ptryBest2->m_iGreen, (unsigned char)ptryBest2->m_iBlue); + m_uiCW1 = ptryBest1->m_uiCW; + m_uiCW2 = ptryBest2->m_uiCW; + + for (unsigned int uiPixelOrder = 0; uiPixelOrder < PIXELS / 2; uiPixelOrder++) + { + unsigned int uiPixel1 = pauiPixelMapping1[uiPixelOrder]; + unsigned int uiPixel2 = pauiPixelMapping2[uiPixelOrder]; + + unsigned int uiSelector1 = ptryBest1->m_auiSelectors[uiPixelOrder]; + unsigned int uiSelector2 = ptryBest2->m_auiSelectors[uiPixelOrder]; + + m_auiSelectors[uiPixel1] = uiSelector1; + m_auiSelectors[uiPixel2] = ptryBest2->m_auiSelectors[uiPixelOrder]; + + float fDeltaRGB1 = s_aafCwTable[m_uiCW1][uiSelector1]; + float fDeltaRGB2 = s_aafCwTable[m_uiCW2][uiSelector2]; + + m_afrgbaDecodedColors[uiPixel1] = (m_frgbaColor1 + fDeltaRGB1).ClampRGB(); + m_afrgbaDecodedColors[uiPixel2] = (m_frgbaColor2 + fDeltaRGB2).ClampRGB(); + } + + m_fError1 = ptryBest1->m_fError; + m_fError2 = ptryBest2->m_fError; + m_boolSeverelyBentDifferentialColors = trys.m_boolSeverelyBentColors; + m_fError = m_fError1 + m_fError2; + + // sanity check + { + int iRed1 = m_frgbaColor1.IntRed(31.0f); + int iGreen1 = m_frgbaColor1.IntGreen(31.0f); + int iBlue1 = m_frgbaColor1.IntBlue(31.0f); + + int iRed2 = m_frgbaColor2.IntRed(31.0f); + int iGreen2 = m_frgbaColor2.IntGreen(31.0f); + int iBlue2 = m_frgbaColor2.IntBlue(31.0f); + + iDRed = iRed2 - iRed1; + iDGreen = iGreen2 - iGreen1; + iDBlue = iBlue2 - iBlue1; + + assert(iDRed >= -4 && iDRed < 4); + assert(iDGreen >= -4 && iDGreen < 4); + assert(iDBlue >= -4 && iDBlue < 4); + } + } + + } + + // ---------------------------------------------------------------------------------------------------- + // try an ETC1 differential mode encoding for a half of a 4x4 block + // vary the basecolor components using a radius + // + void Block4x4Encoding_ETC1::TryDifferentialHalf(DifferentialTrys::Half *a_phalf) + { + + a_phalf->m_ptryBest = nullptr; + float fBestTryError = FLT_MAX; + + a_phalf->m_uiTrys = 0; + for (int iRed = a_phalf->m_iRed - (int)a_phalf->m_uiRadius; + iRed <= a_phalf->m_iRed + (int)a_phalf->m_uiRadius; + iRed++) + { + assert(iRed >= 0 && iRed <= 31); + + for (int iGreen = a_phalf->m_iGreen - (int)a_phalf->m_uiRadius; + iGreen <= a_phalf->m_iGreen + (int)a_phalf->m_uiRadius; + iGreen++) + { + assert(iGreen >= 0 && iGreen <= 31); + + for (int iBlue = a_phalf->m_iBlue - (int)a_phalf->m_uiRadius; + iBlue <= a_phalf->m_iBlue + (int)a_phalf->m_uiRadius; + iBlue++) + { + assert(iBlue >= 0 && iBlue <= 31); + + DifferentialTrys::Try *ptry = &a_phalf->m_atry[a_phalf->m_uiTrys]; + assert(ptry < &a_phalf->m_atry[DifferentialTrys::Half::MAX_TRYS]); + + ptry->m_iRed = iRed; + ptry->m_iGreen = iGreen; + ptry->m_iBlue = iBlue; + ptry->m_fError = FLT_MAX; + ColorFloatRGBA frgbaColor = ColorFloatRGBA::ConvertFromRGB5((unsigned char)iRed, (unsigned char)iGreen, (unsigned char)iBlue); + + // try each CW + for (unsigned int uiCW = 0; uiCW < CW_RANGES; uiCW++) + { + unsigned int auiPixelSelectors[PIXELS / 2]; + ColorFloatRGBA afrgbaDecodedPixels[PIXELS / 2]; + float afPixelErrors[PIXELS / 2] = { FLT_MAX, FLT_MAX, FLT_MAX, FLT_MAX, + FLT_MAX, FLT_MAX, FLT_MAX, FLT_MAX }; + + // pre-compute decoded pixels for each selector + ColorFloatRGBA afrgbaSelectors[SELECTORS]; + assert(SELECTORS == 4); + afrgbaSelectors[0] = (frgbaColor + s_aafCwTable[uiCW][0]).ClampRGB(); + afrgbaSelectors[1] = (frgbaColor + s_aafCwTable[uiCW][1]).ClampRGB(); + afrgbaSelectors[2] = (frgbaColor + s_aafCwTable[uiCW][2]).ClampRGB(); + afrgbaSelectors[3] = (frgbaColor + s_aafCwTable[uiCW][3]).ClampRGB(); + + for (unsigned int uiPixel = 0; uiPixel < 8; uiPixel++) + { + ColorFloatRGBA *pfrgbaSourcePixel = &m_pafrgbaSource[a_phalf->m_pauiPixelMapping[uiPixel]]; + ColorFloatRGBA frgbaDecodedPixel; + + for (unsigned int uiSelector = 0; uiSelector < SELECTORS; uiSelector++) + { + frgbaDecodedPixel = afrgbaSelectors[uiSelector]; + + float fPixelError; + + fPixelError = CalcPixelError(frgbaDecodedPixel, m_afDecodedAlphas[a_phalf->m_pauiPixelMapping[uiPixel]], + *pfrgbaSourcePixel); + + if (fPixelError < afPixelErrors[uiPixel]) + { + auiPixelSelectors[uiPixel] = uiSelector; + afrgbaDecodedPixels[uiPixel] = frgbaDecodedPixel; + afPixelErrors[uiPixel] = fPixelError; + } + + } + } + + // add up all pixel errors + float fCWError = 0.0f; + for (unsigned int uiPixel = 0; uiPixel < 8; uiPixel++) + { + fCWError += afPixelErrors[uiPixel]; + } + + // if best CW so far + if (fCWError < ptry->m_fError) + { + ptry->m_uiCW = uiCW; + for (unsigned int uiPixel = 0; uiPixel < 8; uiPixel++) + { + ptry->m_auiSelectors[uiPixel] = auiPixelSelectors[uiPixel]; + } + ptry->m_fError = fCWError; + } + + } + + if (ptry->m_fError < fBestTryError) + { + a_phalf->m_ptryBest = ptry; + fBestTryError = ptry->m_fError; + } + + assert(ptry->m_fError < FLT_MAX); + + a_phalf->m_uiTrys++; + } + } + } + + } + + // ---------------------------------------------------------------------------------------------------- + // try an ETC1 individual mode encoding + // use a_boolFlip to set the encoding F bit + // use a_uiRadius to alter basecolor components in the range[-a_uiRadius:a_uiRadius] + // replace the encoding if the encoding error is less than previous encoding + // + void Block4x4Encoding_ETC1::TryIndividual(bool a_boolFlip, unsigned int a_uiRadius) + { + + ColorFloatRGBA frgbaColor1; + ColorFloatRGBA frgbaColor2; + + const unsigned int *pauiPixelMapping1; + const unsigned int *pauiPixelMapping2; + + if (a_boolFlip) + { + frgbaColor1 = m_frgbaSourceAverageTop; + frgbaColor2 = m_frgbaSourceAverageBottom; + + pauiPixelMapping1 = s_auiTopPixelMapping; + pauiPixelMapping2 = s_auiBottomPixelMapping; + } + else + { + frgbaColor1 = m_frgbaSourceAverageLeft; + frgbaColor2 = m_frgbaSourceAverageRight; + + pauiPixelMapping1 = s_auiLeftPixelMapping; + pauiPixelMapping2 = s_auiRightPixelMapping; + } + + IndividualTrys trys(frgbaColor1, frgbaColor2, pauiPixelMapping1, pauiPixelMapping2, a_uiRadius); + + Block4x4Encoding_ETC1 encodingTry = *this; + encodingTry.m_boolFlip = a_boolFlip; + + encodingTry.TryIndividualHalf(&trys.m_half1); + encodingTry.TryIndividualHalf(&trys.m_half2); + + // use the best of each half + IndividualTrys::Try *ptryBest1 = trys.m_half1.m_ptryBest; + IndividualTrys::Try *ptryBest2 = trys.m_half2.m_ptryBest; + encodingTry.m_fError = trys.m_half1.m_ptryBest->m_fError + trys.m_half2.m_ptryBest->m_fError; + + if (encodingTry.m_fError < m_fError) + { + m_mode = MODE_ETC1; + m_boolDiff = false; + m_boolFlip = encodingTry.m_boolFlip; + m_frgbaColor1 = ColorFloatRGBA::ConvertFromRGB4((unsigned char)ptryBest1->m_iRed, (unsigned char)ptryBest1->m_iGreen, (unsigned char)ptryBest1->m_iBlue); + m_frgbaColor2 = ColorFloatRGBA::ConvertFromRGB4((unsigned char)ptryBest2->m_iRed, (unsigned char)ptryBest2->m_iGreen, (unsigned char)ptryBest2->m_iBlue); + m_uiCW1 = ptryBest1->m_uiCW; + m_uiCW2 = ptryBest2->m_uiCW; + + for (unsigned int uiPixelOrder = 0; uiPixelOrder < PIXELS / 2; uiPixelOrder++) + { + unsigned int uiPixel1 = pauiPixelMapping1[uiPixelOrder]; + unsigned int uiPixel2 = pauiPixelMapping2[uiPixelOrder]; + + unsigned int uiSelector1 = ptryBest1->m_auiSelectors[uiPixelOrder]; + unsigned int uiSelector2 = ptryBest2->m_auiSelectors[uiPixelOrder]; + + m_auiSelectors[uiPixel1] = uiSelector1; + m_auiSelectors[uiPixel2] = ptryBest2->m_auiSelectors[uiPixelOrder]; + + float fDeltaRGB1 = s_aafCwTable[m_uiCW1][uiSelector1]; + float fDeltaRGB2 = s_aafCwTable[m_uiCW2][uiSelector2]; + + m_afrgbaDecodedColors[uiPixel1] = (m_frgbaColor1 + fDeltaRGB1).ClampRGB(); + m_afrgbaDecodedColors[uiPixel2] = (m_frgbaColor2 + fDeltaRGB2).ClampRGB(); + } + + m_fError1 = ptryBest1->m_fError; + m_fError2 = ptryBest2->m_fError; + m_fError = m_fError1 + m_fError2; + } + + } + + // ---------------------------------------------------------------------------------------------------- + // try an ETC1 differential mode encoding for a half of a 4x4 block + // vary the basecolor components using a radius + // + void Block4x4Encoding_ETC1::TryIndividualHalf(IndividualTrys::Half *a_phalf) + { + + a_phalf->m_ptryBest = nullptr; + float fBestTryError = FLT_MAX; + + a_phalf->m_uiTrys = 0; + for (int iRed = a_phalf->m_iRed - (int)a_phalf->m_uiRadius; + iRed <= a_phalf->m_iRed + (int)a_phalf->m_uiRadius; + iRed++) + { + assert(iRed >= 0 && iRed <= 15); + + for (int iGreen = a_phalf->m_iGreen - (int)a_phalf->m_uiRadius; + iGreen <= a_phalf->m_iGreen + (int)a_phalf->m_uiRadius; + iGreen++) + { + assert(iGreen >= 0 && iGreen <= 15); + + for (int iBlue = a_phalf->m_iBlue - (int)a_phalf->m_uiRadius; + iBlue <= a_phalf->m_iBlue + (int)a_phalf->m_uiRadius; + iBlue++) + { + assert(iBlue >= 0 && iBlue <= 15); + + IndividualTrys::Try *ptry = &a_phalf->m_atry[a_phalf->m_uiTrys]; + assert(ptry < &a_phalf->m_atry[IndividualTrys::Half::MAX_TRYS]); + + ptry->m_iRed = iRed; + ptry->m_iGreen = iGreen; + ptry->m_iBlue = iBlue; + ptry->m_fError = FLT_MAX; + ColorFloatRGBA frgbaColor = ColorFloatRGBA::ConvertFromRGB4((unsigned char)iRed, (unsigned char)iGreen, (unsigned char)iBlue); + + // try each CW + for (unsigned int uiCW = 0; uiCW < CW_RANGES; uiCW++) + { + unsigned int auiPixelSelectors[PIXELS / 2]; + ColorFloatRGBA afrgbaDecodedPixels[PIXELS / 2]; + float afPixelErrors[PIXELS / 2] = { FLT_MAX, FLT_MAX, FLT_MAX, FLT_MAX, + FLT_MAX, FLT_MAX, FLT_MAX, FLT_MAX }; + + // pre-compute decoded pixels for each selector + ColorFloatRGBA afrgbaSelectors[SELECTORS]; + assert(SELECTORS == 4); + afrgbaSelectors[0] = (frgbaColor + s_aafCwTable[uiCW][0]).ClampRGB(); + afrgbaSelectors[1] = (frgbaColor + s_aafCwTable[uiCW][1]).ClampRGB(); + afrgbaSelectors[2] = (frgbaColor + s_aafCwTable[uiCW][2]).ClampRGB(); + afrgbaSelectors[3] = (frgbaColor + s_aafCwTable[uiCW][3]).ClampRGB(); + + for (unsigned int uiPixel = 0; uiPixel < 8; uiPixel++) + { + ColorFloatRGBA *pfrgbaSourcePixel = &m_pafrgbaSource[a_phalf->m_pauiPixelMapping[uiPixel]]; + ColorFloatRGBA frgbaDecodedPixel; + + for (unsigned int uiSelector = 0; uiSelector < SELECTORS; uiSelector++) + { + frgbaDecodedPixel = afrgbaSelectors[uiSelector]; + + float fPixelError; + + fPixelError = CalcPixelError(frgbaDecodedPixel, m_afDecodedAlphas[a_phalf->m_pauiPixelMapping[uiPixel]], + *pfrgbaSourcePixel); + + if (fPixelError < afPixelErrors[uiPixel]) + { + auiPixelSelectors[uiPixel] = uiSelector; + afrgbaDecodedPixels[uiPixel] = frgbaDecodedPixel; + afPixelErrors[uiPixel] = fPixelError; + } + + } + } + + // add up all pixel errors + float fCWError = 0.0f; + for (unsigned int uiPixel = 0; uiPixel < 8; uiPixel++) + { + fCWError += afPixelErrors[uiPixel]; + } + + // if best CW so far + if (fCWError < ptry->m_fError) + { + ptry->m_uiCW = uiCW; + for (unsigned int uiPixel = 0; uiPixel < 8; uiPixel++) + { + ptry->m_auiSelectors[uiPixel] = auiPixelSelectors[uiPixel]; + } + ptry->m_fError = fCWError; + } + + } + + if (ptry->m_fError < fBestTryError) + { + a_phalf->m_ptryBest = ptry; + fBestTryError = ptry->m_fError; + } + + assert(ptry->m_fError < FLT_MAX); + + a_phalf->m_uiTrys++; + } + } + } + + } + + // ---------------------------------------------------------------------------------------------------- + // try version 1 of the degenerate search + // degenerate encodings use basecolor movement and a subset of the selectors to find useful encodings + // each subsequent version of the degenerate search uses more basecolor movement and is less likely to + // be successfull + // + void Block4x4Encoding_ETC1::TryDegenerates1(void) + { + + TryDifferential(m_boolMostLikelyFlip, 1, -2, 0); + TryDifferential(m_boolMostLikelyFlip, 1, 2, 0); + TryDifferential(m_boolMostLikelyFlip, 1, 0, 2); + TryDifferential(m_boolMostLikelyFlip, 1, 0, -2); + + } + + // ---------------------------------------------------------------------------------------------------- + // try version 2 of the degenerate search + // degenerate encodings use basecolor movement and a subset of the selectors to find useful encodings + // each subsequent version of the degenerate search uses more basecolor movement and is less likely to + // be successfull + // + void Block4x4Encoding_ETC1::TryDegenerates2(void) + { + + TryDifferential(!m_boolMostLikelyFlip, 1, -2, 0); + TryDifferential(!m_boolMostLikelyFlip, 1, 2, 0); + TryDifferential(!m_boolMostLikelyFlip, 1, 0, 2); + TryDifferential(!m_boolMostLikelyFlip, 1, 0, -2); + + } + + // ---------------------------------------------------------------------------------------------------- + // try version 3 of the degenerate search + // degenerate encodings use basecolor movement and a subset of the selectors to find useful encodings + // each subsequent version of the degenerate search uses more basecolor movement and is less likely to + // be successfull + // + void Block4x4Encoding_ETC1::TryDegenerates3(void) + { + + TryDifferential(m_boolMostLikelyFlip, 1, -2, -2); + TryDifferential(m_boolMostLikelyFlip, 1, -2, 2); + TryDifferential(m_boolMostLikelyFlip, 1, 2, -2); + TryDifferential(m_boolMostLikelyFlip, 1, 2, 2); + + } + + // ---------------------------------------------------------------------------------------------------- + // try version 4 of the degenerate search + // degenerate encodings use basecolor movement and a subset of the selectors to find useful encodings + // each subsequent version of the degenerate search uses more basecolor movement and is less likely to + // be successfull + // + void Block4x4Encoding_ETC1::TryDegenerates4(void) + { + + TryDifferential(m_boolMostLikelyFlip, 1, -4, 0); + TryDifferential(m_boolMostLikelyFlip, 1, 4, 0); + TryDifferential(m_boolMostLikelyFlip, 1, 0, 4); + TryDifferential(m_boolMostLikelyFlip, 1, 0, -4); + + } + + // ---------------------------------------------------------------------------------------------------- + // find the best selector for each pixel based on a particular basecolor and CW that have been previously set + // calculate the selectors for each half of the block separately + // set the block error as the sum of each half's error + // + void Block4x4Encoding_ETC1::CalculateSelectors() + { + if (m_boolFlip) + { + CalculateHalfOfTheSelectors(0, s_auiTopPixelMapping); + CalculateHalfOfTheSelectors(1, s_auiBottomPixelMapping); + } + else + { + CalculateHalfOfTheSelectors(0, s_auiLeftPixelMapping); + CalculateHalfOfTheSelectors(1, s_auiRightPixelMapping); + } + + m_fError = m_fError1 + m_fError2; + } + + // ---------------------------------------------------------------------------------------------------- + // choose best selectors for half of the block + // calculate the error for half of the block + // + void Block4x4Encoding_ETC1::CalculateHalfOfTheSelectors(unsigned int a_uiHalf, + const unsigned int *pauiPixelMapping) + { + static const bool DEBUG_PRINT = false; + + ColorFloatRGBA *pfrgbaColor = a_uiHalf ? &m_frgbaColor2 : &m_frgbaColor1; + unsigned int *puiCW = a_uiHalf ? &m_uiCW2 : &m_uiCW1; + + float *pfHalfError = a_uiHalf ? &m_fError2 : &m_fError1; + *pfHalfError = FLT_MAX; + + // try each CW + for (unsigned int uiCW = 0; uiCW < CW_RANGES; uiCW++) + { + if (DEBUG_PRINT) + { + printf("\ncw=%u\n", uiCW); + } + + unsigned int auiPixelSelectors[PIXELS / 2]; + ColorFloatRGBA afrgbaDecodedPixels[PIXELS / 2]; + float afPixelErrors[PIXELS / 2] = { FLT_MAX, FLT_MAX, FLT_MAX, FLT_MAX, FLT_MAX, FLT_MAX, FLT_MAX, FLT_MAX }; + + for (unsigned int uiPixel = 0; uiPixel < 8; uiPixel++) + { + if (DEBUG_PRINT) + { + printf("\tsource [%.2f,%.2f,%.2f]\n", m_pafrgbaSource[pauiPixelMapping[uiPixel]].fR, + m_pafrgbaSource[pauiPixelMapping[uiPixel]].fG, m_pafrgbaSource[pauiPixelMapping[uiPixel]].fB); + } + + ColorFloatRGBA *pfrgbaSourcePixel = &m_pafrgbaSource[pauiPixelMapping[uiPixel]]; + ColorFloatRGBA frgbaDecodedPixel; + + for (unsigned int uiSelector = 0; uiSelector < SELECTORS; uiSelector++) + { + float fDeltaRGB = s_aafCwTable[uiCW][uiSelector]; + + frgbaDecodedPixel = (*pfrgbaColor + fDeltaRGB).ClampRGB(); + + float fPixelError; + + fPixelError = CalcPixelError(frgbaDecodedPixel, m_afDecodedAlphas[pauiPixelMapping[uiPixel]], + *pfrgbaSourcePixel); + + if (DEBUG_PRINT) + { + printf("\tpixel %u, index %u [%.2f,%.2f,%.2f], error %.2f", uiPixel, uiSelector, + frgbaDecodedPixel.fR, + frgbaDecodedPixel.fG, + frgbaDecodedPixel.fB, + fPixelError); + } + + if (fPixelError < afPixelErrors[uiPixel]) + { + if (DEBUG_PRINT) + { + printf(" *"); + } + + auiPixelSelectors[uiPixel] = uiSelector; + afrgbaDecodedPixels[uiPixel] = frgbaDecodedPixel; + afPixelErrors[uiPixel] = fPixelError; + } + + if (DEBUG_PRINT) + { + printf("\n"); + } + } + } + + // add up all pixel errors + float fCWError = 0.0f; + for (unsigned int uiPixel = 0; uiPixel < 8; uiPixel++) + { + fCWError += afPixelErrors[uiPixel]; + } + if (DEBUG_PRINT) + { + printf("\terror %.2f\n", fCWError); + } + + // if best CW so far + if (fCWError < *pfHalfError) + { + *pfHalfError = fCWError; + *puiCW = uiCW; + for (unsigned int uiPixel = 0; uiPixel < 8; uiPixel++) + { + m_auiSelectors[pauiPixelMapping[uiPixel]] = auiPixelSelectors[uiPixel]; + m_afrgbaDecodedColors[pauiPixelMapping[uiPixel]] = afrgbaDecodedPixels[uiPixel]; + } + } + } + + } + + // ---------------------------------------------------------------------------------------------------- + // set the encoding bits based on encoding state + // + void Block4x4Encoding_ETC1::SetEncodingBits(void) + { + assert(m_mode == MODE_ETC1); + + if (m_boolDiff) + { + int iRed1 = m_frgbaColor1.IntRed(31.0f); + int iGreen1 = m_frgbaColor1.IntGreen(31.0f); + int iBlue1 = m_frgbaColor1.IntBlue(31.0f); + + int iRed2 = m_frgbaColor2.IntRed(31.0f); + int iGreen2 = m_frgbaColor2.IntGreen(31.0f); + int iBlue2 = m_frgbaColor2.IntBlue(31.0f); + + int iDRed2 = iRed2 - iRed1; + int iDGreen2 = iGreen2 - iGreen1; + int iDBlue2 = iBlue2 - iBlue1; + + assert(iDRed2 >= -4 && iDRed2 < 4); + assert(iDGreen2 >= -4 && iDGreen2 < 4); + assert(iDBlue2 >= -4 && iDBlue2 < 4); + + m_pencodingbitsRGB8->differential.red1 = (unsigned int)iRed1; + m_pencodingbitsRGB8->differential.green1 = (unsigned int)iGreen1; + m_pencodingbitsRGB8->differential.blue1 = (unsigned int)iBlue1; + + m_pencodingbitsRGB8->differential.dred2 = iDRed2; + m_pencodingbitsRGB8->differential.dgreen2 = iDGreen2; + m_pencodingbitsRGB8->differential.dblue2 = iDBlue2; + } + else + { + m_pencodingbitsRGB8->individual.red1 = (unsigned int)m_frgbaColor1.IntRed(15.0f); + m_pencodingbitsRGB8->individual.green1 = (unsigned int)m_frgbaColor1.IntGreen(15.0f); + m_pencodingbitsRGB8->individual.blue1 = (unsigned int)m_frgbaColor1.IntBlue(15.0f); + + m_pencodingbitsRGB8->individual.red2 = (unsigned int)m_frgbaColor2.IntRed(15.0f); + m_pencodingbitsRGB8->individual.green2 = (unsigned int)m_frgbaColor2.IntGreen(15.0f); + m_pencodingbitsRGB8->individual.blue2 = (unsigned int)m_frgbaColor2.IntBlue(15.0f); + } + + m_pencodingbitsRGB8->individual.cw1 = m_uiCW1; + m_pencodingbitsRGB8->individual.cw2 = m_uiCW2; + + SetEncodingBits_Selectors(); + + m_pencodingbitsRGB8->individual.diff = (unsigned int)m_boolDiff; + m_pencodingbitsRGB8->individual.flip = (unsigned int)m_boolFlip; + + } + + // ---------------------------------------------------------------------------------------------------- + // set the selectors in the encoding bits + // + void Block4x4Encoding_ETC1::SetEncodingBits_Selectors(void) + { + + m_pencodingbitsRGB8->individual.selectors = 0; + for (unsigned int uiPixel = 0; uiPixel < PIXELS; uiPixel++) + { + unsigned int uiSelector = m_auiSelectors[uiPixel]; + + // set index msb + m_pencodingbitsRGB8->individual.selectors |= (uiSelector >> 1) << (uiPixel ^ 8); + + // set index lsb + m_pencodingbitsRGB8->individual.selectors |= (uiSelector & 1) << ((16 + uiPixel) ^ 8); + } + + } + + // ---------------------------------------------------------------------------------------------------- + // set the decoded colors and decoded alpha based on the encoding state + // + void Block4x4Encoding_ETC1::Decode(void) + { + + const unsigned int *pauiPixelOrder = m_boolFlip ? s_auiPixelOrderFlip1 : s_auiPixelOrderFlip0; + + for (unsigned int uiPixelOrder = 0; uiPixelOrder < PIXELS; uiPixelOrder++) + { + ColorFloatRGBA *pfrgbaCenter = uiPixelOrder < 8 ? &m_frgbaColor1 : &m_frgbaColor2; + unsigned int uiCW = uiPixelOrder < 8 ? m_uiCW1 : m_uiCW2; + + unsigned int uiPixel = pauiPixelOrder[uiPixelOrder]; + + float fDelta = s_aafCwTable[uiCW][m_auiSelectors[uiPixel]]; + m_afrgbaDecodedColors[uiPixel] = (*pfrgbaCenter + fDelta).ClampRGB(); + m_afDecodedAlphas[uiPixel] = 1.0f; + } + + } + + // ---------------------------------------------------------------------------------------------------- + // + +} // namespace Etc diff --git a/EtcLib/EtcCodec/EtcBlock4x4Encoding_ETC1.h b/EtcLib/EtcCodec/EtcBlock4x4Encoding_ETC1.h new file mode 100644 index 0000000..816aa05 --- /dev/null +++ b/EtcLib/EtcCodec/EtcBlock4x4Encoding_ETC1.h @@ -0,0 +1,186 @@ +/* + * Copyright 2015 The Etc2Comp Authors. + * + * Licensed under the Apache License, Version 2.0 (the "License"); + * you may not use this file except in compliance with the License. + * You may obtain a copy of the License at + * + * http://www.apache.org/licenses/LICENSE-2.0 + * + * Unless required by applicable law or agreed to in writing, software + * distributed under the License is distributed on an "AS IS" BASIS, + * WITHOUT WARRANTIES OR CONDITIONS OF ANY KIND, either express or implied. + * See the License for the specific language governing permissions and + * limitations under the License. + */ + +#pragma once + +#include "EtcBlock4x4Encoding.h" +#include "EtcBlock4x4EncodingBits.h" +#include "EtcDifferentialTrys.h" +#include "EtcIndividualTrys.h" + +namespace Etc +{ + + // base class for Block4x4Encoding_RGB8 + class Block4x4Encoding_ETC1 : public Block4x4Encoding + { + public: + + Block4x4Encoding_ETC1(void); + virtual ~Block4x4Encoding_ETC1(void); + + virtual void InitFromSource(Block4x4 *a_pblockParent, + ColorFloatRGBA *a_pafrgbaSource, + + unsigned char *a_paucEncodingBits, + ErrorMetric a_errormetric); + + virtual void InitFromEncodingBits(Block4x4 *a_pblockParent, + unsigned char *a_paucEncodingBits, + ColorFloatRGBA *a_pafrgbaSource, + + ErrorMetric a_errormetric); + + virtual void PerformIteration(float a_fEffort); + + inline virtual bool GetFlip(void) + { + return m_boolFlip; + } + + inline virtual bool IsDifferential(void) + { + return m_boolDiff; + } + + virtual void SetEncodingBits(void); + + void Decode(void); + + inline ColorFloatRGBA GetColor1(void) const + { + return m_frgbaColor1; + } + + inline ColorFloatRGBA GetColor2(void) const + { + return m_frgbaColor2; + } + + inline const unsigned int * GetSelectors(void) const + { + return m_auiSelectors; + } + + inline unsigned int GetCW1(void) const + { + return m_uiCW1; + } + + inline unsigned int GetCW2(void) const + { + return m_uiCW2; + } + + inline bool HasSeverelyBentDifferentialColors(void) const + { + return m_boolSeverelyBentDifferentialColors; + } + + protected: + + static const unsigned int s_auiPixelOrderFlip0[PIXELS]; + static const unsigned int s_auiPixelOrderFlip1[PIXELS]; + static const unsigned int s_auiPixelOrderHScan[PIXELS]; + + static const unsigned int s_auiLeftPixelMapping[8]; + static const unsigned int s_auiRightPixelMapping[8]; + static const unsigned int s_auiTopPixelMapping[8]; + static const unsigned int s_auiBottomPixelMapping[8]; + + static const unsigned int SELECTOR_BITS = 2; + static const unsigned int SELECTORS = 1 << SELECTOR_BITS; + + static const unsigned int CW_BITS = 3; + static const unsigned int CW_RANGES = 1 << CW_BITS; + + static float s_aafCwTable[CW_RANGES][SELECTORS]; + static unsigned char s_aucDifferentialCwRange[256]; + + static const int MAX_DIFFERENTIAL = 3; + static const int MIN_DIFFERENTIAL = -4; + + void InitFromEncodingBits_Selectors(void); + + void PerformFirstIteration(void); + void CalculateMostLikelyFlip(void); + + void TryDifferential(bool a_boolFlip, unsigned int a_uiRadius, + int a_iGrayOffset1, int a_iGrayOffset2); + void TryDifferentialHalf(DifferentialTrys::Half *a_phalf); + + void TryIndividual(bool a_boolFlip, unsigned int a_uiRadius); + void TryIndividualHalf(IndividualTrys::Half *a_phalf); + + void TryDegenerates1(void); + void TryDegenerates2(void); + void TryDegenerates3(void); + void TryDegenerates4(void); + + void CalculateSelectors(); + void CalculateHalfOfTheSelectors(unsigned int a_uiHalf, + const unsigned int *pauiPixelMapping); + + // calculate the distance2 of r_frgbaPixel from r_frgbaTarget's gray line + inline float CalcGrayDistance2(ColorFloatRGBA &r_frgbaPixel, + ColorFloatRGBA &r_frgbaTarget) + { + float fDeltaGray = ((r_frgbaPixel.fR - r_frgbaTarget.fR) + + (r_frgbaPixel.fG - r_frgbaTarget.fG) + + (r_frgbaPixel.fB - r_frgbaTarget.fB)) / 3.0f; + + ColorFloatRGBA frgbaPointOnGrayLine = (r_frgbaTarget + fDeltaGray).ClampRGB(); + + float fDR = r_frgbaPixel.fR - frgbaPointOnGrayLine.fR; + float fDG = r_frgbaPixel.fG - frgbaPointOnGrayLine.fG; + float fDB = r_frgbaPixel.fB - frgbaPointOnGrayLine.fB; + + return (fDR*fDR) + (fDG*fDG) + (fDB*fDB); + } + + void SetEncodingBits_Selectors(void); + + // intermediate encoding + bool m_boolDiff; + bool m_boolFlip; + ColorFloatRGBA m_frgbaColor1; + ColorFloatRGBA m_frgbaColor2; + unsigned int m_uiCW1; + unsigned int m_uiCW2; + unsigned int m_auiSelectors[PIXELS]; + + // state shared between iterations + ColorFloatRGBA m_frgbaSourceAverageLeft; + ColorFloatRGBA m_frgbaSourceAverageRight; + ColorFloatRGBA m_frgbaSourceAverageTop; + ColorFloatRGBA m_frgbaSourceAverageBottom; + bool m_boolMostLikelyFlip; + + // stats + float m_fError1; // error for Etc1 half 1 + float m_fError2; // error for Etc1 half 2 + bool m_boolSeverelyBentDifferentialColors; // only valid if m_boolDiff; + + // final encoding + Block4x4EncodingBits_RGB8 *m_pencodingbitsRGB8; // or RGB8 portion of Block4x4EncodingBits_RGB8A8 + + private: + + void CalculateSourceAverages(void); + + }; + +} // namespace Etc diff --git a/EtcLib/EtcCodec/EtcBlock4x4Encoding_R11.cpp b/EtcLib/EtcCodec/EtcBlock4x4Encoding_R11.cpp new file mode 100644 index 0000000..5dd9884 --- /dev/null +++ b/EtcLib/EtcCodec/EtcBlock4x4Encoding_R11.cpp @@ -0,0 +1,429 @@ +/* + * Copyright 2015 The Etc2Comp Authors. + * + * Licensed under the Apache License, Version 2.0 (the "License"); + * you may not use this file except in compliance with the License. + * You may obtain a copy of the License at + * + * http://www.apache.org/licenses/LICENSE-2.0 + * + * Unless required by applicable law or agreed to in writing, software + * distributed under the License is distributed on an "AS IS" BASIS, + * WITHOUT WARRANTIES OR CONDITIONS OF ANY KIND, either express or implied. + * See the License for the specific language governing permissions and + * limitations under the License. + */ + +/* +EtcBlock4x4Encoding_R11.cpp + +Block4x4Encoding_R11 is the encoder to use when targetting file format R11 and SR11 (signed R11). + +*/ + +#include "EtcConfig.h" +#include "EtcBlock4x4Encoding_R11.h" + +#include "EtcBlock4x4EncodingBits.h" +#include "EtcBlock4x4.h" + +#include +#include +#include +#include +#include + +namespace Etc +{ + + // modifier values to use for R11, SR11, RG11 and SRG11 + float Block4x4Encoding_R11::s_aafModifierTable[MODIFIER_TABLE_ENTRYS][SELECTORS] + { + { -3.0f / 255.0f, -6.0f / 255.0f, -9.0f / 255.0f, -15.0f / 255.0f, 2.0f / 255.0f, 5.0f / 255.0f, 8.0f / 255.0f, 14.0f / 255.0f }, + { -3.0f / 255.0f, -7.0f / 255.0f, -10.0f / 255.0f, -13.0f / 255.0f, 2.0f / 255.0f, 6.0f / 255.0f, 9.0f / 255.0f, 12.0f / 255.0f }, + { -2.0f / 255.0f, -5.0f / 255.0f, -8.0f / 255.0f, -13.0f / 255.0f, 1.0f / 255.0f, 4.0f / 255.0f, 7.0f / 255.0f, 12.0f / 255.0f }, + { -2.0f / 255.0f, -4.0f / 255.0f, -6.0f / 255.0f, -13.0f / 255.0f, 1.0f / 255.0f, 3.0f / 255.0f, 5.0f / 255.0f, 12.0f / 255.0f }, + + { -3.0f / 255.0f, -6.0f / 255.0f, -8.0f / 255.0f, -12.0f / 255.0f, 2.0f / 255.0f, 5.0f / 255.0f, 7.0f / 255.0f, 11.0f / 255.0f }, + { -3.0f / 255.0f, -7.0f / 255.0f, -9.0f / 255.0f, -11.0f / 255.0f, 2.0f / 255.0f, 6.0f / 255.0f, 8.0f / 255.0f, 10.0f / 255.0f }, + { -4.0f / 255.0f, -7.0f / 255.0f, -8.0f / 255.0f, -11.0f / 255.0f, 3.0f / 255.0f, 6.0f / 255.0f, 7.0f / 255.0f, 10.0f / 255.0f }, + { -3.0f / 255.0f, -5.0f / 255.0f, -8.0f / 255.0f, -11.0f / 255.0f, 2.0f / 255.0f, 4.0f / 255.0f, 7.0f / 255.0f, 10.0f / 255.0f }, + + { -2.0f / 255.0f, -6.0f / 255.0f, -8.0f / 255.0f, -10.0f / 255.0f, 1.0f / 255.0f, 5.0f / 255.0f, 7.0f / 255.0f, 9.0f / 255.0f }, + { -2.0f / 255.0f, -5.0f / 255.0f, -8.0f / 255.0f, -10.0f / 255.0f, 1.0f / 255.0f, 4.0f / 255.0f, 7.0f / 255.0f, 9.0f / 255.0f }, + { -2.0f / 255.0f, -4.0f / 255.0f, -8.0f / 255.0f, -10.0f / 255.0f, 1.0f / 255.0f, 3.0f / 255.0f, 7.0f / 255.0f, 9.0f / 255.0f }, + { -2.0f / 255.0f, -5.0f / 255.0f, -7.0f / 255.0f, -10.0f / 255.0f, 1.0f / 255.0f, 4.0f / 255.0f, 6.0f / 255.0f, 9.0f / 255.0f }, + + { -3.0f / 255.0f, -4.0f / 255.0f, -7.0f / 255.0f, -10.0f / 255.0f, 2.0f / 255.0f, 3.0f / 255.0f, 6.0f / 255.0f, 9.0f / 255.0f }, + { -1.0f / 255.0f, -2.0f / 255.0f, -3.0f / 255.0f, -10.0f / 255.0f, 0.0f / 255.0f, 1.0f / 255.0f, 2.0f / 255.0f, 9.0f / 255.0f }, + { -4.0f / 255.0f, -6.0f / 255.0f, -8.0f / 255.0f, -9.0f / 255.0f, 3.0f / 255.0f, 5.0f / 255.0f, 7.0f / 255.0f, 8.0f / 255.0f }, + { -3.0f / 255.0f, -5.0f / 255.0f, -7.0f / 255.0f, -9.0f / 255.0f, 2.0f / 255.0f, 4.0f / 255.0f, 6.0f / 255.0f, 8.0f / 255.0f } + }; + + // ---------------------------------------------------------------------------------------------------- + // + Block4x4Encoding_R11::Block4x4Encoding_R11(void) + { + + m_pencodingbitsR11 = nullptr; + + } + + Block4x4Encoding_R11::~Block4x4Encoding_R11(void) {} + // ---------------------------------------------------------------------------------------------------- + // initialization prior to encoding + // a_pblockParent points to the block associated with this encoding + // a_errormetric is used to choose the best encoding + // a_pafrgbaSource points to a 4x4 block subset of the source image + // a_paucEncodingBits points to the final encoding bits + // + void Block4x4Encoding_R11::InitFromSource(Block4x4 *a_pblockParent, + ColorFloatRGBA *a_pafrgbaSource, + unsigned char *a_paucEncodingBits, ErrorMetric a_errormetric) + { + Block4x4Encoding::Init(a_pblockParent, a_pafrgbaSource,a_errormetric); + + m_pencodingbitsR11 = (Block4x4EncodingBits_R11 *)a_paucEncodingBits; + } + + // ---------------------------------------------------------------------------------------------------- + // initialization from the encoding bits of a previous encoding + // a_pblockParent points to the block associated with this encoding + // a_errormetric is used to choose the best encoding + // a_pafrgbaSource points to a 4x4 block subset of the source image + // a_paucEncodingBits points to the final encoding bits of a previous encoding + // + void Block4x4Encoding_R11::InitFromEncodingBits(Block4x4 *a_pblockParent, + unsigned char *a_paucEncodingBits, + ColorFloatRGBA *a_pafrgbaSource, + ErrorMetric a_errormetric) + { + m_pencodingbitsR11 = (Block4x4EncodingBits_R11 *)a_paucEncodingBits; + + // init RGB portion + Block4x4Encoding_RGB8::InitFromEncodingBits(a_pblockParent, + (unsigned char *)m_pencodingbitsR11, + a_pafrgbaSource, + a_errormetric); + + // init R11 portion + { + m_mode = MODE_R11; + if (a_pblockParent->GetImageSource()->GetFormat() == Image::Format::SIGNED_R11 || a_pblockParent->GetImageSource()->GetFormat() == Image::Format::SIGNED_RG11) + { + m_fRedBase = (float)(signed char)m_pencodingbitsR11->data.base; + } + else + { + m_fRedBase = (float)(unsigned char)m_pencodingbitsR11->data.base; + } + m_fRedMultiplier = (float)m_pencodingbitsR11->data.multiplier; + m_uiRedModifierTableIndex = m_pencodingbitsR11->data.table; + + unsigned long long int ulliSelectorBits = 0; + ulliSelectorBits |= (unsigned long long int)m_pencodingbitsR11->data.selectors0 << 40; + ulliSelectorBits |= (unsigned long long int)m_pencodingbitsR11->data.selectors1 << 32; + ulliSelectorBits |= (unsigned long long int)m_pencodingbitsR11->data.selectors2 << 24; + ulliSelectorBits |= (unsigned long long int)m_pencodingbitsR11->data.selectors3 << 16; + ulliSelectorBits |= (unsigned long long int)m_pencodingbitsR11->data.selectors4 << 8; + ulliSelectorBits |= (unsigned long long int)m_pencodingbitsR11->data.selectors5; + for (unsigned int uiPixel = 0; uiPixel < PIXELS; uiPixel++) + { + unsigned int uiShift = 45 - (3 * uiPixel); + m_auiRedSelectors[uiPixel] = (ulliSelectorBits >> uiShift) & (SELECTORS - 1); + } + + // decode the red channel + // calc red error + for (unsigned int uiPixel = 0; uiPixel < PIXELS; uiPixel++) + { + float fDecodedPixelData = 0.0f; + if (a_pblockParent->GetImageSource()->GetFormat() == Image::Format::R11 || a_pblockParent->GetImageSource()->GetFormat() == Image::Format::RG11) + { + fDecodedPixelData = DecodePixelRed(m_fRedBase, m_fRedMultiplier, + m_uiRedModifierTableIndex, + m_auiRedSelectors[uiPixel]); + } + else if (a_pblockParent->GetImageSource()->GetFormat() == Image::Format::SIGNED_R11 || a_pblockParent->GetImageSource()->GetFormat() == Image::Format::SIGNED_RG11) + { + fDecodedPixelData = DecodePixelRed(m_fRedBase + 128, m_fRedMultiplier, + m_uiRedModifierTableIndex, + m_auiRedSelectors[uiPixel]); + } + else + { + assert(0); + } + m_afrgbaDecodedColors[uiPixel] = ColorFloatRGBA(fDecodedPixelData, 0.0f, 0.0f, 1.0f); + } + CalcBlockError(); + } + + } + + // ---------------------------------------------------------------------------------------------------- + // perform a single encoding iteration + // replace the encoding if a better encoding was found + // subsequent iterations generally take longer for each iteration + // set m_boolDone if encoding is perfect or encoding is finished based on a_fEffort + // + void Block4x4Encoding_R11::PerformIteration(float a_fEffort) + { + assert(!m_boolDone); + m_mode = MODE_R11; + + switch (m_uiEncodingIterations) + { + case 0: + m_fError = FLT_MAX; + m_fRedBlockError = FLT_MAX; // artificially high value + CalculateR11(8, 0.0f, 0.0f); + m_fError = m_fRedBlockError; + break; + + case 1: + CalculateR11(8, 2.0f, 1.0f); + m_fError = m_fRedBlockError; + if (a_fEffort <= 24.5f) + { + m_boolDone = true; + } + break; + + case 2: + CalculateR11(8, 12.0f, 1.0f); + m_fError = m_fRedBlockError; + if (a_fEffort <= 49.5f) + { + m_boolDone = true; + } + break; + + case 3: + CalculateR11(7, 6.0f, 1.0f); + m_fError = m_fRedBlockError; + break; + + case 4: + CalculateR11(6, 3.0f, 1.0f); + m_fError = m_fRedBlockError; + break; + + case 5: + CalculateR11(5, 1.0f, 0.0f); + m_fError = m_fRedBlockError; + m_boolDone = true; + break; + + default: + assert(0); + break; + } + + m_uiEncodingIterations++; + SetDoneIfPerfect(); + } + + // ---------------------------------------------------------------------------------------------------- + // find the best combination of base color, multiplier and selectors + // + // a_uiSelectorsUsed limits the number of selector combinations to try + // a_fBaseRadius limits the range of base colors to try + // a_fMultiplierRadius limits the range of multipliers to try + // + void Block4x4Encoding_R11::CalculateR11(unsigned int a_uiSelectorsUsed, + float a_fBaseRadius, float a_fMultiplierRadius) + { + // maps from virtual (monotonic) selector to ETC selector + static const unsigned int auiVirtualSelectorMap[8] = {3, 2, 1, 0, 4, 5, 6, 7}; + + // find min/max red + float fMinRed = 1.0f; + float fMaxRed = 0.0f; + for (unsigned int uiPixel = 0; uiPixel < PIXELS; uiPixel++) + { + // ignore border pixels + float fAlpha = m_pafrgbaSource[uiPixel].fA; + if (isnan(fAlpha)) + { + continue; + } + + float fRed = m_pafrgbaSource[uiPixel].fR; + + if (fRed < fMinRed) + { + fMinRed = fRed; + } + if (fRed > fMaxRed) + { + fMaxRed = fRed; + } + } + assert(fMinRed <= fMaxRed); + + float fRedRange = (fMaxRed - fMinRed); + + // try each modifier table entry + for (unsigned int uiTableEntry = 0; uiTableEntry < MODIFIER_TABLE_ENTRYS; uiTableEntry++) + { + for (unsigned int uiMinVirtualSelector = 0; + uiMinVirtualSelector <= (8- a_uiSelectorsUsed); + uiMinVirtualSelector++) + { + unsigned int uiMaxVirtualSelector = uiMinVirtualSelector + a_uiSelectorsUsed - 1; + + unsigned int uiMinSelector = auiVirtualSelectorMap[uiMinVirtualSelector]; + unsigned int uiMaxSelector = auiVirtualSelectorMap[uiMaxVirtualSelector]; + + float fTableEntryCenter = -s_aafModifierTable[uiTableEntry][uiMinSelector]; + + float fTableEntryRange = s_aafModifierTable[uiTableEntry][uiMaxSelector] - + s_aafModifierTable[uiTableEntry][uiMinSelector]; + + float fCenterRatio = fTableEntryCenter / fTableEntryRange; + + float fCenter = fMinRed + fCenterRatio*fRedRange; + fCenter = roundf(255.0f * fCenter) / 255.0f; + + float fMinBase = fCenter - (a_fBaseRadius / 255.0f); + if (fMinBase < 0.0f) + { + fMinBase = 0.0f; + } + + float fMaxBase = fCenter + (a_fBaseRadius / 255.0f); + if (fMaxBase > 1.0f) + { + fMaxBase = 1.0f; + } + + for (float fBase = fMinBase; fBase <= fMaxBase; fBase += (0.999999f / 255.0f)) + { + float fRangeMultiplier = roundf(fRedRange / fTableEntryRange); + + float fMinMultiplier = fRangeMultiplier - a_fMultiplierRadius; + if (fMinMultiplier < 1.0f) + { + fMinMultiplier = 0.0f; + } + else if (fMinMultiplier > 15.0f) + { + fMinMultiplier = 15.0f; + } + + float fMaxMultiplier = fRangeMultiplier + a_fMultiplierRadius; + if (fMaxMultiplier < 1.0f) + { + fMaxMultiplier = 1.0f; + } + else if (fMaxMultiplier > 15.0f) + { + fMaxMultiplier = 15.0f; + } + + for (float fMultiplier = fMinMultiplier; fMultiplier <= fMaxMultiplier; fMultiplier += 1.0f) + { + // find best selector for each pixel + unsigned int auiBestSelectors[PIXELS]; + float afBestRedError[PIXELS]; + float afBestPixelRed[PIXELS]; + + for (unsigned int uiPixel = 0; uiPixel < PIXELS; uiPixel++) + { + float fBestPixelRedError = FLT_MAX; + + for (unsigned int uiSelector = 0; uiSelector < SELECTORS; uiSelector++) + { + float fPixelRed = DecodePixelRed(fBase * 255.0f, fMultiplier, uiTableEntry, uiSelector); + + ColorFloatRGBA frgba(fPixelRed, m_pafrgbaSource[uiPixel].fG,0.0f,1.0f); + + float fPixelRedError = CalcPixelError(frgba, 1.0f, m_pafrgbaSource[uiPixel]); + + if (fPixelRedError < fBestPixelRedError) + { + fBestPixelRedError = fPixelRedError; + auiBestSelectors[uiPixel] = uiSelector; + afBestRedError[uiPixel] = fBestPixelRedError; + afBestPixelRed[uiPixel] = fPixelRed; + } + } + } + float fBlockError = 0.0f; + for (unsigned int uiPixel = 0; uiPixel < PIXELS; uiPixel++) + { + fBlockError += afBestRedError[uiPixel]; + } + if (fBlockError < m_fRedBlockError) + { + m_fRedBlockError = fBlockError; + + if (m_pblockParent->GetImageSource()->GetFormat() == Image::Format::R11 || m_pblockParent->GetImageSource()->GetFormat() == Image::Format::RG11) + { + m_fRedBase = 255.0f * fBase; + } + else if (m_pblockParent->GetImageSource()->GetFormat() == Image::Format::SIGNED_R11 || m_pblockParent->GetImageSource()->GetFormat() == Image::Format::SIGNED_RG11) + { + m_fRedBase = (fBase * 255) - 128; + } + else + { + assert(0); + } + m_fRedMultiplier = fMultiplier; + m_uiRedModifierTableIndex = uiTableEntry; + + for (unsigned int uiPixel = 0; uiPixel < PIXELS; uiPixel++) + { + m_auiRedSelectors[uiPixel] = auiBestSelectors[uiPixel]; + float fBestPixelRed = afBestPixelRed[uiPixel]; + m_afrgbaDecodedColors[uiPixel] = ColorFloatRGBA(fBestPixelRed, 0.0f, 0.0f, 1.0f); + m_afDecodedAlphas[uiPixel] = 1.0f; + } + } + } + } + + } + } + } + + // ---------------------------------------------------------------------------------------------------- + // set the encoding bits based on encoding state + // + void Block4x4Encoding_R11::SetEncodingBits(void) + { + if (m_pblockParent->GetImageSource()->GetFormat() == Image::Format::R11 || m_pblockParent->GetImageSource()->GetFormat() == Image::Format::RG11) + { + m_pencodingbitsR11->data.base = (unsigned char)roundf(m_fRedBase); + } + else if (m_pblockParent->GetImageSource()->GetFormat() == Image::Format::SIGNED_R11 || m_pblockParent->GetImageSource()->GetFormat() == Image::Format::SIGNED_RG11) + { + m_pencodingbitsR11->data.base = (signed char)roundf(m_fRedBase); + } + else + { + assert(0); + } + m_pencodingbitsR11->data.table = m_uiRedModifierTableIndex; + m_pencodingbitsR11->data.multiplier = (unsigned char)roundf(m_fRedMultiplier); + + unsigned long long int ulliSelectorBits = 0; + for (unsigned int uiPixel = 0; uiPixel < PIXELS; uiPixel++) + { + unsigned int uiShift = 45 - (3 * uiPixel); + ulliSelectorBits |= ((unsigned long long int)m_auiRedSelectors[uiPixel]) << uiShift; + } + + m_pencodingbitsR11->data.selectors0 = ulliSelectorBits >> 40; + m_pencodingbitsR11->data.selectors1 = ulliSelectorBits >> 32; + m_pencodingbitsR11->data.selectors2 = ulliSelectorBits >> 24; + m_pencodingbitsR11->data.selectors3 = ulliSelectorBits >> 16; + m_pencodingbitsR11->data.selectors4 = ulliSelectorBits >> 8; + m_pencodingbitsR11->data.selectors5 = ulliSelectorBits; + } + + // ---------------------------------------------------------------------------------------------------- + // +} diff --git a/EtcLib/EtcCodec/EtcBlock4x4Encoding_R11.h b/EtcLib/EtcCodec/EtcBlock4x4Encoding_R11.h new file mode 100644 index 0000000..99cb4df --- /dev/null +++ b/EtcLib/EtcCodec/EtcBlock4x4Encoding_R11.h @@ -0,0 +1,122 @@ +/* + * Copyright 2015 The Etc2Comp Authors. + * + * Licensed under the Apache License, Version 2.0 (the "License"); + * you may not use this file except in compliance with the License. + * You may obtain a copy of the License at + * + * http://www.apache.org/licenses/LICENSE-2.0 + * + * Unless required by applicable law or agreed to in writing, software + * distributed under the License is distributed on an "AS IS" BASIS, + * WITHOUT WARRANTIES OR CONDITIONS OF ANY KIND, either express or implied. + * See the License for the specific language governing permissions and + * limitations under the License. + */ + +#pragma once + +#include "EtcBlock4x4Encoding_RGB8.h" + +namespace Etc +{ + class Block4x4EncodingBits_R11; + + // ################################################################################ + // Block4x4Encoding_R11 + // ################################################################################ + + class Block4x4Encoding_R11 : public Block4x4Encoding_RGB8 + { + public: + + Block4x4Encoding_R11(void); + virtual ~Block4x4Encoding_R11(void); + + virtual void InitFromSource(Block4x4 *a_pblockParent, + ColorFloatRGBA *a_pafrgbaSource, + unsigned char *a_paucEncodingBits, ErrorMetric a_errormetric); + + virtual void InitFromEncodingBits(Block4x4 *a_pblockParent, + unsigned char *a_paucEncodingBits, + ColorFloatRGBA *a_pafrgbaSource, + ErrorMetric a_errormetric); + + virtual void PerformIteration(float a_fEffort); + + virtual void SetEncodingBits(void); + + inline float GetRedBase(void) const + { + return m_fRedBase; + } + + inline float GetRedMultiplier(void) const + { + return m_fRedMultiplier; + } + + inline int GetRedTableIndex(void) const + { + return m_uiRedModifierTableIndex; + } + + inline const unsigned int * GetRedSelectors(void) const + { + return m_auiRedSelectors; + } + + protected: + + static const unsigned int MODIFIER_TABLE_ENTRYS = 16; + static const unsigned int SELECTOR_BITS = 3; + static const unsigned int SELECTORS = 1 << SELECTOR_BITS; + + static float s_aafModifierTable[MODIFIER_TABLE_ENTRYS][SELECTORS]; + + void CalculateR11(unsigned int a_uiSelectorsUsed, + float a_fBaseRadius, float a_fMultiplierRadius); + + + + + inline float DecodePixelRed(float a_fBase, float a_fMultiplier, + unsigned int a_uiTableIndex, unsigned int a_uiSelector) + { + float fMultiplier = a_fMultiplier; + if (fMultiplier <= 0.0f) + { + fMultiplier = 1.0f / 8.0f; + } + + float fPixelRed = a_fBase * 8 + 4 + + 8 * fMultiplier*s_aafModifierTable[a_uiTableIndex][a_uiSelector]*255; + fPixelRed /= 2047.0f; + + if (fPixelRed < 0.0f) + { + fPixelRed = 0.0f; + } + else if (fPixelRed > 1.0f) + { + fPixelRed = 1.0f; + } + + return fPixelRed; + } + + Block4x4EncodingBits_R11 *m_pencodingbitsR11; + + float m_fRedBase; + float m_fRedMultiplier; + float m_fRedBlockError; + unsigned int m_uiRedModifierTableIndex; + unsigned int m_auiRedSelectors[PIXELS]; + + + }; + + // ---------------------------------------------------------------------------------------------------- + // + +} // namespace Etc diff --git a/EtcLib/EtcCodec/EtcBlock4x4Encoding_RG11.cpp b/EtcLib/EtcCodec/EtcBlock4x4Encoding_RG11.cpp new file mode 100644 index 0000000..a1bc9c9 --- /dev/null +++ b/EtcLib/EtcCodec/EtcBlock4x4Encoding_RG11.cpp @@ -0,0 +1,447 @@ +/* + * Copyright 2015 The Etc2Comp Authors. + * + * Licensed under the Apache License, Version 2.0 (the "License"); + * you may not use this file except in compliance with the License. + * You may obtain a copy of the License at + * + * http://www.apache.org/licenses/LICENSE-2.0 + * + * Unless required by applicable law or agreed to in writing, software + * distributed under the License is distributed on an "AS IS" BASIS, + * WITHOUT WARRANTIES OR CONDITIONS OF ANY KIND, either express or implied. + * See the License for the specific language governing permissions and + * limitations under the License. + */ + +/* +EtcBlock4x4Encoding_RG11.cpp + +Block4x4Encoding_RG11 is the encoder to use when targetting file format RG11 and SRG11 (signed RG11). + +*/ + +#include "EtcConfig.h" +#include "EtcBlock4x4Encoding_RG11.h" + +#include "EtcBlock4x4EncodingBits.h" +#include "EtcBlock4x4.h" + +#include +#include +#include +#include +#include + +namespace Etc +{ + // ---------------------------------------------------------------------------------------------------- + // + Block4x4Encoding_RG11::Block4x4Encoding_RG11(void) + { + m_pencodingbitsRG11 = nullptr; + } + + Block4x4Encoding_RG11::~Block4x4Encoding_RG11(void) {} + // ---------------------------------------------------------------------------------------------------- + // initialization prior to encoding + // a_pblockParent points to the block associated with this encoding + // a_errormetric is used to choose the best encoding + // a_pafrgbaSource points to a 4x4 block subset of the source image + // a_paucEncodingBits points to the final encoding bits + // + void Block4x4Encoding_RG11::InitFromSource(Block4x4 *a_pblockParent, + ColorFloatRGBA *a_pafrgbaSource, + unsigned char *a_paucEncodingBits, ErrorMetric a_errormetric) + { + Block4x4Encoding::Init(a_pblockParent, a_pafrgbaSource,a_errormetric); + + m_pencodingbitsRG11 = (Block4x4EncodingBits_RG11 *)a_paucEncodingBits; + } + + // ---------------------------------------------------------------------------------------------------- + // initialization from the encoding bits of a previous encoding + // a_pblockParent points to the block associated with this encoding + // a_errormetric is used to choose the best encoding + // a_pafrgbaSource points to a 4x4 block subset of the source image + // a_paucEncodingBits points to the final encoding bits of a previous encoding + // + void Block4x4Encoding_RG11::InitFromEncodingBits(Block4x4 *a_pblockParent, + unsigned char *a_paucEncodingBits, + ColorFloatRGBA *a_pafrgbaSource, + ErrorMetric a_errormetric) + { + + m_pencodingbitsRG11 = (Block4x4EncodingBits_RG11 *)a_paucEncodingBits; + + // init RGB portion + Block4x4Encoding_RGB8::InitFromEncodingBits(a_pblockParent, + (unsigned char *)m_pencodingbitsRG11, + a_pafrgbaSource, + a_errormetric); + m_fError = 0.0f; + + { + m_mode = MODE_RG11; + if (a_pblockParent->GetImageSource()->GetFormat() == Image::Format::SIGNED_RG11) + { + m_fRedBase = (float)(signed char)m_pencodingbitsRG11->data.baseR; + m_fGrnBase = (float)(signed char)m_pencodingbitsRG11->data.baseG; + } + else + { + m_fRedBase = (float)(unsigned char)m_pencodingbitsRG11->data.baseR; + m_fGrnBase = (float)(unsigned char)m_pencodingbitsRG11->data.baseG; + } + m_fRedMultiplier = (float)m_pencodingbitsRG11->data.multiplierR; + m_fGrnMultiplier = (float)m_pencodingbitsRG11->data.multiplierG; + m_uiRedModifierTableIndex = m_pencodingbitsRG11->data.tableIndexR; + m_uiGrnModifierTableIndex = m_pencodingbitsRG11->data.tableIndexG; + + unsigned long long int ulliSelectorBitsR = 0; + ulliSelectorBitsR |= (unsigned long long int)m_pencodingbitsRG11->data.selectorsR0 << 40; + ulliSelectorBitsR |= (unsigned long long int)m_pencodingbitsRG11->data.selectorsR1 << 32; + ulliSelectorBitsR |= (unsigned long long int)m_pencodingbitsRG11->data.selectorsR2 << 24; + ulliSelectorBitsR |= (unsigned long long int)m_pencodingbitsRG11->data.selectorsR3 << 16; + ulliSelectorBitsR |= (unsigned long long int)m_pencodingbitsRG11->data.selectorsR4 << 8; + ulliSelectorBitsR |= (unsigned long long int)m_pencodingbitsRG11->data.selectorsR5; + + unsigned long long int ulliSelectorBitsG = 0; + ulliSelectorBitsG |= (unsigned long long int)m_pencodingbitsRG11->data.selectorsG0 << 40; + ulliSelectorBitsG |= (unsigned long long int)m_pencodingbitsRG11->data.selectorsG1 << 32; + ulliSelectorBitsG |= (unsigned long long int)m_pencodingbitsRG11->data.selectorsG2 << 24; + ulliSelectorBitsG |= (unsigned long long int)m_pencodingbitsRG11->data.selectorsG3 << 16; + ulliSelectorBitsG |= (unsigned long long int)m_pencodingbitsRG11->data.selectorsG4 << 8; + ulliSelectorBitsG |= (unsigned long long int)m_pencodingbitsRG11->data.selectorsG5; + + + for (unsigned int uiPixel = 0; uiPixel < PIXELS; uiPixel++) + { + unsigned int uiShift = 45 - (3 * uiPixel); + m_auiRedSelectors[uiPixel] = (ulliSelectorBitsR >> uiShift) & (SELECTORS - 1); + m_auiGrnSelectors[uiPixel] = (ulliSelectorBitsG >> uiShift) & (SELECTORS - 1); + } + + + for (unsigned int uiPixel = 0; uiPixel < PIXELS; uiPixel++) + { + float fRedDecodedData = 0.0f; + float fGrnDecodedData = 0.0f; + if (a_pblockParent->GetImageSource()->GetFormat() == Image::Format::RG11) + { + fRedDecodedData = DecodePixelRed(m_fRedBase, m_fRedMultiplier, m_uiRedModifierTableIndex, m_auiRedSelectors[uiPixel]); + fGrnDecodedData = DecodePixelRed(m_fGrnBase, m_fGrnMultiplier, m_uiGrnModifierTableIndex, m_auiGrnSelectors[uiPixel]); + } + else if (a_pblockParent->GetImageSource()->GetFormat() == Image::Format::SIGNED_RG11) + { + fRedDecodedData = DecodePixelRed(m_fRedBase + 128, m_fRedMultiplier, m_uiRedModifierTableIndex, m_auiRedSelectors[uiPixel]); + fGrnDecodedData = DecodePixelRed(m_fGrnBase + 128, m_fGrnMultiplier, m_uiGrnModifierTableIndex, m_auiGrnSelectors[uiPixel]); + } + else + { + assert(0); + } + m_afrgbaDecodedColors[uiPixel] = ColorFloatRGBA(fRedDecodedData, fGrnDecodedData, 0.0f, 1.0f); + } + + } + + CalcBlockError(); + } + + // ---------------------------------------------------------------------------------------------------- + // perform a single encoding iteration + // replace the encoding if a better encoding was found + // subsequent iterations generally take longer for each iteration + // set m_boolDone if encoding is perfect or encoding is finished based on a_fEffort + // + void Block4x4Encoding_RG11::PerformIteration(float a_fEffort) + { + assert(!m_boolDone); + + switch (m_uiEncodingIterations) + { + case 0: + m_fError = FLT_MAX; + m_fGrnBlockError = FLT_MAX; // artificially high value + m_fRedBlockError = FLT_MAX; + CalculateR11(8, 0.0f, 0.0f); + CalculateG11(8, 0.0f, 0.0f); + m_fError = (m_fGrnBlockError + m_fRedBlockError); + break; + + case 1: + CalculateR11(8, 2.0f, 1.0f); + CalculateG11(8, 2.0f, 1.0f); + m_fError = (m_fGrnBlockError + m_fRedBlockError); + if (a_fEffort <= 24.5f) + { + m_boolDone = true; + } + break; + + case 2: + CalculateR11(8, 12.0f, 1.0f); + CalculateG11(8, 12.0f, 1.0f); + m_fError = (m_fGrnBlockError + m_fRedBlockError); + if (a_fEffort <= 49.5f) + { + m_boolDone = true; + } + break; + + case 3: + CalculateR11(7, 6.0f, 1.0f); + CalculateG11(7, 6.0f, 1.0f); + m_fError = (m_fGrnBlockError + m_fRedBlockError); + break; + + case 4: + CalculateR11(6, 3.0f, 1.0f); + CalculateG11(6, 3.0f, 1.0f); + m_fError = (m_fGrnBlockError + m_fRedBlockError); + break; + + case 5: + CalculateR11(5, 1.0f, 0.0f); + CalculateG11(5, 1.0f, 0.0f); + m_fError = (m_fGrnBlockError + m_fRedBlockError); + m_boolDone = true; + break; + + default: + assert(0); + break; + } + + m_uiEncodingIterations++; + SetDoneIfPerfect(); + } + + // ---------------------------------------------------------------------------------------------------- + // find the best combination of base color, multiplier and selectors + // + // a_uiSelectorsUsed limits the number of selector combinations to try + // a_fBaseRadius limits the range of base colors to try + // a_fMultiplierRadius limits the range of multipliers to try + // + void Block4x4Encoding_RG11::CalculateG11(unsigned int a_uiSelectorsUsed, + float a_fBaseRadius, float a_fMultiplierRadius) + { + // maps from virtual (monotonic) selector to etc selector + static const unsigned int auiVirtualSelectorMap[8] = { 3, 2, 1, 0, 4, 5, 6, 7 }; + + // find min/max Grn + float fMinGrn = 1.0f; + float fMaxGrn = 0.0f; + for (unsigned int uiPixel = 0; uiPixel < PIXELS; uiPixel++) + { + // ignore border pixels + float fAlpha = m_pafrgbaSource[uiPixel].fA; + if (isnan(fAlpha)) + { + continue; + } + + float fGrn = m_pafrgbaSource[uiPixel].fG; + + if (fGrn < fMinGrn) + { + fMinGrn = fGrn; + } + if (fGrn > fMaxGrn) + { + fMaxGrn = fGrn; + } + } + assert(fMinGrn <= fMaxGrn); + + float fGrnRange = (fMaxGrn - fMinGrn); + + // try each modifier table entry + for (unsigned int uiTableEntry = 0; uiTableEntry < MODIFIER_TABLE_ENTRYS; uiTableEntry++) + { + for (unsigned int uiMinVirtualSelector = 0; + uiMinVirtualSelector <= (8 - a_uiSelectorsUsed); + uiMinVirtualSelector++) + { + unsigned int uiMaxVirtualSelector = uiMinVirtualSelector + a_uiSelectorsUsed - 1; + + unsigned int uiMinSelector = auiVirtualSelectorMap[uiMinVirtualSelector]; + unsigned int uiMaxSelector = auiVirtualSelectorMap[uiMaxVirtualSelector]; + + float fTableEntryCenter = -s_aafModifierTable[uiTableEntry][uiMinSelector]; + + float fTableEntryRange = s_aafModifierTable[uiTableEntry][uiMaxSelector] - + s_aafModifierTable[uiTableEntry][uiMinSelector]; + + float fCenterRatio = fTableEntryCenter / fTableEntryRange; + + float fCenter = fMinGrn + fCenterRatio*fGrnRange; + fCenter = roundf(255.0f * fCenter) / 255.0f; + + float fMinBase = fCenter - (a_fBaseRadius / 255.0f); + if (fMinBase < 0.0f) + { + fMinBase = 0.0f; + } + + float fMaxBase = fCenter + (a_fBaseRadius / 255.0f); + if (fMaxBase > 1.0f) + { + fMaxBase = 1.0f; + } + + for (float fBase = fMinBase; fBase <= fMaxBase; fBase += (0.999999f / 255.0f)) + { + float fRangeMultiplier = roundf(fGrnRange / fTableEntryRange); + + float fMinMultiplier = fRangeMultiplier - a_fMultiplierRadius; + if (fMinMultiplier < 1.0f) + { + fMinMultiplier = 0.0f; + } + else if (fMinMultiplier > 15.0f) + { + fMinMultiplier = 15.0f; + } + + float fMaxMultiplier = fRangeMultiplier + a_fMultiplierRadius; + if (fMaxMultiplier < 1.0f) + { + fMaxMultiplier = 1.0f; + } + else if (fMaxMultiplier > 15.0f) + { + fMaxMultiplier = 15.0f; + } + + for (float fMultiplier = fMinMultiplier; fMultiplier <= fMaxMultiplier; fMultiplier += 1.0f) + { + // find best selector for each pixel + unsigned int auiBestSelectors[PIXELS]; + float afBestGrnError[PIXELS]; + float afBestPixelGrn[PIXELS]; + + for (unsigned int uiPixel = 0; uiPixel < PIXELS; uiPixel++) + { + float fBestPixelGrnError = FLT_MAX; + + for (unsigned int uiSelector = 0; uiSelector < SELECTORS; uiSelector++) + { + //DecodePixelRed is not red channel specific + float fPixelGrn = DecodePixelRed(fBase * 255.0f, fMultiplier, uiTableEntry, uiSelector); + + ColorFloatRGBA frgba(m_pafrgbaSource[uiPixel].fR, fPixelGrn, 0.0f, 1.0f); + + float fPixelGrnError = CalcPixelError(frgba, 1.0f, m_pafrgbaSource[uiPixel]); + + if (fPixelGrnError < fBestPixelGrnError) + { + fBestPixelGrnError = fPixelGrnError; + auiBestSelectors[uiPixel] = uiSelector; + afBestGrnError[uiPixel] = fBestPixelGrnError; + afBestPixelGrn[uiPixel] = fPixelGrn; + } + } + } + float fBlockError = 0.0f; + for (unsigned int uiPixel = 0; uiPixel < PIXELS; uiPixel++) + { + fBlockError += afBestGrnError[uiPixel]; + } + + if (fBlockError < m_fGrnBlockError) + { + m_fGrnBlockError = fBlockError; + + if (m_pblockParent->GetImageSource()->GetFormat() == Image::Format::RG11) + { + m_fGrnBase = 255.0f * fBase; + } + else if (m_pblockParent->GetImageSource()->GetFormat() == Image::Format::SIGNED_RG11) + { + m_fGrnBase = (fBase * 255) - 128; + } + else + { + assert(0); + } + m_fGrnMultiplier = fMultiplier; + m_uiGrnModifierTableIndex = uiTableEntry; + for (unsigned int uiPixel = 0; uiPixel < PIXELS; uiPixel++) + { + m_auiGrnSelectors[uiPixel] = auiBestSelectors[uiPixel]; + m_afrgbaDecodedColors[uiPixel].fG = afBestPixelGrn[uiPixel]; + m_afDecodedAlphas[uiPixel] = 1.0f; + } + } + } + } + + } + } + } + + // ---------------------------------------------------------------------------------------------------- + // set the encoding bits based on encoding state + // + void Block4x4Encoding_RG11::SetEncodingBits(void) + { + unsigned long long int ulliSelectorBitsR = 0; + unsigned long long int ulliSelectorBitsG = 0; + for (unsigned int uiPixel = 0; uiPixel < PIXELS; uiPixel++) + { + unsigned int uiShift = 45 - (3 * uiPixel); + ulliSelectorBitsR |= ((unsigned long long int)m_auiRedSelectors[uiPixel]) << uiShift; + ulliSelectorBitsG |= ((unsigned long long int)m_auiGrnSelectors[uiPixel]) << uiShift; + } + if (m_pblockParent->GetImageSource()->GetFormat() == Image::Format::RG11) + { + m_pencodingbitsRG11->data.baseR = (unsigned char)roundf(m_fRedBase); + } + else if (m_pblockParent->GetImageSource()->GetFormat() == Image::Format::SIGNED_RG11) + { + m_pencodingbitsRG11->data.baseR = (signed char)roundf(m_fRedBase); + } + else + { + assert(0); + } + m_pencodingbitsRG11->data.tableIndexR = m_uiRedModifierTableIndex; + m_pencodingbitsRG11->data.multiplierR = (unsigned char)roundf(m_fRedMultiplier); + + m_pencodingbitsRG11->data.selectorsR0 = ulliSelectorBitsR >> 40; + m_pencodingbitsRG11->data.selectorsR1 = ulliSelectorBitsR >> 32; + m_pencodingbitsRG11->data.selectorsR2 = ulliSelectorBitsR >> 24; + m_pencodingbitsRG11->data.selectorsR3 = ulliSelectorBitsR >> 16; + m_pencodingbitsRG11->data.selectorsR4 = ulliSelectorBitsR >> 8; + m_pencodingbitsRG11->data.selectorsR5 = ulliSelectorBitsR; + + if (m_pblockParent->GetImageSource()->GetFormat() == Image::Format::RG11) + { + m_pencodingbitsRG11->data.baseG = (unsigned char)roundf(m_fGrnBase); + } + else if (m_pblockParent->GetImageSource()->GetFormat() == Image::Format::SIGNED_RG11) + { + m_pencodingbitsRG11->data.baseG = (signed char)roundf(m_fGrnBase); + } + else + { + assert(0); + } + m_pencodingbitsRG11->data.tableIndexG = m_uiGrnModifierTableIndex; + m_pencodingbitsRG11->data.multiplierG = (unsigned char)roundf(m_fGrnMultiplier); + + m_pencodingbitsRG11->data.selectorsG0 = ulliSelectorBitsG >> 40; + m_pencodingbitsRG11->data.selectorsG1 = ulliSelectorBitsG >> 32; + m_pencodingbitsRG11->data.selectorsG2 = ulliSelectorBitsG >> 24; + m_pencodingbitsRG11->data.selectorsG3 = ulliSelectorBitsG >> 16; + m_pencodingbitsRG11->data.selectorsG4 = ulliSelectorBitsG >> 8; + m_pencodingbitsRG11->data.selectorsG5 = ulliSelectorBitsG; + + } + + // ---------------------------------------------------------------------------------------------------- + // +} diff --git a/EtcLib/EtcCodec/EtcBlock4x4Encoding_RG11.h b/EtcLib/EtcCodec/EtcBlock4x4Encoding_RG11.h new file mode 100644 index 0000000..4caac1d --- /dev/null +++ b/EtcLib/EtcCodec/EtcBlock4x4Encoding_RG11.h @@ -0,0 +1,86 @@ +/* + * Copyright 2015 The Etc2Comp Authors. + * + * Licensed under the Apache License, Version 2.0 (the "License"); + * you may not use this file except in compliance with the License. + * You may obtain a copy of the License at + * + * http://www.apache.org/licenses/LICENSE-2.0 + * + * Unless required by applicable law or agreed to in writing, software + * distributed under the License is distributed on an "AS IS" BASIS, + * WITHOUT WARRANTIES OR CONDITIONS OF ANY KIND, either express or implied. + * See the License for the specific language governing permissions and + * limitations under the License. + */ + +#pragma once + +#include "EtcBlock4x4Encoding_RGB8.h" +#include "EtcBlock4x4Encoding_R11.h" + +namespace Etc +{ + class Block4x4EncodingBits_RG11; + + // ################################################################################ + // Block4x4Encoding_RG11 + // ################################################################################ + + class Block4x4Encoding_RG11 : public Block4x4Encoding_R11 + { + float m_fGrnBase; + float m_fGrnMultiplier; + float m_fGrnBlockError; + unsigned int m_auiGrnSelectors[PIXELS]; + unsigned int m_uiGrnModifierTableIndex; + public: + + Block4x4Encoding_RG11(void); + virtual ~Block4x4Encoding_RG11(void); + + virtual void InitFromSource(Block4x4 *a_pblockParent, + ColorFloatRGBA *a_pafrgbaSource, + + unsigned char *a_paucEncodingBits, ErrorMetric a_errormetric); + + virtual void InitFromEncodingBits(Block4x4 *a_pblockParent, + unsigned char *a_paucEncodingBits, + ColorFloatRGBA *a_pafrgbaSource, + + ErrorMetric a_errormetric); + + virtual void PerformIteration(float a_fEffort); + + virtual void SetEncodingBits(void); + + Block4x4EncodingBits_RG11 *m_pencodingbitsRG11; + + void CalculateG11(unsigned int a_uiSelectorsUsed, float a_fBaseRadius, float a_fMultiplierRadius); + + inline float GetGrnBase(void) const + { + return m_fGrnBase; + } + + inline float GetGrnMultiplier(void) const + { + return m_fGrnMultiplier; + } + + inline int GetGrnTableIndex(void) const + { + return m_uiGrnModifierTableIndex; + } + + inline const unsigned int * GetGrnSelectors(void) const + { + return m_auiGrnSelectors; + } + + }; + + // ---------------------------------------------------------------------------------------------------- + // + +} // namespace Etc diff --git a/EtcLib/EtcCodec/EtcBlock4x4Encoding_RGB8.cpp b/EtcLib/EtcCodec/EtcBlock4x4Encoding_RGB8.cpp new file mode 100644 index 0000000..6e55c46 --- /dev/null +++ b/EtcLib/EtcCodec/EtcBlock4x4Encoding_RGB8.cpp @@ -0,0 +1,1727 @@ +/* + * Copyright 2015 The Etc2Comp Authors. + * + * Licensed under the Apache License, Version 2.0 (the "License"); + * you may not use this file except in compliance with the License. + * You may obtain a copy of the License at + * + * http://www.apache.org/licenses/LICENSE-2.0 + * + * Unless required by applicable law or agreed to in writing, software + * distributed under the License is distributed on an "AS IS" BASIS, + * WITHOUT WARRANTIES OR CONDITIONS OF ANY KIND, either express or implied. + * See the License for the specific language governing permissions and + * limitations under the License. + */ + +/* +EtcBlock4x4Encoding_RGB8.cpp + +Block4x4Encoding_RGB8 is the encoder to use for the ETC2 extensions when targetting file format RGB8. +This encoder is also used for the ETC2 subset of file format RGBA8. + +Block4x4Encoding_ETC1 encodes the ETC1 subset of RGB8. + +*/ + +#include "EtcConfig.h" +#include "EtcBlock4x4Encoding_RGB8.h" + +#include "EtcBlock4x4EncodingBits.h" +#include "EtcBlock4x4.h" +#include "EtcMath.h" + +#include +#include +#include +#include +#include + +namespace Etc +{ + float Block4x4Encoding_RGB8::s_afTHDistanceTable[TH_DISTANCES] = + { + 3.0f / 255.0f, + 6.0f / 255.0f, + 11.0f / 255.0f, + 16.0f / 255.0f, + 23.0f / 255.0f, + 32.0f / 255.0f, + 41.0f / 255.0f, + 64.0f / 255.0f + }; + + // ---------------------------------------------------------------------------------------------------- + // + Block4x4Encoding_RGB8::Block4x4Encoding_RGB8(void) + { + + m_pencodingbitsRGB8 = nullptr; + + } + + Block4x4Encoding_RGB8::~Block4x4Encoding_RGB8(void) {} + // ---------------------------------------------------------------------------------------------------- + // initialization from the encoding bits of a previous encoding + // a_pblockParent points to the block associated with this encoding + // a_errormetric is used to choose the best encoding + // a_pafrgbaSource points to a 4x4 block subset of the source image + // a_paucEncodingBits points to the final encoding bits of a previous encoding + // + void Block4x4Encoding_RGB8::InitFromEncodingBits(Block4x4 *a_pblockParent, + unsigned char *a_paucEncodingBits, + ColorFloatRGBA *a_pafrgbaSource, + ErrorMetric a_errormetric) + { + + // handle ETC1 modes + Block4x4Encoding_ETC1::InitFromEncodingBits(a_pblockParent, + a_paucEncodingBits, a_pafrgbaSource,a_errormetric); + + m_pencodingbitsRGB8 = (Block4x4EncodingBits_RGB8 *)a_paucEncodingBits; + + // detect if there is a T, H or Planar mode present + if (m_pencodingbitsRGB8->differential.diff) + { + int iRed1 = (int)m_pencodingbitsRGB8->differential.red1; + int iDRed2 = m_pencodingbitsRGB8->differential.dred2; + int iRed2 = iRed1 + iDRed2; + + int iGreen1 = (int)m_pencodingbitsRGB8->differential.green1; + int iDGreen2 = m_pencodingbitsRGB8->differential.dgreen2; + int iGreen2 = iGreen1 + iDGreen2; + + int iBlue1 = (int)m_pencodingbitsRGB8->differential.blue1; + int iDBlue2 = m_pencodingbitsRGB8->differential.dblue2; + int iBlue2 = iBlue1 + iDBlue2; + + if (iRed2 < 0 || iRed2 > 31) + { + InitFromEncodingBits_T(); + } + else if (iGreen2 < 0 || iGreen2 > 31) + { + InitFromEncodingBits_H(); + } + else if (iBlue2 < 0 || iBlue2 > 31) + { + InitFromEncodingBits_Planar(); + } + } + + } + + // ---------------------------------------------------------------------------------------------------- + // initialization from the encoding bits of a previous encoding if T mode is detected + // + void Block4x4Encoding_RGB8::InitFromEncodingBits_T(void) + { + + m_mode = MODE_T; + + unsigned char ucRed1 = (unsigned char)((m_pencodingbitsRGB8->t.red1a << 2) + + m_pencodingbitsRGB8->t.red1b); + unsigned char ucGreen1 = m_pencodingbitsRGB8->t.green1; + unsigned char ucBlue1 = m_pencodingbitsRGB8->t.blue1; + + unsigned char ucRed2 = m_pencodingbitsRGB8->t.red2; + unsigned char ucGreen2 = m_pencodingbitsRGB8->t.green2; + unsigned char ucBlue2 = m_pencodingbitsRGB8->t.blue2; + + m_frgbaColor1 = ColorFloatRGBA::ConvertFromRGB4(ucRed1, ucGreen1, ucBlue1); + m_frgbaColor2 = ColorFloatRGBA::ConvertFromRGB4(ucRed2, ucGreen2, ucBlue2); + + m_uiCW1 = (m_pencodingbitsRGB8->t.da << 1) + m_pencodingbitsRGB8->t.db; + + Block4x4Encoding_ETC1::InitFromEncodingBits_Selectors(); + + DecodePixels_T(); + + CalcBlockError(); + + } + + // ---------------------------------------------------------------------------------------------------- + // initialization from the encoding bits of a previous encoding if H mode is detected + // + void Block4x4Encoding_RGB8::InitFromEncodingBits_H(void) + { + + m_mode = MODE_H; + + unsigned char ucRed1 = m_pencodingbitsRGB8->h.red1; + unsigned char ucGreen1 = (unsigned char)((m_pencodingbitsRGB8->h.green1a << 1) + + m_pencodingbitsRGB8->h.green1b); + unsigned char ucBlue1 = (unsigned char)((m_pencodingbitsRGB8->h.blue1a << 3) + + (m_pencodingbitsRGB8->h.blue1b << 1) + + m_pencodingbitsRGB8->h.blue1c); + + unsigned char ucRed2 = m_pencodingbitsRGB8->h.red2; + unsigned char ucGreen2 = (unsigned char)((m_pencodingbitsRGB8->h.green2a << 1) + + m_pencodingbitsRGB8->h.green2b); + unsigned char ucBlue2 = m_pencodingbitsRGB8->h.blue2; + + m_frgbaColor1 = ColorFloatRGBA::ConvertFromRGB4(ucRed1, ucGreen1, ucBlue1); + m_frgbaColor2 = ColorFloatRGBA::ConvertFromRGB4(ucRed2, ucGreen2, ucBlue2); + + // used to determine the LSB of the CW + unsigned int uiRGB1 = (unsigned int)(((int)ucRed1 << 16) + ((int)ucGreen1 << 8) + (int)ucBlue1); + unsigned int uiRGB2 = (unsigned int)(((int)ucRed2 << 16) + ((int)ucGreen2 << 8) + (int)ucBlue2); + + m_uiCW1 = (m_pencodingbitsRGB8->h.da << 2) + (m_pencodingbitsRGB8->h.db << 1); + if (uiRGB1 >= uiRGB2) + { + m_uiCW1++; + } + + Block4x4Encoding_ETC1::InitFromEncodingBits_Selectors(); + + DecodePixels_H(); + + CalcBlockError(); + + } + + // ---------------------------------------------------------------------------------------------------- + // initialization from the encoding bits of a previous encoding if Planar mode is detected + // + void Block4x4Encoding_RGB8::InitFromEncodingBits_Planar(void) + { + + m_mode = MODE_PLANAR; + + unsigned char ucOriginRed = m_pencodingbitsRGB8->planar.originRed; + unsigned char ucOriginGreen = (unsigned char)((m_pencodingbitsRGB8->planar.originGreen1 << 6) + + m_pencodingbitsRGB8->planar.originGreen2); + unsigned char ucOriginBlue = (unsigned char)((m_pencodingbitsRGB8->planar.originBlue1 << 5) + + (m_pencodingbitsRGB8->planar.originBlue2 << 3) + + (m_pencodingbitsRGB8->planar.originBlue3 << 1) + + m_pencodingbitsRGB8->planar.originBlue4); + + unsigned char ucHorizRed = (unsigned char)((m_pencodingbitsRGB8->planar.horizRed1 << 1) + + m_pencodingbitsRGB8->planar.horizRed2); + unsigned char ucHorizGreen = m_pencodingbitsRGB8->planar.horizGreen; + unsigned char ucHorizBlue = (unsigned char)((m_pencodingbitsRGB8->planar.horizBlue1 << 5) + + m_pencodingbitsRGB8->planar.horizBlue2); + + unsigned char ucVertRed = (unsigned char)((m_pencodingbitsRGB8->planar.vertRed1 << 3) + + m_pencodingbitsRGB8->planar.vertRed2); + unsigned char ucVertGreen = (unsigned char)((m_pencodingbitsRGB8->planar.vertGreen1 << 2) + + m_pencodingbitsRGB8->planar.vertGreen2); + unsigned char ucVertBlue = m_pencodingbitsRGB8->planar.vertBlue; + + m_frgbaColor1 = ColorFloatRGBA::ConvertFromR6G7B6(ucOriginRed, ucOriginGreen, ucOriginBlue); + m_frgbaColor2 = ColorFloatRGBA::ConvertFromR6G7B6(ucHorizRed, ucHorizGreen, ucHorizBlue); + m_frgbaColor3 = ColorFloatRGBA::ConvertFromR6G7B6(ucVertRed, ucVertGreen, ucVertBlue); + + DecodePixels_Planar(); + + CalcBlockError(); + + } + + // ---------------------------------------------------------------------------------------------------- + // perform a single encoding iteration + // replace the encoding if a better encoding was found + // subsequent iterations generally take longer for each iteration + // set m_boolDone if encoding is perfect or encoding is finished based on a_fEffort + // + void Block4x4Encoding_RGB8::PerformIteration(float a_fEffort) + { + assert(!m_boolDone); + + switch (m_uiEncodingIterations) + { + case 0: + Block4x4Encoding_ETC1::PerformFirstIteration(); + if (m_boolDone) + { + break; + } + TryPlanar(0); + SetDoneIfPerfect(); + if (m_boolDone) + { + break; + } + TryTAndH(0); + break; + + case 1: + Block4x4Encoding_ETC1::TryDifferential(m_boolMostLikelyFlip, 1, 0, 0); + break; + + case 2: + Block4x4Encoding_ETC1::TryIndividual(m_boolMostLikelyFlip, 1); + break; + + case 3: + Block4x4Encoding_ETC1::TryDifferential(!m_boolMostLikelyFlip, 1, 0, 0); + break; + + case 4: + Block4x4Encoding_ETC1::TryIndividual(!m_boolMostLikelyFlip, 1); + break; + + case 5: + TryPlanar(1); + if (a_fEffort <= 49.5f) + { + m_boolDone = true; + } + break; + + case 6: + TryTAndH(1); + if (a_fEffort <= 59.5f) + { + m_boolDone = true; + } + break; + + case 7: + Block4x4Encoding_ETC1::TryDegenerates1(); + if (a_fEffort <= 69.5f) + { + m_boolDone = true; + } + break; + + case 8: + Block4x4Encoding_ETC1::TryDegenerates2(); + if (a_fEffort <= 79.5f) + { + m_boolDone = true; + } + break; + + case 9: + Block4x4Encoding_ETC1::TryDegenerates3(); + if (a_fEffort <= 89.5f) + { + m_boolDone = true; + } + break; + + case 10: + Block4x4Encoding_ETC1::TryDegenerates4(); + m_boolDone = true; + break; + + default: + assert(0); + break; + } + + m_uiEncodingIterations++; + + SetDoneIfPerfect(); + } + + // ---------------------------------------------------------------------------------------------------- + // try encoding in Planar mode + // save this encoding if it improves the error + // + void Block4x4Encoding_RGB8::TryPlanar(unsigned int a_uiRadius) + { + Block4x4Encoding_RGB8 encodingTry = *this; + + // init "try" + { + encodingTry.m_mode = MODE_PLANAR; + encodingTry.m_boolDiff = true; + encodingTry.m_boolFlip = false; + } + + encodingTry.CalculatePlanarCornerColors(); + + encodingTry.DecodePixels_Planar(); + + encodingTry.CalcBlockError(); + + if (a_uiRadius > 0) + { + encodingTry.TwiddlePlanar(); + } + + if (encodingTry.m_fError < m_fError) + { + m_mode = MODE_PLANAR; + m_boolDiff = true; + m_boolFlip = false; + m_frgbaColor1 = encodingTry.m_frgbaColor1; + m_frgbaColor2 = encodingTry.m_frgbaColor2; + m_frgbaColor3 = encodingTry.m_frgbaColor3; + + for (unsigned int uiPixel = 0; uiPixel < PIXELS; uiPixel++) + { + m_afrgbaDecodedColors[uiPixel] = encodingTry.m_afrgbaDecodedColors[uiPixel]; + } + + m_fError = encodingTry.m_fError; + } + + } + + // ---------------------------------------------------------------------------------------------------- + // try encoding in T mode or H mode + // save this encoding if it improves the error + // + void Block4x4Encoding_RGB8::TryTAndH(unsigned int a_uiRadius) + { + + CalculateBaseColorsForTAndH(); + + TryT(a_uiRadius); + + TryH(a_uiRadius); + + } + + // ---------------------------------------------------------------------------------------------------- + // calculate original values for base colors + // store them in m_frgbaOriginalColor1 and m_frgbaOriginalColor2 + // + void Block4x4Encoding_RGB8::CalculateBaseColorsForTAndH(void) + { + + ColorFloatRGBA frgbaBlockAverage = (m_frgbaSourceAverageLeft + m_frgbaSourceAverageRight) * 0.5f; + + // find pixel farthest from average gray line + unsigned int uiFarthestPixel = 0; + float fFarthestGrayDistance2 = 0.0f; + unsigned int uiTransparentPixels = 0; + for (unsigned int uiPixel = 0; uiPixel < PIXELS; uiPixel++) + { + // don't count transparent + if (m_pafrgbaSource[uiPixel].fA == 0.0f) + { + uiTransparentPixels++; + } + else + { + float fGrayDistance2 = CalcGrayDistance2(m_pafrgbaSource[uiPixel], frgbaBlockAverage); + + if (fGrayDistance2 > fFarthestGrayDistance2) + { + uiFarthestPixel = uiPixel; + fFarthestGrayDistance2 = fGrayDistance2; + } + } + } + // a transparent block should not reach this method + assert(uiTransparentPixels < PIXELS); + + // set the original base colors to: + // half way to the farthest pixel and + // the mirror color on the other side of the average + ColorFloatRGBA frgbaOffset = (m_pafrgbaSource[uiFarthestPixel] - frgbaBlockAverage) * 0.5f; + m_frgbaOriginalColor1_TAndH = (frgbaBlockAverage + frgbaOffset).QuantizeR4G4B4(); + m_frgbaOriginalColor2_TAndH = (frgbaBlockAverage - frgbaOffset).ClampRGB().QuantizeR4G4B4(); // the "other side" might be out of range + + // move base colors to find best fit + for (unsigned int uiIteration = 0; uiIteration < 10; uiIteration++) + { + // find the center of pixels closest to each color + float fPixelsCloserToColor1 = 0.0f; + ColorFloatRGBA frgbSumPixelsCloserToColor1; + float fPixelsCloserToColor2 = 0.0f; + ColorFloatRGBA frgbSumPixelsCloserToColor2; + for (unsigned int uiPixel = 0; uiPixel < PIXELS; uiPixel++) + { + // don't count transparent pixels + if (m_pafrgbaSource[uiPixel].fA == 0.0f) + { + continue; + } + + float fGrayDistance2ToColor1 = CalcGrayDistance2(m_pafrgbaSource[uiPixel], m_frgbaOriginalColor1_TAndH); + float fGrayDistance2ToColor2 = CalcGrayDistance2(m_pafrgbaSource[uiPixel], m_frgbaOriginalColor2_TAndH); + + ColorFloatRGBA frgbaAlphaWeightedSource = m_pafrgbaSource[uiPixel] * m_pafrgbaSource[uiPixel].fA; + + if (fGrayDistance2ToColor1 <= fGrayDistance2ToColor2) + { + fPixelsCloserToColor1 += m_pafrgbaSource[uiPixel].fA; + frgbSumPixelsCloserToColor1 = frgbSumPixelsCloserToColor1 + frgbaAlphaWeightedSource; + } + else + { + fPixelsCloserToColor2 += m_pafrgbaSource[uiPixel].fA; + frgbSumPixelsCloserToColor2 = frgbSumPixelsCloserToColor2 + frgbaAlphaWeightedSource; + } + } + if (fPixelsCloserToColor1 == 0.0f || fPixelsCloserToColor2 == 0.0f) + { + break; + } + + ColorFloatRGBA frgbAvgColor1Pixels = (frgbSumPixelsCloserToColor1 * (1.0f / fPixelsCloserToColor1)).QuantizeR4G4B4(); + ColorFloatRGBA frgbAvgColor2Pixels = (frgbSumPixelsCloserToColor2 * (1.0f / fPixelsCloserToColor2)).QuantizeR4G4B4(); + + if (frgbAvgColor1Pixels.fR == m_frgbaOriginalColor1_TAndH.fR && + frgbAvgColor1Pixels.fG == m_frgbaOriginalColor1_TAndH.fG && + frgbAvgColor1Pixels.fB == m_frgbaOriginalColor1_TAndH.fB && + frgbAvgColor2Pixels.fR == m_frgbaOriginalColor2_TAndH.fR && + frgbAvgColor2Pixels.fG == m_frgbaOriginalColor2_TAndH.fG && + frgbAvgColor2Pixels.fB == m_frgbaOriginalColor2_TAndH.fB) + { + break; + } + + m_frgbaOriginalColor1_TAndH = frgbAvgColor1Pixels; + m_frgbaOriginalColor2_TAndH = frgbAvgColor2Pixels; + } + + } + + // ---------------------------------------------------------------------------------------------------- + // try encoding in T mode + // save this encoding if it improves the error + // + // since pixels that use base color1 don't use the distance table, color1 and color2 can be twiddled independently + // better encoding can be found if TWIDDLE_RADIUS is set to 2, but it will be much slower + // + void Block4x4Encoding_RGB8::TryT(unsigned int a_uiRadius) + { + Block4x4Encoding_RGB8 encodingTry = *this; + + // init "try" + { + encodingTry.m_mode = MODE_T; + encodingTry.m_boolDiff = true; + encodingTry.m_boolFlip = false; + encodingTry.m_fError = FLT_MAX; + } + + int iColor1Red = m_frgbaOriginalColor1_TAndH.IntRed(15.0f); + int iColor1Green = m_frgbaOriginalColor1_TAndH.IntGreen(15.0f); + int iColor1Blue = m_frgbaOriginalColor1_TAndH.IntBlue(15.0f); + + int iMinRed1 = iColor1Red - (int)a_uiRadius; + if (iMinRed1 < 0) + { + iMinRed1 = 0; + } + int iMaxRed1 = iColor1Red + (int)a_uiRadius; + if (iMaxRed1 > 15) + { + iMinRed1 = 15; + } + + int iMinGreen1 = iColor1Green - (int)a_uiRadius; + if (iMinGreen1 < 0) + { + iMinGreen1 = 0; + } + int iMaxGreen1 = iColor1Green + (int)a_uiRadius; + if (iMaxGreen1 > 15) + { + iMinGreen1 = 15; + } + + int iMinBlue1 = iColor1Blue - (int)a_uiRadius; + if (iMinBlue1 < 0) + { + iMinBlue1 = 0; + } + int iMaxBlue1 = iColor1Blue + (int)a_uiRadius; + if (iMaxBlue1 > 15) + { + iMinBlue1 = 15; + } + + int iColor2Red = m_frgbaOriginalColor2_TAndH.IntRed(15.0f); + int iColor2Green = m_frgbaOriginalColor2_TAndH.IntGreen(15.0f); + int iColor2Blue = m_frgbaOriginalColor2_TAndH.IntBlue(15.0f); + + int iMinRed2 = iColor2Red - (int)a_uiRadius; + if (iMinRed2 < 0) + { + iMinRed2 = 0; + } + int iMaxRed2 = iColor2Red + (int)a_uiRadius; + if (iMaxRed2 > 15) + { + iMinRed2 = 15; + } + + int iMinGreen2 = iColor2Green - (int)a_uiRadius; + if (iMinGreen2 < 0) + { + iMinGreen2 = 0; + } + int iMaxGreen2 = iColor2Green + (int)a_uiRadius; + if (iMaxGreen2 > 15) + { + iMinGreen2 = 15; + } + + int iMinBlue2 = iColor2Blue - (int)a_uiRadius; + if (iMinBlue2 < 0) + { + iMinBlue2 = 0; + } + int iMaxBlue2 = iColor2Blue + (int)a_uiRadius; + if (iMaxBlue2 > 15) + { + iMinBlue2 = 15; + } + + for (unsigned int uiDistance = 0; uiDistance < TH_DISTANCES; uiDistance++) + { + encodingTry.m_uiCW1 = uiDistance; + + // twiddle m_frgbaOriginalColor2_TAndH + // twiddle color2 first, since it affects 3 selectors, while color1 only affects one selector + // + for (int iRed2 = iMinRed2; iRed2 <= iMaxRed2; iRed2++) + { + for (int iGreen2 = iMinGreen2; iGreen2 <= iMaxGreen2; iGreen2++) + { + for (int iBlue2 = iMinBlue2; iBlue2 <= iMaxBlue2; iBlue2++) + { + for (unsigned int uiBaseColorSwaps = 0; uiBaseColorSwaps < 2; uiBaseColorSwaps++) + { + if (uiBaseColorSwaps == 0) + { + encodingTry.m_frgbaColor1 = m_frgbaOriginalColor1_TAndH; + encodingTry.m_frgbaColor2 = ColorFloatRGBA::ConvertFromRGB4((unsigned char)iRed2, (unsigned char)iGreen2, (unsigned char)iBlue2); + } + else + { + encodingTry.m_frgbaColor1 = ColorFloatRGBA::ConvertFromRGB4((unsigned char)iRed2, (unsigned char)iGreen2, (unsigned char)iBlue2); + encodingTry.m_frgbaColor2 = m_frgbaOriginalColor1_TAndH; + } + + encodingTry.TryT_BestSelectorCombination(); + + if (encodingTry.m_fError < m_fError) + { + m_mode = encodingTry.m_mode; + m_boolDiff = encodingTry.m_boolDiff; + m_boolFlip = encodingTry.m_boolFlip; + + m_frgbaColor1 = encodingTry.m_frgbaColor1; + m_frgbaColor2 = encodingTry.m_frgbaColor2; + m_uiCW1 = encodingTry.m_uiCW1; + + for (unsigned int uiPixel = 0; uiPixel < PIXELS; uiPixel++) + { + m_auiSelectors[uiPixel] = encodingTry.m_auiSelectors[uiPixel]; + m_afrgbaDecodedColors[uiPixel] = encodingTry.m_afrgbaDecodedColors[uiPixel]; + } + + m_fError = encodingTry.m_fError; + } + } + } + } + } + + // twiddle m_frgbaOriginalColor1_TAndH + for (int iRed1 = iMinRed1; iRed1 <= iMaxRed1; iRed1++) + { + for (int iGreen1 = iMinGreen1; iGreen1 <= iMaxGreen1; iGreen1++) + { + for (int iBlue1 = iMinBlue1; iBlue1 <= iMaxBlue1; iBlue1++) + { + for (unsigned int uiBaseColorSwaps = 0; uiBaseColorSwaps < 2; uiBaseColorSwaps++) + { + if (uiBaseColorSwaps == 0) + { + encodingTry.m_frgbaColor1 = ColorFloatRGBA::ConvertFromRGB4((unsigned char)iRed1, (unsigned char)iGreen1, (unsigned char)iBlue1); + encodingTry.m_frgbaColor2 = m_frgbaOriginalColor2_TAndH; + } + else + { + encodingTry.m_frgbaColor1 = m_frgbaOriginalColor2_TAndH; + encodingTry.m_frgbaColor2 = ColorFloatRGBA::ConvertFromRGB4((unsigned char)iRed1, (unsigned char)iGreen1, (unsigned char)iBlue1); + } + + encodingTry.TryT_BestSelectorCombination(); + + if (encodingTry.m_fError < m_fError) + { + m_mode = encodingTry.m_mode; + m_boolDiff = encodingTry.m_boolDiff; + m_boolFlip = encodingTry.m_boolFlip; + + m_frgbaColor1 = encodingTry.m_frgbaColor1; + m_frgbaColor2 = encodingTry.m_frgbaColor2; + m_uiCW1 = encodingTry.m_uiCW1; + + for (unsigned int uiPixel = 0; uiPixel < PIXELS; uiPixel++) + { + m_auiSelectors[uiPixel] = encodingTry.m_auiSelectors[uiPixel]; + m_afrgbaDecodedColors[uiPixel] = encodingTry.m_afrgbaDecodedColors[uiPixel]; + } + + m_fError = encodingTry.m_fError; + } + } + } + } + } + + } + + } + + // ---------------------------------------------------------------------------------------------------- + // find best selector combination for TryT + // called on an encodingTry + // + void Block4x4Encoding_RGB8::TryT_BestSelectorCombination(void) + { + + float fDistance = s_afTHDistanceTable[m_uiCW1]; + + unsigned int auiBestPixelSelectors[PIXELS]; + float afBestPixelErrors[PIXELS] = { FLT_MAX, FLT_MAX, FLT_MAX, FLT_MAX, FLT_MAX, FLT_MAX, FLT_MAX, FLT_MAX, + FLT_MAX, FLT_MAX, FLT_MAX, FLT_MAX, FLT_MAX, FLT_MAX, FLT_MAX, FLT_MAX }; + ColorFloatRGBA afrgbaBestDecodedPixels[PIXELS]; + ColorFloatRGBA afrgbaDecodedPixel[SELECTORS]; + + assert(SELECTORS == 4); + afrgbaDecodedPixel[0] = m_frgbaColor1; + afrgbaDecodedPixel[1] = (m_frgbaColor2 + fDistance).ClampRGB(); + afrgbaDecodedPixel[2] = m_frgbaColor2; + afrgbaDecodedPixel[3] = (m_frgbaColor2 - fDistance).ClampRGB(); + + // try each selector + for (unsigned int uiSelector = 0; uiSelector < SELECTORS; uiSelector++) + { + for (unsigned int uiPixel = 0; uiPixel < PIXELS; uiPixel++) + { + + float fPixelError = CalcPixelError(afrgbaDecodedPixel[uiSelector], m_afDecodedAlphas[uiPixel], + m_pafrgbaSource[uiPixel]); + + if (fPixelError < afBestPixelErrors[uiPixel]) + { + afBestPixelErrors[uiPixel] = fPixelError; + auiBestPixelSelectors[uiPixel] = uiSelector; + afrgbaBestDecodedPixels[uiPixel] = afrgbaDecodedPixel[uiSelector]; + } + } + } + + + // add up all of the pixel errors + float fBlockError = 0.0f; + for (unsigned int uiPixel = 0; uiPixel < PIXELS; uiPixel++) + { + fBlockError += afBestPixelErrors[uiPixel]; + } + + if (fBlockError < m_fError) + { + m_fError = fBlockError; + + for (unsigned int uiPixel = 0; uiPixel < PIXELS; uiPixel++) + { + m_auiSelectors[uiPixel] = auiBestPixelSelectors[uiPixel]; + m_afrgbaDecodedColors[uiPixel] = afrgbaBestDecodedPixels[uiPixel]; + } + } + + } + + // ---------------------------------------------------------------------------------------------------- + // try encoding in T mode + // save this encoding if it improves the error + // + // since all pixels use the distance table, color1 and color2 can NOT be twiddled independently + // TWIDDLE_RADIUS of 2 is WAY too slow + // + void Block4x4Encoding_RGB8::TryH(unsigned int a_uiRadius) + { + Block4x4Encoding_RGB8 encodingTry = *this; + + // init "try" + { + encodingTry.m_mode = MODE_H; + encodingTry.m_boolDiff = true; + encodingTry.m_boolFlip = false; + encodingTry.m_fError = FLT_MAX; + } + + int iColor1Red = m_frgbaOriginalColor1_TAndH.IntRed(15.0f); + int iColor1Green = m_frgbaOriginalColor1_TAndH.IntGreen(15.0f); + int iColor1Blue = m_frgbaOriginalColor1_TAndH.IntBlue(15.0f); + + int iMinRed1 = iColor1Red - (int)a_uiRadius; + if (iMinRed1 < 0) + { + iMinRed1 = 0; + } + int iMaxRed1 = iColor1Red + (int)a_uiRadius; + if (iMaxRed1 > 15) + { + iMinRed1 = 15; + } + + int iMinGreen1 = iColor1Green - (int)a_uiRadius; + if (iMinGreen1 < 0) + { + iMinGreen1 = 0; + } + int iMaxGreen1 = iColor1Green + (int)a_uiRadius; + if (iMaxGreen1 > 15) + { + iMinGreen1 = 15; + } + + int iMinBlue1 = iColor1Blue - (int)a_uiRadius; + if (iMinBlue1 < 0) + { + iMinBlue1 = 0; + } + int iMaxBlue1 = iColor1Blue + (int)a_uiRadius; + if (iMaxBlue1 > 15) + { + iMinBlue1 = 15; + } + + int iColor2Red = m_frgbaOriginalColor2_TAndH.IntRed(15.0f); + int iColor2Green = m_frgbaOriginalColor2_TAndH.IntGreen(15.0f); + int iColor2Blue = m_frgbaOriginalColor2_TAndH.IntBlue(15.0f); + + int iMinRed2 = iColor2Red - (int)a_uiRadius; + if (iMinRed2 < 0) + { + iMinRed2 = 0; + } + int iMaxRed2 = iColor2Red + (int)a_uiRadius; + if (iMaxRed2 > 15) + { + iMinRed2 = 15; + } + + int iMinGreen2 = iColor2Green - (int)a_uiRadius; + if (iMinGreen2 < 0) + { + iMinGreen2 = 0; + } + int iMaxGreen2 = iColor2Green + (int)a_uiRadius; + if (iMaxGreen2 > 15) + { + iMinGreen2 = 15; + } + + int iMinBlue2 = iColor2Blue - (int)a_uiRadius; + if (iMinBlue2 < 0) + { + iMinBlue2 = 0; + } + int iMaxBlue2 = iColor2Blue + (int)a_uiRadius; + if (iMaxBlue2 > 15) + { + iMinBlue2 = 15; + } + + for (unsigned int uiDistance = 0; uiDistance < TH_DISTANCES; uiDistance++) + { + encodingTry.m_uiCW1 = uiDistance; + + // twiddle m_frgbaOriginalColor1_TAndH + for (int iRed1 = iMinRed1; iRed1 <= iMaxRed1; iRed1++) + { + for (int iGreen1 = iMinGreen1; iGreen1 <= iMaxGreen1; iGreen1++) + { + for (int iBlue1 = iMinBlue1; iBlue1 <= iMaxBlue1; iBlue1++) + { + encodingTry.m_frgbaColor1 = ColorFloatRGBA::ConvertFromRGB4((unsigned char)iRed1, (unsigned char)iGreen1, (unsigned char)iBlue1); + encodingTry.m_frgbaColor2 = m_frgbaOriginalColor2_TAndH; + + // if color1 == color2, H encoding issues can pop up, so abort + if (iRed1 == iColor2Red && iGreen1 == iColor2Green && iBlue1 == iColor2Blue) + { + continue; + } + + encodingTry.TryH_BestSelectorCombination(); + + if (encodingTry.m_fError < m_fError) + { + m_mode = encodingTry.m_mode; + m_boolDiff = encodingTry.m_boolDiff; + m_boolFlip = encodingTry.m_boolFlip; + + m_frgbaColor1 = encodingTry.m_frgbaColor1; + m_frgbaColor2 = encodingTry.m_frgbaColor2; + m_uiCW1 = encodingTry.m_uiCW1; + + for (unsigned int uiPixel = 0; uiPixel < PIXELS; uiPixel++) + { + m_auiSelectors[uiPixel] = encodingTry.m_auiSelectors[uiPixel]; + m_afrgbaDecodedColors[uiPixel] = encodingTry.m_afrgbaDecodedColors[uiPixel]; + } + + m_fError = encodingTry.m_fError; + } + } + } + } + + // twiddle m_frgbaOriginalColor2_TAndH + for (int iRed2 = iMinRed2; iRed2 <= iMaxRed2; iRed2++) + { + for (int iGreen2 = iMinGreen2; iGreen2 <= iMaxGreen2; iGreen2++) + { + for (int iBlue2 = iMinBlue2; iBlue2 <= iMaxBlue2; iBlue2++) + { + encodingTry.m_frgbaColor1 = m_frgbaOriginalColor1_TAndH; + encodingTry.m_frgbaColor2 = ColorFloatRGBA::ConvertFromRGB4((unsigned char)iRed2, (unsigned char)iGreen2, (unsigned char)iBlue2); + + // if color1 == color2, H encoding issues can pop up, so abort + if (iRed2 == iColor1Red && iGreen2 == iColor1Green && iBlue2 == iColor1Blue) + { + continue; + } + + encodingTry.TryH_BestSelectorCombination(); + + if (encodingTry.m_fError < m_fError) + { + m_mode = encodingTry.m_mode; + m_boolDiff = encodingTry.m_boolDiff; + m_boolFlip = encodingTry.m_boolFlip; + + m_frgbaColor1 = encodingTry.m_frgbaColor1; + m_frgbaColor2 = encodingTry.m_frgbaColor2; + m_uiCW1 = encodingTry.m_uiCW1; + + for (unsigned int uiPixel = 0; uiPixel < PIXELS; uiPixel++) + { + m_auiSelectors[uiPixel] = encodingTry.m_auiSelectors[uiPixel]; + m_afrgbaDecodedColors[uiPixel] = encodingTry.m_afrgbaDecodedColors[uiPixel]; + } + + m_fError = encodingTry.m_fError; + } + } + } + } + + } + + } + + // ---------------------------------------------------------------------------------------------------- + // find best selector combination for TryH + // called on an encodingTry + // + void Block4x4Encoding_RGB8::TryH_BestSelectorCombination(void) + { + + float fDistance = s_afTHDistanceTable[m_uiCW1]; + + unsigned int auiBestPixelSelectors[PIXELS]; + float afBestPixelErrors[PIXELS] = { FLT_MAX, FLT_MAX, FLT_MAX, FLT_MAX, FLT_MAX, FLT_MAX, FLT_MAX, FLT_MAX, + FLT_MAX, FLT_MAX, FLT_MAX, FLT_MAX, FLT_MAX, FLT_MAX, FLT_MAX, FLT_MAX }; + ColorFloatRGBA afrgbaBestDecodedPixels[PIXELS]; + ColorFloatRGBA afrgbaDecodedPixel[SELECTORS]; + + assert(SELECTORS == 4); + afrgbaDecodedPixel[0] = (m_frgbaColor1 + fDistance).ClampRGB(); + afrgbaDecodedPixel[1] = (m_frgbaColor1 - fDistance).ClampRGB(); + afrgbaDecodedPixel[2] = (m_frgbaColor2 + fDistance).ClampRGB(); + afrgbaDecodedPixel[3] = (m_frgbaColor2 - fDistance).ClampRGB(); + + // try each selector + for (unsigned int uiSelector = 0; uiSelector < SELECTORS; uiSelector++) + { + for (unsigned int uiPixel = 0; uiPixel < PIXELS; uiPixel++) + { + + float fPixelError = CalcPixelError(afrgbaDecodedPixel[uiSelector], m_afDecodedAlphas[uiPixel], + m_pafrgbaSource[uiPixel]); + + if (fPixelError < afBestPixelErrors[uiPixel]) + { + afBestPixelErrors[uiPixel] = fPixelError; + auiBestPixelSelectors[uiPixel] = uiSelector; + afrgbaBestDecodedPixels[uiPixel] = afrgbaDecodedPixel[uiSelector]; + } + } + } + + + // add up all of the pixel errors + float fBlockError = 0.0f; + for (unsigned int uiPixel = 0; uiPixel < PIXELS; uiPixel++) + { + fBlockError += afBestPixelErrors[uiPixel]; + } + + if (fBlockError < m_fError) + { + m_fError = fBlockError; + + for (unsigned int uiPixel = 0; uiPixel < PIXELS; uiPixel++) + { + m_auiSelectors[uiPixel] = auiBestPixelSelectors[uiPixel]; + m_afrgbaDecodedColors[uiPixel] = afrgbaBestDecodedPixels[uiPixel]; + } + } + + } + + // ---------------------------------------------------------------------------------------------------- + // use linear regression to find the best fit for colors along the edges of the 4x4 block + // + void Block4x4Encoding_RGB8::CalculatePlanarCornerColors(void) + { + ColorFloatRGBA afrgbaRegression[MAX_PLANAR_REGRESSION_SIZE]; + ColorFloatRGBA frgbaSlope; + ColorFloatRGBA frgbaOffset; + + // top edge + afrgbaRegression[0] = m_pafrgbaSource[0]; + afrgbaRegression[1] = m_pafrgbaSource[4]; + afrgbaRegression[2] = m_pafrgbaSource[8]; + afrgbaRegression[3] = m_pafrgbaSource[12]; + ColorRegression(afrgbaRegression, 4, &frgbaSlope, &frgbaOffset); + m_frgbaColor1 = frgbaOffset; + m_frgbaColor2 = (frgbaSlope * 4.0f) + frgbaOffset; + + // left edge + afrgbaRegression[0] = m_pafrgbaSource[0]; + afrgbaRegression[1] = m_pafrgbaSource[1]; + afrgbaRegression[2] = m_pafrgbaSource[2]; + afrgbaRegression[3] = m_pafrgbaSource[3]; + ColorRegression(afrgbaRegression, 4, &frgbaSlope, &frgbaOffset); + m_frgbaColor1 = (m_frgbaColor1 + frgbaOffset) * 0.5f; // average with top edge + m_frgbaColor3 = (frgbaSlope * 4.0f) + frgbaOffset; + + // right edge + afrgbaRegression[0] = m_pafrgbaSource[12]; + afrgbaRegression[1] = m_pafrgbaSource[13]; + afrgbaRegression[2] = m_pafrgbaSource[14]; + afrgbaRegression[3] = m_pafrgbaSource[15]; + ColorRegression(afrgbaRegression, 4, &frgbaSlope, &frgbaOffset); + m_frgbaColor2 = (m_frgbaColor2 + frgbaOffset) * 0.5f; // average with top edge + + // bottom edge + afrgbaRegression[0] = m_pafrgbaSource[3]; + afrgbaRegression[1] = m_pafrgbaSource[7]; + afrgbaRegression[2] = m_pafrgbaSource[11]; + afrgbaRegression[3] = m_pafrgbaSource[15]; + ColorRegression(afrgbaRegression, 4, &frgbaSlope, &frgbaOffset); + m_frgbaColor3 = (m_frgbaColor3 + frgbaOffset) * 0.5f; // average with left edge + + // quantize corner colors to 6/7/6 + m_frgbaColor1 = m_frgbaColor1.QuantizeR6G7B6(); + m_frgbaColor2 = m_frgbaColor2.QuantizeR6G7B6(); + m_frgbaColor3 = m_frgbaColor3.QuantizeR6G7B6(); + + } + + // ---------------------------------------------------------------------------------------------------- + // try different corner colors by slightly changing R, G and B independently + // + // R, G and B decoding and errors are independent, so R, G and B twiddles can be independent + // + // return true if improvement + // + bool Block4x4Encoding_RGB8::TwiddlePlanar(void) + { + bool boolImprovement = false; + + while (TwiddlePlanarR()) + { + boolImprovement = true; + } + + while (TwiddlePlanarG()) + { + boolImprovement = true; + } + + while (TwiddlePlanarB()) + { + boolImprovement = true; + } + + return boolImprovement; + } + + // ---------------------------------------------------------------------------------------------------- + // try different corner colors by slightly changing R + // + bool Block4x4Encoding_RGB8::TwiddlePlanarR() + { + bool boolImprovement = false; + + Block4x4Encoding_RGB8 encodingTry = *this; + + // init "try" + { + encodingTry.m_mode = MODE_PLANAR; + encodingTry.m_boolDiff = true; + encodingTry.m_boolFlip = false; + } + + int iOriginRed = encodingTry.m_frgbaColor1.IntRed(63.0f); + int iHorizRed = encodingTry.m_frgbaColor2.IntRed(63.0f); + int iVertRed = encodingTry.m_frgbaColor3.IntRed(63.0f); + + for (int iTryOriginRed = iOriginRed - 1; iTryOriginRed <= iOriginRed + 1; iTryOriginRed++) + { + // check for out of range + if (iTryOriginRed < 0 || iTryOriginRed > 63) + { + continue; + } + + encodingTry.m_frgbaColor1.fR = ((iTryOriginRed << 2) + (iTryOriginRed >> 4)) / 255.0f; + + for (int iTryHorizRed = iHorizRed - 1; iTryHorizRed <= iHorizRed + 1; iTryHorizRed++) + { + // check for out of range + if (iTryHorizRed < 0 || iTryHorizRed > 63) + { + continue; + } + + encodingTry.m_frgbaColor2.fR = ((iTryHorizRed << 2) + (iTryHorizRed >> 4)) / 255.0f; + + for (int iTryVertRed = iVertRed - 1; iTryVertRed <= iVertRed + 1; iTryVertRed++) + { + // check for out of range + if (iTryVertRed < 0 || iTryVertRed > 63) + { + continue; + } + + // don't bother with null twiddle + if (iTryOriginRed == iOriginRed && iTryHorizRed == iHorizRed && iTryVertRed == iVertRed) + { + continue; + } + + encodingTry.m_frgbaColor3.fR = ((iTryVertRed << 2) + (iTryVertRed >> 4)) / 255.0f; + + encodingTry.DecodePixels_Planar(); + + encodingTry.CalcBlockError(); + + if (encodingTry.m_fError < m_fError) + { + m_mode = MODE_PLANAR; + m_boolDiff = true; + m_boolFlip = false; + m_frgbaColor1 = encodingTry.m_frgbaColor1; + m_frgbaColor2 = encodingTry.m_frgbaColor2; + m_frgbaColor3 = encodingTry.m_frgbaColor3; + + for (unsigned int uiPixel = 0; uiPixel < PIXELS; uiPixel++) + { + m_afrgbaDecodedColors[uiPixel] = encodingTry.m_afrgbaDecodedColors[uiPixel]; + } + + m_fError = encodingTry.m_fError; + + boolImprovement = true; + } + } + } + } + + return boolImprovement; + } + + // ---------------------------------------------------------------------------------------------------- + // try different corner colors by slightly changing G + // + bool Block4x4Encoding_RGB8::TwiddlePlanarG() + { + bool boolImprovement = false; + + Block4x4Encoding_RGB8 encodingTry = *this; + + // init "try" + { + encodingTry.m_mode = MODE_PLANAR; + encodingTry.m_boolDiff = true; + encodingTry.m_boolFlip = false; + } + + int iOriginGreen = encodingTry.m_frgbaColor1.IntGreen(127.0f); + int iHorizGreen = encodingTry.m_frgbaColor2.IntGreen(127.0f); + int iVertGreen = encodingTry.m_frgbaColor3.IntGreen(127.0f); + + for (int iTryOriginGreen = iOriginGreen - 1; iTryOriginGreen <= iOriginGreen + 1; iTryOriginGreen++) + { + // check for out of range + if (iTryOriginGreen < 0 || iTryOriginGreen > 127) + { + continue; + } + + encodingTry.m_frgbaColor1.fG = ((iTryOriginGreen << 1) + (iTryOriginGreen >> 6)) / 255.0f; + + for (int iTryHorizGreen = iHorizGreen - 1; iTryHorizGreen <= iHorizGreen + 1; iTryHorizGreen++) + { + // check for out of range + if (iTryHorizGreen < 0 || iTryHorizGreen > 127) + { + continue; + } + + encodingTry.m_frgbaColor2.fG = ((iTryHorizGreen << 1) + (iTryHorizGreen >> 6)) / 255.0f; + + for (int iTryVertGreen = iVertGreen - 1; iTryVertGreen <= iVertGreen + 1; iTryVertGreen++) + { + // check for out of range + if (iTryVertGreen < 0 || iTryVertGreen > 127) + { + continue; + } + + // don't bother with null twiddle + if (iTryOriginGreen == iOriginGreen && + iTryHorizGreen == iHorizGreen && + iTryVertGreen == iVertGreen) + { + continue; + } + + encodingTry.m_frgbaColor3.fG = ((iTryVertGreen << 1) + (iTryVertGreen >> 6)) / 255.0f; + + encodingTry.DecodePixels_Planar(); + + encodingTry.CalcBlockError(); + + if (encodingTry.m_fError < m_fError) + { + m_mode = MODE_PLANAR; + m_boolDiff = true; + m_boolFlip = false; + m_frgbaColor1 = encodingTry.m_frgbaColor1; + m_frgbaColor2 = encodingTry.m_frgbaColor2; + m_frgbaColor3 = encodingTry.m_frgbaColor3; + + for (unsigned int uiPixel = 0; uiPixel < PIXELS; uiPixel++) + { + m_afrgbaDecodedColors[uiPixel] = encodingTry.m_afrgbaDecodedColors[uiPixel]; + } + + m_fError = encodingTry.m_fError; + + boolImprovement = true; + } + } + } + } + + return boolImprovement; + } + + // ---------------------------------------------------------------------------------------------------- + // try different corner colors by slightly changing B + // + bool Block4x4Encoding_RGB8::TwiddlePlanarB() + { + bool boolImprovement = false; + + Block4x4Encoding_RGB8 encodingTry = *this; + + // init "try" + { + encodingTry.m_mode = MODE_PLANAR; + encodingTry.m_boolDiff = true; + encodingTry.m_boolFlip = false; + } + + int iOriginBlue = encodingTry.m_frgbaColor1.IntBlue(63.0f); + int iHorizBlue = encodingTry.m_frgbaColor2.IntBlue(63.0f); + int iVertBlue = encodingTry.m_frgbaColor3.IntBlue(63.0f); + + for (int iTryOriginBlue = iOriginBlue - 1; iTryOriginBlue <= iOriginBlue + 1; iTryOriginBlue++) + { + // check for out of range + if (iTryOriginBlue < 0 || iTryOriginBlue > 63) + { + continue; + } + + encodingTry.m_frgbaColor1.fB = ((iTryOriginBlue << 2) + (iTryOriginBlue >> 4)) / 255.0f; + + for (int iTryHorizBlue = iHorizBlue - 1; iTryHorizBlue <= iHorizBlue + 1; iTryHorizBlue++) + { + // check for out of range + if (iTryHorizBlue < 0 || iTryHorizBlue > 63) + { + continue; + } + + encodingTry.m_frgbaColor2.fB = ((iTryHorizBlue << 2) + (iTryHorizBlue >> 4)) / 255.0f; + + for (int iTryVertBlue = iVertBlue - 1; iTryVertBlue <= iVertBlue + 1; iTryVertBlue++) + { + // check for out of range + if (iTryVertBlue < 0 || iTryVertBlue > 63) + { + continue; + } + + // don't bother with null twiddle + if (iTryOriginBlue == iOriginBlue && iTryHorizBlue == iHorizBlue && iTryVertBlue == iVertBlue) + { + continue; + } + + encodingTry.m_frgbaColor3.fB = ((iTryVertBlue << 2) + (iTryVertBlue >> 4)) / 255.0f; + + encodingTry.DecodePixels_Planar(); + + encodingTry.CalcBlockError(); + + if (encodingTry.m_fError < m_fError) + { + m_mode = MODE_PLANAR; + m_boolDiff = true; + m_boolFlip = false; + m_frgbaColor1 = encodingTry.m_frgbaColor1; + m_frgbaColor2 = encodingTry.m_frgbaColor2; + m_frgbaColor3 = encodingTry.m_frgbaColor3; + + for (unsigned int uiPixel = 0; uiPixel < PIXELS; uiPixel++) + { + m_afrgbaDecodedColors[uiPixel] = encodingTry.m_afrgbaDecodedColors[uiPixel]; + } + + m_fError = encodingTry.m_fError; + + boolImprovement = true; + } + } + } + } + + return boolImprovement; + } + + // ---------------------------------------------------------------------------------------------------- + // set the encoding bits based on encoding state + // + void Block4x4Encoding_RGB8::SetEncodingBits(void) + { + + switch (m_mode) + { + case MODE_ETC1: + Block4x4Encoding_ETC1::SetEncodingBits(); + break; + + case MODE_T: + SetEncodingBits_T(); + break; + + case MODE_H: + SetEncodingBits_H(); + break; + + case MODE_PLANAR: + SetEncodingBits_Planar(); + break; + + default: + assert(false); + } + + } + + // ---------------------------------------------------------------------------------------------------- + // set the encoding bits based on encoding state for T mode + // + void Block4x4Encoding_RGB8::SetEncodingBits_T(void) + { + static const bool SANITY_CHECK = true; + + assert(m_mode == MODE_T); + assert(m_boolDiff == true); + + unsigned int uiRed1 = (unsigned int)m_frgbaColor1.IntRed(15.0f); + unsigned int uiGreen1 = (unsigned int)m_frgbaColor1.IntGreen(15.0f); + unsigned int uiBlue1 = (unsigned int)m_frgbaColor1.IntBlue(15.0f); + + unsigned int uiRed2 = (unsigned int)m_frgbaColor2.IntRed(15.0f); + unsigned int uiGreen2 = (unsigned int)m_frgbaColor2.IntGreen(15.0f); + unsigned int uiBlue2 = (unsigned int)m_frgbaColor2.IntBlue(15.0f); + + m_pencodingbitsRGB8->t.red1a = uiRed1 >> 2; + m_pencodingbitsRGB8->t.red1b = uiRed1; + m_pencodingbitsRGB8->t.green1 = uiGreen1; + m_pencodingbitsRGB8->t.blue1 = uiBlue1; + + m_pencodingbitsRGB8->t.red2 = uiRed2; + m_pencodingbitsRGB8->t.green2 = uiGreen2; + m_pencodingbitsRGB8->t.blue2 = uiBlue2; + + m_pencodingbitsRGB8->t.da = m_uiCW1 >> 1; + m_pencodingbitsRGB8->t.db = m_uiCW1; + + m_pencodingbitsRGB8->t.diff = 1; + + Block4x4Encoding_ETC1::SetEncodingBits_Selectors(); + + // create an invalid R differential to trigger T mode + m_pencodingbitsRGB8->t.detect1 = 0; + m_pencodingbitsRGB8->t.detect2 = 0; + int iRed2 = (int)m_pencodingbitsRGB8->differential.red1 + (int)m_pencodingbitsRGB8->differential.dred2; + if (iRed2 >= 4) + { + m_pencodingbitsRGB8->t.detect1 = 7; + m_pencodingbitsRGB8->t.detect2 = 0; + } + else + { + m_pencodingbitsRGB8->t.detect1 = 0; + m_pencodingbitsRGB8->t.detect2 = 1; + } + + if (SANITY_CHECK) + { + iRed2 = (int)m_pencodingbitsRGB8->differential.red1 + (int)m_pencodingbitsRGB8->differential.dred2; + + // make sure red overflows + assert(iRed2 < 0 || iRed2 > 31); + } + + } + + // ---------------------------------------------------------------------------------------------------- + // set the encoding bits based on encoding state for H mode + // + // colors and selectors may need to swap in order to generate lsb of distance index + // + void Block4x4Encoding_RGB8::SetEncodingBits_H(void) + { + static const bool SANITY_CHECK = true; + + assert(m_mode == MODE_H); + assert(m_boolDiff == true); + + unsigned int uiRed1 = (unsigned int)m_frgbaColor1.IntRed(15.0f); + unsigned int uiGreen1 = (unsigned int)m_frgbaColor1.IntGreen(15.0f); + unsigned int uiBlue1 = (unsigned int)m_frgbaColor1.IntBlue(15.0f); + + unsigned int uiRed2 = (unsigned int)m_frgbaColor2.IntRed(15.0f); + unsigned int uiGreen2 = (unsigned int)m_frgbaColor2.IntGreen(15.0f); + unsigned int uiBlue2 = (unsigned int)m_frgbaColor2.IntBlue(15.0f); + + unsigned int uiColor1 = (uiRed1 << 16) + (uiGreen1 << 8) + uiBlue1; + unsigned int uiColor2 = (uiRed2 << 16) + (uiGreen2 << 8) + uiBlue2; + + bool boolOddDistance = m_uiCW1 & 1; + bool boolSwapColors = (uiColor1 < uiColor2) ^ !boolOddDistance; + + if (boolSwapColors) + { + m_pencodingbitsRGB8->h.red1 = uiRed2; + m_pencodingbitsRGB8->h.green1a = uiGreen2 >> 1; + m_pencodingbitsRGB8->h.green1b = uiGreen2; + m_pencodingbitsRGB8->h.blue1a = uiBlue2 >> 3; + m_pencodingbitsRGB8->h.blue1b = uiBlue2 >> 1; + m_pencodingbitsRGB8->h.blue1c = uiBlue2; + + m_pencodingbitsRGB8->h.red2 = uiRed1; + m_pencodingbitsRGB8->h.green2a = uiGreen1 >> 1; + m_pencodingbitsRGB8->h.green2b = uiGreen1; + m_pencodingbitsRGB8->h.blue2 = uiBlue1; + + m_pencodingbitsRGB8->h.da = m_uiCW1 >> 2; + m_pencodingbitsRGB8->h.db = m_uiCW1 >> 1; + } + else + { + m_pencodingbitsRGB8->h.red1 = uiRed1; + m_pencodingbitsRGB8->h.green1a = uiGreen1 >> 1; + m_pencodingbitsRGB8->h.green1b = uiGreen1; + m_pencodingbitsRGB8->h.blue1a = uiBlue1 >> 3; + m_pencodingbitsRGB8->h.blue1b = uiBlue1 >> 1; + m_pencodingbitsRGB8->h.blue1c = uiBlue1; + + m_pencodingbitsRGB8->h.red2 = uiRed2; + m_pencodingbitsRGB8->h.green2a = uiGreen2 >> 1; + m_pencodingbitsRGB8->h.green2b = uiGreen2; + m_pencodingbitsRGB8->h.blue2 = uiBlue2; + + m_pencodingbitsRGB8->h.da = m_uiCW1 >> 2; + m_pencodingbitsRGB8->h.db = m_uiCW1 >> 1; + } + + m_pencodingbitsRGB8->h.diff = 1; + + Block4x4Encoding_ETC1::SetEncodingBits_Selectors(); + + if (boolSwapColors) + { + m_pencodingbitsRGB8->h.selectors ^= 0x0000FFFF; + } + + // create an invalid R differential to trigger T mode + m_pencodingbitsRGB8->h.detect1 = 0; + m_pencodingbitsRGB8->h.detect2 = 0; + m_pencodingbitsRGB8->h.detect3 = 0; + int iRed2 = (int)m_pencodingbitsRGB8->differential.red1 + (int)m_pencodingbitsRGB8->differential.dred2; + int iGreen2 = (int)m_pencodingbitsRGB8->differential.green1 + (int)m_pencodingbitsRGB8->differential.dgreen2; + if (iRed2 < 0 || iRed2 > 31) + { + m_pencodingbitsRGB8->h.detect1 = 1; + } + if (iGreen2 >= 4) + { + m_pencodingbitsRGB8->h.detect2 = 7; + m_pencodingbitsRGB8->h.detect3 = 0; + } + else + { + m_pencodingbitsRGB8->h.detect2 = 0; + m_pencodingbitsRGB8->h.detect3 = 1; + } + + if (SANITY_CHECK) + { + iRed2 = (int)m_pencodingbitsRGB8->differential.red1 + (int)m_pencodingbitsRGB8->differential.dred2; + iGreen2 = (int)m_pencodingbitsRGB8->differential.green1 + (int)m_pencodingbitsRGB8->differential.dgreen2; + + // make sure red doesn't overflow and green does + assert(iRed2 >= 0 && iRed2 <= 31); + assert(iGreen2 < 0 || iGreen2 > 31); + } + + } + + // ---------------------------------------------------------------------------------------------------- + // set the encoding bits based on encoding state for Planar mode + // + void Block4x4Encoding_RGB8::SetEncodingBits_Planar(void) + { + static const bool SANITY_CHECK = true; + + assert(m_mode == MODE_PLANAR); + assert(m_boolDiff == true); + + unsigned int uiOriginRed = (unsigned int)m_frgbaColor1.IntRed(63.0f); + unsigned int uiOriginGreen = (unsigned int)m_frgbaColor1.IntGreen(127.0f); + unsigned int uiOriginBlue = (unsigned int)m_frgbaColor1.IntBlue(63.0f); + + unsigned int uiHorizRed = (unsigned int)m_frgbaColor2.IntRed(63.0f); + unsigned int uiHorizGreen = (unsigned int)m_frgbaColor2.IntGreen(127.0f); + unsigned int uiHorizBlue = (unsigned int)m_frgbaColor2.IntBlue(63.0f); + + unsigned int uiVertRed = (unsigned int)m_frgbaColor3.IntRed(63.0f); + unsigned int uiVertGreen = (unsigned int)m_frgbaColor3.IntGreen(127.0f); + unsigned int uiVertBlue = (unsigned int)m_frgbaColor3.IntBlue(63.0f); + + m_pencodingbitsRGB8->planar.originRed = uiOriginRed; + m_pencodingbitsRGB8->planar.originGreen1 = uiOriginGreen >> 6; + m_pencodingbitsRGB8->planar.originGreen2 = uiOriginGreen; + m_pencodingbitsRGB8->planar.originBlue1 = uiOriginBlue >> 5; + m_pencodingbitsRGB8->planar.originBlue2 = uiOriginBlue >> 3; + m_pencodingbitsRGB8->planar.originBlue3 = uiOriginBlue >> 1; + m_pencodingbitsRGB8->planar.originBlue4 = uiOriginBlue; + + m_pencodingbitsRGB8->planar.horizRed1 = uiHorizRed >> 1; + m_pencodingbitsRGB8->planar.horizRed2 = uiHorizRed; + m_pencodingbitsRGB8->planar.horizGreen = uiHorizGreen; + m_pencodingbitsRGB8->planar.horizBlue1 = uiHorizBlue >> 5; + m_pencodingbitsRGB8->planar.horizBlue2 = uiHorizBlue; + + m_pencodingbitsRGB8->planar.vertRed1 = uiVertRed >> 3; + m_pencodingbitsRGB8->planar.vertRed2 = uiVertRed; + m_pencodingbitsRGB8->planar.vertGreen1 = uiVertGreen >> 2; + m_pencodingbitsRGB8->planar.vertGreen2 = uiVertGreen; + m_pencodingbitsRGB8->planar.vertBlue = uiVertBlue; + + m_pencodingbitsRGB8->planar.diff = 1; + + // create valid RG differentials and an invalid B differential to trigger planar mode + m_pencodingbitsRGB8->planar.detect1 = 0; + m_pencodingbitsRGB8->planar.detect2 = 0; + m_pencodingbitsRGB8->planar.detect3 = 0; + m_pencodingbitsRGB8->planar.detect4 = 0; + int iRed2 = (int)m_pencodingbitsRGB8->differential.red1 + (int)m_pencodingbitsRGB8->differential.dred2; + int iGreen2 = (int)m_pencodingbitsRGB8->differential.green1 + (int)m_pencodingbitsRGB8->differential.dgreen2; + int iBlue2 = (int)m_pencodingbitsRGB8->differential.blue1 + (int)m_pencodingbitsRGB8->differential.dblue2; + if (iRed2 < 0 || iRed2 > 31) + { + m_pencodingbitsRGB8->planar.detect1 = 1; + } + if (iGreen2 < 0 || iGreen2 > 31) + { + m_pencodingbitsRGB8->planar.detect2 = 1; + } + if (iBlue2 >= 4) + { + m_pencodingbitsRGB8->planar.detect3 = 7; + m_pencodingbitsRGB8->planar.detect4 = 0; + } + else + { + m_pencodingbitsRGB8->planar.detect3 = 0; + m_pencodingbitsRGB8->planar.detect4 = 1; + } + + if (SANITY_CHECK) + { + iRed2 = (int)m_pencodingbitsRGB8->differential.red1 + (int)m_pencodingbitsRGB8->differential.dred2; + iGreen2 = (int)m_pencodingbitsRGB8->differential.green1 + (int)m_pencodingbitsRGB8->differential.dgreen2; + iBlue2 = (int)m_pencodingbitsRGB8->differential.blue1 + (int)m_pencodingbitsRGB8->differential.dblue2; + + // make sure red and green don't overflow and blue does + assert(iRed2 >= 0 && iRed2 <= 31); + assert(iGreen2 >= 0 && iGreen2 <= 31); + assert(iBlue2 < 0 || iBlue2 > 31); + } + + } + + // ---------------------------------------------------------------------------------------------------- + // set the decoded colors and decoded alpha based on the encoding state for T mode + // + void Block4x4Encoding_RGB8::DecodePixels_T(void) + { + + float fDistance = s_afTHDistanceTable[m_uiCW1]; + ColorFloatRGBA frgbaDistance(fDistance, fDistance, fDistance, 0.0f); + + for (unsigned int uiPixel = 0; uiPixel < PIXELS; uiPixel++) + { + switch (m_auiSelectors[uiPixel]) + { + case 0: + m_afrgbaDecodedColors[uiPixel] = m_frgbaColor1; + break; + + case 1: + m_afrgbaDecodedColors[uiPixel] = (m_frgbaColor2 + frgbaDistance).ClampRGB(); + break; + + case 2: + m_afrgbaDecodedColors[uiPixel] = m_frgbaColor2; + break; + + case 3: + m_afrgbaDecodedColors[uiPixel] = (m_frgbaColor2 - frgbaDistance).ClampRGB(); + break; + } + + } + + } + + // ---------------------------------------------------------------------------------------------------- + // set the decoded colors and decoded alpha based on the encoding state for H mode + // + void Block4x4Encoding_RGB8::DecodePixels_H(void) + { + + float fDistance = s_afTHDistanceTable[m_uiCW1]; + ColorFloatRGBA frgbaDistance(fDistance, fDistance, fDistance, 0.0f); + + for (unsigned int uiPixel = 0; uiPixel < PIXELS; uiPixel++) + { + switch (m_auiSelectors[uiPixel]) + { + case 0: + m_afrgbaDecodedColors[uiPixel] = (m_frgbaColor1 + frgbaDistance).ClampRGB(); + break; + + case 1: + m_afrgbaDecodedColors[uiPixel] = (m_frgbaColor1 - frgbaDistance).ClampRGB(); + break; + + case 2: + m_afrgbaDecodedColors[uiPixel] = (m_frgbaColor2 + frgbaDistance).ClampRGB(); + break; + + case 3: + m_afrgbaDecodedColors[uiPixel] = (m_frgbaColor2 - frgbaDistance).ClampRGB(); + break; + } + + } + + } + + // ---------------------------------------------------------------------------------------------------- + // set the decoded colors and decoded alpha based on the encoding state for Planar mode + // + void Block4x4Encoding_RGB8::DecodePixels_Planar(void) + { + + int iRO = (int)roundf(m_frgbaColor1.fR * 255.0f); + int iGO = (int)roundf(m_frgbaColor1.fG * 255.0f); + int iBO = (int)roundf(m_frgbaColor1.fB * 255.0f); + + int iRH = (int)roundf(m_frgbaColor2.fR * 255.0f); + int iGH = (int)roundf(m_frgbaColor2.fG * 255.0f); + int iBH = (int)roundf(m_frgbaColor2.fB * 255.0f); + + int iRV = (int)roundf(m_frgbaColor3.fR * 255.0f); + int iGV = (int)roundf(m_frgbaColor3.fG * 255.0f); + int iBV = (int)roundf(m_frgbaColor3.fB * 255.0f); + + for (unsigned int uiPixel = 0; uiPixel < PIXELS; uiPixel++) + { + int iX = (int)(uiPixel >> 2); + int iY = (int)(uiPixel & 3); + + int iR = (iX*(iRH - iRO) + iY*(iRV - iRO) + 4*iRO + 2) >> 2; + int iG = (iX*(iGH - iGO) + iY*(iGV - iGO) + 4*iGO + 2) >> 2; + int iB = (iX*(iBH - iBO) + iY*(iBV - iBO) + 4*iBO + 2) >> 2; + + ColorFloatRGBA frgba; + frgba.fR = (float)iR / 255.0f; + frgba.fG = (float)iG / 255.0f; + frgba.fB = (float)iB / 255.0f; + frgba.fA = 1.0f; + + m_afrgbaDecodedColors[uiPixel] = frgba.ClampRGB(); + } + + } + + // ---------------------------------------------------------------------------------------------------- + // perform a linear regression for the a_uiPixels in a_pafrgbaPixels[] + // + // output the closest color line using a_pfrgbaSlope and a_pfrgbaOffset + // + void Block4x4Encoding_RGB8::ColorRegression(ColorFloatRGBA *a_pafrgbaPixels, unsigned int a_uiPixels, + ColorFloatRGBA *a_pfrgbaSlope, ColorFloatRGBA *a_pfrgbaOffset) + { + typedef struct + { + float f[4]; + } Float4; + + Float4 *paf4Pixels = (Float4 *)(a_pafrgbaPixels); + Float4 *pf4Slope = (Float4 *)(a_pfrgbaSlope); + Float4 *pf4Offset = (Float4 *)(a_pfrgbaOffset); + + float afX[MAX_PLANAR_REGRESSION_SIZE]; + float afY[MAX_PLANAR_REGRESSION_SIZE]; + + // handle r, g and b separately. don't bother with a + for (unsigned int uiComponent = 0; uiComponent < 3; uiComponent++) + { + for (unsigned int uiPixel = 0; uiPixel < a_uiPixels; uiPixel++) + { + afX[uiPixel] = (float)uiPixel; + afY[uiPixel] = paf4Pixels[uiPixel].f[uiComponent]; + Etc::Regression(afX, afY, a_uiPixels, + &(pf4Slope->f[uiComponent]), &(pf4Offset->f[uiComponent])); + } + } + + } + + // ---------------------------------------------------------------------------------------------------- + // +} diff --git a/EtcLib/EtcCodec/EtcBlock4x4Encoding_RGB8.h b/EtcLib/EtcCodec/EtcBlock4x4Encoding_RGB8.h new file mode 100644 index 0000000..e405223 --- /dev/null +++ b/EtcLib/EtcCodec/EtcBlock4x4Encoding_RGB8.h @@ -0,0 +1,96 @@ +/* + * Copyright 2015 The Etc2Comp Authors. + * + * Licensed under the Apache License, Version 2.0 (the "License"); + * you may not use this file except in compliance with the License. + * You may obtain a copy of the License at + * + * http://www.apache.org/licenses/LICENSE-2.0 + * + * Unless required by applicable law or agreed to in writing, software + * distributed under the License is distributed on an "AS IS" BASIS, + * WITHOUT WARRANTIES OR CONDITIONS OF ANY KIND, either express or implied. + * See the License for the specific language governing permissions and + * limitations under the License. + */ + +#pragma once + +#include "EtcBlock4x4Encoding_ETC1.h" + +namespace Etc +{ + + class Block4x4Encoding_RGB8 : public Block4x4Encoding_ETC1 + { + public: + + Block4x4Encoding_RGB8(void); + virtual ~Block4x4Encoding_RGB8(void); + + virtual void InitFromEncodingBits(Block4x4 *a_pblockParent, + unsigned char *a_paucEncodingBits, + ColorFloatRGBA *a_pafrgbaSource, + + ErrorMetric a_errormetric); + + virtual void PerformIteration(float a_fEffort); + + virtual void SetEncodingBits(void); + + inline ColorFloatRGBA GetColor3(void) const + { + return m_frgbaColor3; + } + + protected: + + static const unsigned int PLANAR_CORNER_COLORS = 3; + static const unsigned int MAX_PLANAR_REGRESSION_SIZE = 4; + static const unsigned int TH_DISTANCES = 8; + + static float s_afTHDistanceTable[TH_DISTANCES]; + + void TryPlanar(unsigned int a_uiRadius); + void TryTAndH(unsigned int a_uiRadius); + + void InitFromEncodingBits_Planar(void); + + ColorFloatRGBA m_frgbaColor3; // used for planar + + void SetEncodingBits_T(void); + void SetEncodingBits_H(void); + void SetEncodingBits_Planar(void); + + // state shared between iterations + ColorFloatRGBA m_frgbaOriginalColor1_TAndH; + ColorFloatRGBA m_frgbaOriginalColor2_TAndH; + + void CalculateBaseColorsForTAndH(void); + void TryT(unsigned int a_uiRadius); + void TryT_BestSelectorCombination(void); + void TryH(unsigned int a_uiRadius); + void TryH_BestSelectorCombination(void); + + private: + + void InitFromEncodingBits_T(void); + void InitFromEncodingBits_H(void); + + void CalculatePlanarCornerColors(void); + + void ColorRegression(ColorFloatRGBA *a_pafrgbaPixels, unsigned int a_uiPixels, + ColorFloatRGBA *a_pfrgbaSlope, ColorFloatRGBA *a_pfrgbaOffset); + + bool TwiddlePlanar(void); + bool TwiddlePlanarR(); + bool TwiddlePlanarG(); + bool TwiddlePlanarB(); + + void DecodePixels_T(void); + void DecodePixels_H(void); + void DecodePixels_Planar(void); + + }; + +} // namespace Etc diff --git a/EtcLib/EtcCodec/EtcBlock4x4Encoding_RGB8A1.cpp b/EtcLib/EtcCodec/EtcBlock4x4Encoding_RGB8A1.cpp new file mode 100644 index 0000000..fc0678d --- /dev/null +++ b/EtcLib/EtcCodec/EtcBlock4x4Encoding_RGB8A1.cpp @@ -0,0 +1,1819 @@ +/* + * Copyright 2015 The Etc2Comp Authors. + * + * Licensed under the Apache License, Version 2.0 (the "License"); + * you may not use this file except in compliance with the License. + * You may obtain a copy of the License at + * + * http://www.apache.org/licenses/LICENSE-2.0 + * + * Unless required by applicable law or agreed to in writing, software + * distributed under the License is distributed on an "AS IS" BASIS, + * WITHOUT WARRANTIES OR CONDITIONS OF ANY KIND, either express or implied. + * See the License for the specific language governing permissions and + * limitations under the License. + */ + +/* +EtcBlock4x4Encoding_RGB8A1.cpp contains: + Block4x4Encoding_RGB8A1 + Block4x4Encoding_RGB8A1_Opaque + Block4x4Encoding_RGB8A1_Transparent + +These encoders are used when targetting file format RGB8A1. + +Block4x4Encoding_RGB8A1_Opaque is used when all pixels in the 4x4 block are opaque +Block4x4Encoding_RGB8A1_Transparent is used when all pixels in the 4x4 block are transparent +Block4x4Encoding_RGB8A1 is used when there is a mixture of alphas in the 4x4 block + +*/ + +#include "EtcConfig.h" +#include "EtcBlock4x4Encoding_RGB8A1.h" + +#include "EtcBlock4x4.h" +#include "EtcBlock4x4EncodingBits.h" +#include "EtcBlock4x4Encoding_RGB8.h" + +#include +#include +#include + +namespace Etc +{ + + // #################################################################################################### + // Block4x4Encoding_RGB8A1 + // #################################################################################################### + + float Block4x4Encoding_RGB8A1::s_aafCwOpaqueUnsetTable[CW_RANGES][SELECTORS] = + { + { 0.0f / 255.0f, 8.0f / 255.0f, 0.0f / 255.0f, -8.0f / 255.0f }, + { 0.0f / 255.0f, 17.0f / 255.0f, 0.0f / 255.0f, -17.0f / 255.0f }, + { 0.0f / 255.0f, 29.0f / 255.0f, 0.0f / 255.0f, -29.0f / 255.0f }, + { 0.0f / 255.0f, 42.0f / 255.0f, 0.0f / 255.0f, -42.0f / 255.0f }, + { 0.0f / 255.0f, 60.0f / 255.0f, 0.0f / 255.0f, -60.0f / 255.0f }, + { 0.0f / 255.0f, 80.0f / 255.0f, 0.0f / 255.0f, -80.0f / 255.0f }, + { 0.0f / 255.0f, 106.0f / 255.0f, 0.0f / 255.0f, -106.0f / 255.0f }, + { 0.0f / 255.0f, 183.0f / 255.0f, 0.0f / 255.0f, -183.0f / 255.0f } + }; + + // ---------------------------------------------------------------------------------------------------- + // + Block4x4Encoding_RGB8A1::Block4x4Encoding_RGB8A1(void) + { + m_pencodingbitsRGB8 = nullptr; + m_boolOpaque = false; + m_boolTransparent = false; + m_boolPunchThroughPixels = true; + + } + Block4x4Encoding_RGB8A1::~Block4x4Encoding_RGB8A1(void) {} + // ---------------------------------------------------------------------------------------------------- + // initialization prior to encoding + // a_pblockParent points to the block associated with this encoding + // a_errormetric is used to choose the best encoding + // a_pafrgbaSource points to a 4x4 block subset of the source image + // a_paucEncodingBits points to the final encoding bits + // + void Block4x4Encoding_RGB8A1::InitFromSource(Block4x4 *a_pblockParent, + ColorFloatRGBA *a_pafrgbaSource, + unsigned char *a_paucEncodingBits, + ErrorMetric a_errormetric) + { + + Block4x4Encoding_RGB8::InitFromSource(a_pblockParent, + a_pafrgbaSource, + a_paucEncodingBits, + a_errormetric); + + m_boolOpaque = a_pblockParent->GetSourceAlphaMix() == Block4x4::SourceAlphaMix::OPAQUE; + m_boolTransparent = a_pblockParent->GetSourceAlphaMix() == Block4x4::SourceAlphaMix::TRANSPARENT; + m_boolPunchThroughPixels = a_pblockParent->HasPunchThroughPixels(); + + for (unsigned int uiPixel = 0; uiPixel < PIXELS; uiPixel++) + { + if (m_pafrgbaSource[uiPixel].fA >= 0.5f) + { + m_afDecodedAlphas[uiPixel] = 1.0f; + } + else + { + m_afDecodedAlphas[uiPixel] = 0.0f; + } + } + + } + + // ---------------------------------------------------------------------------------------------------- + // initialization from the encoding bits of a previous encoding + // a_pblockParent points to the block associated with this encoding + // a_errormetric is used to choose the best encoding + // a_pafrgbaSource points to a 4x4 block subset of the source image + // a_paucEncodingBits points to the final encoding bits of a previous encoding + // + void Block4x4Encoding_RGB8A1::InitFromEncodingBits(Block4x4 *a_pblockParent, + unsigned char *a_paucEncodingBits, + ColorFloatRGBA *a_pafrgbaSource, + ErrorMetric a_errormetric) + { + + + InitFromEncodingBits_ETC1(a_pblockParent, + a_paucEncodingBits, + a_pafrgbaSource, + a_errormetric); + + m_pencodingbitsRGB8 = (Block4x4EncodingBits_RGB8 *)a_paucEncodingBits; + + // detect if there is a T, H or Planar mode present + int iRed1 = m_pencodingbitsRGB8->differential.red1; + int iDRed2 = m_pencodingbitsRGB8->differential.dred2; + int iRed2 = iRed1 + iDRed2; + + int iGreen1 = m_pencodingbitsRGB8->differential.green1; + int iDGreen2 = m_pencodingbitsRGB8->differential.dgreen2; + int iGreen2 = iGreen1 + iDGreen2; + + int iBlue1 = m_pencodingbitsRGB8->differential.blue1; + int iDBlue2 = m_pencodingbitsRGB8->differential.dblue2; + int iBlue2 = iBlue1 + iDBlue2; + + if (iRed2 < 0 || iRed2 > 31) + { + InitFromEncodingBits_T(); + } + else if (iGreen2 < 0 || iGreen2 > 31) + { + InitFromEncodingBits_H(); + } + else if (iBlue2 < 0 || iBlue2 > 31) + { + Block4x4Encoding_RGB8::InitFromEncodingBits_Planar(); + } + + } + + // ---------------------------------------------------------------------------------------------------- + // initialization from the encoding bits of a previous encoding assuming the encoding is an ETC1 mode. + // if it isn't an ETC1 mode, this will be overwritten later + // + void Block4x4Encoding_RGB8A1::InitFromEncodingBits_ETC1(Block4x4 *a_pblockParent, + unsigned char *a_paucEncodingBits, + ColorFloatRGBA *a_pafrgbaSource, + ErrorMetric a_errormetric) + { + Block4x4Encoding::Init(a_pblockParent, a_pafrgbaSource, + a_errormetric); + + m_pencodingbitsRGB8 = (Block4x4EncodingBits_RGB8 *)a_paucEncodingBits; + + m_mode = MODE_ETC1; + m_boolDiff = true; + m_boolFlip = m_pencodingbitsRGB8->differential.flip; + m_boolOpaque = m_pencodingbitsRGB8->differential.diff; + + int iR2 = m_pencodingbitsRGB8->differential.red1 + m_pencodingbitsRGB8->differential.dred2; + if (iR2 < 0) + { + iR2 = 0; + } + else if (iR2 > 31) + { + iR2 = 31; + } + + int iG2 = m_pencodingbitsRGB8->differential.green1 + m_pencodingbitsRGB8->differential.dgreen2; + if (iG2 < 0) + { + iG2 = 0; + } + else if (iG2 > 31) + { + iG2 = 31; + } + + int iB2 = m_pencodingbitsRGB8->differential.blue1 + m_pencodingbitsRGB8->differential.dblue2; + if (iB2 < 0) + { + iB2 = 0; + } + else if (iB2 > 31) + { + iB2 = 31; + } + + m_frgbaColor1 = ColorFloatRGBA::ConvertFromRGB5(m_pencodingbitsRGB8->differential.red1, m_pencodingbitsRGB8->differential.green1, m_pencodingbitsRGB8->differential.blue1); + m_frgbaColor2 = ColorFloatRGBA::ConvertFromRGB5((unsigned char)iR2, (unsigned char)iG2, (unsigned char)iB2); + + m_uiCW1 = m_pencodingbitsRGB8->differential.cw1; + m_uiCW2 = m_pencodingbitsRGB8->differential.cw2; + + Block4x4Encoding_ETC1::InitFromEncodingBits_Selectors(); + + Decode_ETC1(); + + CalcBlockError(); + + } + + // ---------------------------------------------------------------------------------------------------- + // initialization from the encoding bits of a previous encoding if T mode is detected + // + void Block4x4Encoding_RGB8A1::InitFromEncodingBits_T(void) + { + m_mode = MODE_T; + + unsigned char ucRed1 = (unsigned char)((m_pencodingbitsRGB8->t.red1a << 2) + + m_pencodingbitsRGB8->t.red1b); + unsigned char ucGreen1 = m_pencodingbitsRGB8->t.green1; + unsigned char ucBlue1 = m_pencodingbitsRGB8->t.blue1; + + unsigned char ucRed2 = m_pencodingbitsRGB8->t.red2; + unsigned char ucGreen2 = m_pencodingbitsRGB8->t.green2; + unsigned char ucBlue2 = m_pencodingbitsRGB8->t.blue2; + + m_frgbaColor1 = ColorFloatRGBA::ConvertFromRGB4(ucRed1, ucGreen1, ucBlue1); + m_frgbaColor2 = ColorFloatRGBA::ConvertFromRGB4(ucRed2, ucGreen2, ucBlue2); + + m_uiCW1 = (m_pencodingbitsRGB8->t.da << 1) + m_pencodingbitsRGB8->t.db; + + Block4x4Encoding_ETC1::InitFromEncodingBits_Selectors(); + + DecodePixels_T(); + + CalcBlockError(); + } + + // ---------------------------------------------------------------------------------------------------- + // initialization from the encoding bits of a previous encoding if H mode is detected + // + void Block4x4Encoding_RGB8A1::InitFromEncodingBits_H(void) + { + m_mode = MODE_H; + + unsigned char ucRed1 = m_pencodingbitsRGB8->h.red1; + unsigned char ucGreen1 = (unsigned char)((m_pencodingbitsRGB8->h.green1a << 1) + + m_pencodingbitsRGB8->h.green1b); + unsigned char ucBlue1 = (unsigned char)((m_pencodingbitsRGB8->h.blue1a << 3) + + (m_pencodingbitsRGB8->h.blue1b << 1) + + m_pencodingbitsRGB8->h.blue1c); + + unsigned char ucRed2 = m_pencodingbitsRGB8->h.red2; + unsigned char ucGreen2 = (unsigned char)((m_pencodingbitsRGB8->h.green2a << 1) + + m_pencodingbitsRGB8->h.green2b); + unsigned char ucBlue2 = m_pencodingbitsRGB8->h.blue2; + + m_frgbaColor1 = ColorFloatRGBA::ConvertFromRGB4(ucRed1, ucGreen1, ucBlue1); + m_frgbaColor2 = ColorFloatRGBA::ConvertFromRGB4(ucRed2, ucGreen2, ucBlue2); + + // used to determine the LSB of the CW + unsigned int uiRGB1 = (unsigned int)(((int)ucRed1 << 16) + ((int)ucGreen1 << 8) + (int)ucBlue1); + unsigned int uiRGB2 = (unsigned int)(((int)ucRed2 << 16) + ((int)ucGreen2 << 8) + (int)ucBlue2); + + m_uiCW1 = (m_pencodingbitsRGB8->h.da << 2) + (m_pencodingbitsRGB8->h.db << 1); + if (uiRGB1 >= uiRGB2) + { + m_uiCW1++; + } + + Block4x4Encoding_ETC1::InitFromEncodingBits_Selectors(); + + DecodePixels_H(); + + CalcBlockError(); + } + + // ---------------------------------------------------------------------------------------------------- + // for ETC1 modes, set the decoded colors and decoded alpha based on the encoding state + // + void Block4x4Encoding_RGB8A1::Decode_ETC1(void) + { + + const unsigned int *pauiPixelOrder = m_boolFlip ? s_auiPixelOrderFlip1 : s_auiPixelOrderFlip0; + + for (unsigned int uiPixelOrder = 0; uiPixelOrder < PIXELS; uiPixelOrder++) + { + ColorFloatRGBA *pfrgbaCenter = uiPixelOrder < 8 ? &m_frgbaColor1 : &m_frgbaColor2; + unsigned int uiCW = uiPixelOrder < 8 ? m_uiCW1 : m_uiCW2; + + unsigned int uiPixel = pauiPixelOrder[uiPixelOrder]; + + float fDelta; + if (m_boolOpaque) + fDelta = Block4x4Encoding_ETC1::s_aafCwTable[uiCW][m_auiSelectors[uiPixel]]; + else + fDelta = s_aafCwOpaqueUnsetTable[uiCW][m_auiSelectors[uiPixel]]; + + if (m_boolOpaque == false && m_auiSelectors[uiPixel] == TRANSPARENT_SELECTOR) + { + m_afrgbaDecodedColors[uiPixel] = ColorFloatRGBA(); + m_afDecodedAlphas[uiPixel] = 0.0f; + } + else + { + m_afrgbaDecodedColors[uiPixel] = (*pfrgbaCenter + fDelta).ClampRGB(); + m_afDecodedAlphas[uiPixel] = 1.0f; + } + } + + } + + // ---------------------------------------------------------------------------------------------------- + // for T mode, set the decoded colors and decoded alpha based on the encoding state + // + void Block4x4Encoding_RGB8A1::DecodePixels_T(void) + { + + float fDistance = s_afTHDistanceTable[m_uiCW1]; + ColorFloatRGBA frgbaDistance(fDistance, fDistance, fDistance, 0.0f); + + for (unsigned int uiPixel = 0; uiPixel < PIXELS; uiPixel++) + { + switch (m_auiSelectors[uiPixel]) + { + case 0: + m_afrgbaDecodedColors[uiPixel] = m_frgbaColor1; + m_afDecodedAlphas[uiPixel] = 1.0f; + break; + + case 1: + m_afrgbaDecodedColors[uiPixel] = (m_frgbaColor2 + frgbaDistance).ClampRGB(); + m_afDecodedAlphas[uiPixel] = 1.0f; + break; + + case 2: + if (m_boolOpaque == false) + { + m_afrgbaDecodedColors[uiPixel] = ColorFloatRGBA(); + m_afDecodedAlphas[uiPixel] = 0.0f; + } + else + { + m_afrgbaDecodedColors[uiPixel] = m_frgbaColor2; + m_afDecodedAlphas[uiPixel] = 1.0f; + } + break; + + case 3: + m_afrgbaDecodedColors[uiPixel] = (m_frgbaColor2 - frgbaDistance).ClampRGB(); + m_afDecodedAlphas[uiPixel] = 1.0f; + break; + } + + } + + } + + // ---------------------------------------------------------------------------------------------------- + // for H mode, set the decoded colors and decoded alpha based on the encoding state + // + void Block4x4Encoding_RGB8A1::DecodePixels_H(void) + { + + float fDistance = s_afTHDistanceTable[m_uiCW1]; + ColorFloatRGBA frgbaDistance(fDistance, fDistance, fDistance, 0.0f); + + for (unsigned int uiPixel = 0; uiPixel < PIXELS; uiPixel++) + { + switch (m_auiSelectors[uiPixel]) + { + case 0: + m_afrgbaDecodedColors[uiPixel] = (m_frgbaColor1 + frgbaDistance).ClampRGB(); + m_afDecodedAlphas[uiPixel] = 1.0f; + break; + + case 1: + m_afrgbaDecodedColors[uiPixel] = (m_frgbaColor1 - frgbaDistance).ClampRGB(); + m_afDecodedAlphas[uiPixel] = 1.0f; + break; + + case 2: + if (m_boolOpaque == false) + { + m_afrgbaDecodedColors[uiPixel] = ColorFloatRGBA(); + m_afDecodedAlphas[uiPixel] = 0.0f; + } + else + { + m_afrgbaDecodedColors[uiPixel] = (m_frgbaColor2 + frgbaDistance).ClampRGB(); + m_afDecodedAlphas[uiPixel] = 1.0f; + } + break; + + case 3: + m_afrgbaDecodedColors[uiPixel] = (m_frgbaColor2 - frgbaDistance).ClampRGB(); + m_afDecodedAlphas[uiPixel] = 1.0f; + break; + } + + } + + } + + + // ---------------------------------------------------------------------------------------------------- + // perform a single encoding iteration + // replace the encoding if a better encoding was found + // subsequent iterations generally take longer for each iteration + // set m_boolDone if encoding is perfect or encoding is finished based on a_fEffort + // + // RGB8A1 can't use individual mode + // RGB8A1 with transparent pixels can't use planar mode + // + void Block4x4Encoding_RGB8A1::PerformIteration(float a_fEffort) + { + assert(!m_boolOpaque); + assert(!m_boolTransparent); + assert(!m_boolDone); + + switch (m_uiEncodingIterations) + { + case 0: + PerformFirstIteration(); + break; + + case 1: + TryDifferential(m_boolMostLikelyFlip, 1, 0, 0); + break; + + case 2: + TryDifferential(!m_boolMostLikelyFlip, 1, 0, 0); + if (a_fEffort <= 39.5f) + { + m_boolDone = true; + } + break; + + case 3: + Block4x4Encoding_RGB8::CalculateBaseColorsForTAndH(); + TryT(1); + TryH(1); + if (a_fEffort <= 49.5f) + { + m_boolDone = true; + } + break; + + case 4: + TryDegenerates1(); + if (a_fEffort <= 59.5f) + { + m_boolDone = true; + } + break; + + case 5: + TryDegenerates2(); + if (a_fEffort <= 69.5f) + { + m_boolDone = true; + } + break; + + case 6: + TryDegenerates3(); + if (a_fEffort <= 79.5f) + { + m_boolDone = true; + } + break; + + case 7: + TryDegenerates4(); + m_boolDone = true; + break; + + default: + assert(0); + break; + } + + m_uiEncodingIterations++; + + SetDoneIfPerfect(); + + } + + // ---------------------------------------------------------------------------------------------------- + // find best initial encoding to ensure block has a valid encoding + // + void Block4x4Encoding_RGB8A1::PerformFirstIteration(void) + { + Block4x4Encoding_ETC1::CalculateMostLikelyFlip(); + + m_fError = FLT_MAX; + + TryDifferential(m_boolMostLikelyFlip, 0, 0, 0); + SetDoneIfPerfect(); + if (m_boolDone) + { + return; + } + TryDifferential(!m_boolMostLikelyFlip, 0, 0, 0); + SetDoneIfPerfect(); + + } + + // ---------------------------------------------------------------------------------------------------- + // mostly copied from ETC1 + // differences: + // Block4x4Encoding_RGB8A1 encodingTry = *this; + // + void Block4x4Encoding_RGB8A1::TryDifferential(bool a_boolFlip, unsigned int a_uiRadius, + int a_iGrayOffset1, int a_iGrayOffset2) + { + + ColorFloatRGBA frgbaColor1; + ColorFloatRGBA frgbaColor2; + + const unsigned int *pauiPixelMapping1; + const unsigned int *pauiPixelMapping2; + + if (a_boolFlip) + { + frgbaColor1 = m_frgbaSourceAverageTop; + frgbaColor2 = m_frgbaSourceAverageBottom; + + pauiPixelMapping1 = s_auiTopPixelMapping; + pauiPixelMapping2 = s_auiBottomPixelMapping; + } + else + { + frgbaColor1 = m_frgbaSourceAverageLeft; + frgbaColor2 = m_frgbaSourceAverageRight; + + pauiPixelMapping1 = s_auiLeftPixelMapping; + pauiPixelMapping2 = s_auiRightPixelMapping; + } + + DifferentialTrys trys(frgbaColor1, frgbaColor2, pauiPixelMapping1, pauiPixelMapping2, + a_uiRadius, a_iGrayOffset1, a_iGrayOffset2); + + Block4x4Encoding_RGB8A1 encodingTry = *this; + encodingTry.m_boolFlip = a_boolFlip; + + encodingTry.TryDifferentialHalf(&trys.m_half1); + encodingTry.TryDifferentialHalf(&trys.m_half2); + + // find best halves that are within differential range + DifferentialTrys::Try *ptryBest1 = nullptr; + DifferentialTrys::Try *ptryBest2 = nullptr; + encodingTry.m_fError = FLT_MAX; + + // see if the best of each half are in differential range + int iDRed = trys.m_half2.m_ptryBest->m_iRed - trys.m_half1.m_ptryBest->m_iRed; + int iDGreen = trys.m_half2.m_ptryBest->m_iGreen - trys.m_half1.m_ptryBest->m_iGreen; + int iDBlue = trys.m_half2.m_ptryBest->m_iBlue - trys.m_half1.m_ptryBest->m_iBlue; + if (iDRed >= -4 && iDRed <= 3 && iDGreen >= -4 && iDGreen <= 3 && iDBlue >= -4 && iDBlue <= 3) + { + ptryBest1 = trys.m_half1.m_ptryBest; + ptryBest2 = trys.m_half2.m_ptryBest; + encodingTry.m_fError = trys.m_half1.m_ptryBest->m_fError + trys.m_half2.m_ptryBest->m_fError; + } + else + { + // else, find the next best halves that are in differential range + for (DifferentialTrys::Try *ptry1 = &trys.m_half1.m_atry[0]; + ptry1 < &trys.m_half1.m_atry[trys.m_half1.m_uiTrys]; + ptry1++) + { + for (DifferentialTrys::Try *ptry2 = &trys.m_half2.m_atry[0]; + ptry2 < &trys.m_half2.m_atry[trys.m_half2.m_uiTrys]; + ptry2++) + { + iDRed = ptry2->m_iRed - ptry1->m_iRed; + bool boolValidRedDelta = iDRed <= 3 && iDRed >= -4; + iDGreen = ptry2->m_iGreen - ptry1->m_iGreen; + bool boolValidGreenDelta = iDGreen <= 3 && iDGreen >= -4; + iDBlue = ptry2->m_iBlue - ptry1->m_iBlue; + bool boolValidBlueDelta = iDBlue <= 3 && iDBlue >= -4; + + if (boolValidRedDelta && boolValidGreenDelta && boolValidBlueDelta) + { + float fError = ptry1->m_fError + ptry2->m_fError; + + if (fError < encodingTry.m_fError) + { + encodingTry.m_fError = fError; + + ptryBest1 = ptry1; + ptryBest2 = ptry2; + } + } + + } + } + assert(encodingTry.m_fError < FLT_MAX); + assert(ptryBest1 != nullptr); + assert(ptryBest2 != nullptr); + } + + if (encodingTry.m_fError < m_fError) + { + m_mode = MODE_ETC1; + m_boolDiff = true; + m_boolFlip = encodingTry.m_boolFlip; + m_frgbaColor1 = ColorFloatRGBA::ConvertFromRGB5((unsigned char)ptryBest1->m_iRed, (unsigned char)ptryBest1->m_iGreen, (unsigned char)ptryBest1->m_iBlue); + m_frgbaColor2 = ColorFloatRGBA::ConvertFromRGB5((unsigned char)ptryBest2->m_iRed, (unsigned char)ptryBest2->m_iGreen, (unsigned char)ptryBest2->m_iBlue); + m_uiCW1 = ptryBest1->m_uiCW; + m_uiCW2 = ptryBest2->m_uiCW; + + m_fError = 0.0f; + for (unsigned int uiPixelOrder = 0; uiPixelOrder < PIXELS / 2; uiPixelOrder++) + { + unsigned int uiPixel1 = pauiPixelMapping1[uiPixelOrder]; + unsigned int uiPixel2 = pauiPixelMapping2[uiPixelOrder]; + + unsigned int uiSelector1 = ptryBest1->m_auiSelectors[uiPixelOrder]; + unsigned int uiSelector2 = ptryBest2->m_auiSelectors[uiPixelOrder]; + + m_auiSelectors[uiPixel1] = uiSelector1; + m_auiSelectors[uiPixel2] = ptryBest2->m_auiSelectors[uiPixelOrder]; + + if (uiSelector1 == TRANSPARENT_SELECTOR) + { + m_afrgbaDecodedColors[uiPixel1] = ColorFloatRGBA(); + m_afDecodedAlphas[uiPixel1] = 0.0f; + } + else + { + float fDeltaRGB1 = s_aafCwOpaqueUnsetTable[m_uiCW1][uiSelector1]; + m_afrgbaDecodedColors[uiPixel1] = (m_frgbaColor1 + fDeltaRGB1).ClampRGB(); + m_afDecodedAlphas[uiPixel1] = 1.0f; + } + + if (uiSelector2 == TRANSPARENT_SELECTOR) + { + m_afrgbaDecodedColors[uiPixel2] = ColorFloatRGBA(); + m_afDecodedAlphas[uiPixel2] = 0.0f; + } + else + { + float fDeltaRGB2 = s_aafCwOpaqueUnsetTable[m_uiCW2][uiSelector2]; + m_afrgbaDecodedColors[uiPixel2] = (m_frgbaColor2 + fDeltaRGB2).ClampRGB(); + m_afDecodedAlphas[uiPixel2] = 1.0f; + } + + float fDeltaA1 = m_afDecodedAlphas[uiPixel1] - m_pafrgbaSource[uiPixel1].fA; + m_fError += fDeltaA1 * fDeltaA1; + float fDeltaA2 = m_afDecodedAlphas[uiPixel2] - m_pafrgbaSource[uiPixel2].fA; + m_fError += fDeltaA2 * fDeltaA2; + } + + m_fError1 = ptryBest1->m_fError; + m_fError2 = ptryBest2->m_fError; + m_boolSeverelyBentDifferentialColors = trys.m_boolSeverelyBentColors; + m_fError = m_fError1 + m_fError2; + + // sanity check + { + int iRed1 = m_frgbaColor1.IntRed(31.0f); + int iGreen1 = m_frgbaColor1.IntGreen(31.0f); + int iBlue1 = m_frgbaColor1.IntBlue(31.0f); + + int iRed2 = m_frgbaColor2.IntRed(31.0f); + int iGreen2 = m_frgbaColor2.IntGreen(31.0f); + int iBlue2 = m_frgbaColor2.IntBlue(31.0f); + + iDRed = iRed2 - iRed1; + iDGreen = iGreen2 - iGreen1; + iDBlue = iBlue2 - iBlue1; + + assert(iDRed >= -4 && iDRed < 4); + assert(iDGreen >= -4 && iDGreen < 4); + assert(iDBlue >= -4 && iDBlue < 4); + } + } + + } + + // ---------------------------------------------------------------------------------------------------- + // mostly copied from ETC1 + // differences: + // uses s_aafCwOpaqueUnsetTable + // color for selector set to 0,0,0,0 + // + void Block4x4Encoding_RGB8A1::TryDifferentialHalf(DifferentialTrys::Half *a_phalf) + { + + a_phalf->m_ptryBest = nullptr; + float fBestTryError = FLT_MAX; + + a_phalf->m_uiTrys = 0; + for (int iRed = a_phalf->m_iRed - (int)a_phalf->m_uiRadius; + iRed <= a_phalf->m_iRed + (int)a_phalf->m_uiRadius; + iRed++) + { + assert(iRed >= 0 && iRed <= 31); + + for (int iGreen = a_phalf->m_iGreen - (int)a_phalf->m_uiRadius; + iGreen <= a_phalf->m_iGreen + (int)a_phalf->m_uiRadius; + iGreen++) + { + assert(iGreen >= 0 && iGreen <= 31); + + for (int iBlue = a_phalf->m_iBlue - (int)a_phalf->m_uiRadius; + iBlue <= a_phalf->m_iBlue + (int)a_phalf->m_uiRadius; + iBlue++) + { + assert(iBlue >= 0 && iBlue <= 31); + + DifferentialTrys::Try *ptry = &a_phalf->m_atry[a_phalf->m_uiTrys]; + assert(ptry < &a_phalf->m_atry[DifferentialTrys::Half::MAX_TRYS]); + + ptry->m_iRed = iRed; + ptry->m_iGreen = iGreen; + ptry->m_iBlue = iBlue; + ptry->m_fError = FLT_MAX; + ColorFloatRGBA frgbaColor = ColorFloatRGBA::ConvertFromRGB5((unsigned char)iRed, (unsigned char)iGreen, (unsigned char)iBlue); + + // try each CW + for (unsigned int uiCW = 0; uiCW < CW_RANGES; uiCW++) + { + unsigned int auiPixelSelectors[PIXELS / 2]; + ColorFloatRGBA afrgbaDecodedColors[PIXELS / 2]; + float afPixelErrors[PIXELS / 2] = { FLT_MAX, FLT_MAX, FLT_MAX, FLT_MAX, + FLT_MAX, FLT_MAX, FLT_MAX, FLT_MAX }; + + // pre-compute decoded pixels for each selector + ColorFloatRGBA afrgbaSelectors[SELECTORS]; + assert(SELECTORS == 4); + afrgbaSelectors[0] = (frgbaColor + s_aafCwOpaqueUnsetTable[uiCW][0]).ClampRGB(); + afrgbaSelectors[1] = (frgbaColor + s_aafCwOpaqueUnsetTable[uiCW][1]).ClampRGB(); + afrgbaSelectors[2] = ColorFloatRGBA(); + afrgbaSelectors[3] = (frgbaColor + s_aafCwOpaqueUnsetTable[uiCW][3]).ClampRGB(); + + for (unsigned int uiPixel = 0; uiPixel < 8; uiPixel++) + { + ColorFloatRGBA *pfrgbaSourcePixel = &m_pafrgbaSource[a_phalf->m_pauiPixelMapping[uiPixel]]; + ColorFloatRGBA frgbaDecodedPixel; + + for (unsigned int uiSelector = 0; uiSelector < SELECTORS; uiSelector++) + { + if (pfrgbaSourcePixel->fA < 0.5f) + { + uiSelector = TRANSPARENT_SELECTOR; + } + else if (uiSelector == TRANSPARENT_SELECTOR) + { + continue; + } + + frgbaDecodedPixel = afrgbaSelectors[uiSelector]; + + float fPixelError; + + fPixelError = CalcPixelError(frgbaDecodedPixel, m_afDecodedAlphas[a_phalf->m_pauiPixelMapping[uiPixel]], + *pfrgbaSourcePixel); + + if (fPixelError < afPixelErrors[uiPixel]) + { + auiPixelSelectors[uiPixel] = uiSelector; + afrgbaDecodedColors[uiPixel] = frgbaDecodedPixel; + afPixelErrors[uiPixel] = fPixelError; + } + + if (uiSelector == TRANSPARENT_SELECTOR) + { + break; + } + } + } + + // add up all pixel errors + float fCWError = 0.0f; + for (unsigned int uiPixel = 0; uiPixel < 8; uiPixel++) + { + fCWError += afPixelErrors[uiPixel]; + } + + // if best CW so far + if (fCWError < ptry->m_fError) + { + ptry->m_uiCW = uiCW; + for (unsigned int uiPixel = 0; uiPixel < 8; uiPixel++) + { + ptry->m_auiSelectors[uiPixel] = auiPixelSelectors[uiPixel]; + } + ptry->m_fError = fCWError; + } + + } + + if (ptry->m_fError < fBestTryError) + { + a_phalf->m_ptryBest = ptry; + fBestTryError = ptry->m_fError; + } + + assert(ptry->m_fError < FLT_MAX); + + a_phalf->m_uiTrys++; + } + } + } + + } + + // ---------------------------------------------------------------------------------------------------- + // try encoding in T mode + // save this encoding if it improves the error + // + // since pixels that use base color1 don't use the distance table, color1 and color2 can be twiddled independently + // better encoding can be found if TWIDDLE_RADIUS is set to 2, but it will be much slower + // + void Block4x4Encoding_RGB8A1::TryT(unsigned int a_uiRadius) + { + Block4x4Encoding_RGB8A1 encodingTry = *this; + + // init "try" + { + encodingTry.m_mode = MODE_T; + encodingTry.m_boolDiff = true; + encodingTry.m_boolFlip = false; + encodingTry.m_fError = FLT_MAX; + } + + int iColor1Red = m_frgbaOriginalColor1_TAndH.IntRed(15.0f); + int iColor1Green = m_frgbaOriginalColor1_TAndH.IntGreen(15.0f); + int iColor1Blue = m_frgbaOriginalColor1_TAndH.IntBlue(15.0f); + + int iMinRed1 = iColor1Red - (int)a_uiRadius; + if (iMinRed1 < 0) + { + iMinRed1 = 0; + } + int iMaxRed1 = iColor1Red + (int)a_uiRadius; + if (iMaxRed1 > 15) + { + iMinRed1 = 15; + } + + int iMinGreen1 = iColor1Green - (int)a_uiRadius; + if (iMinGreen1 < 0) + { + iMinGreen1 = 0; + } + int iMaxGreen1 = iColor1Green + (int)a_uiRadius; + if (iMaxGreen1 > 15) + { + iMinGreen1 = 15; + } + + int iMinBlue1 = iColor1Blue - (int)a_uiRadius; + if (iMinBlue1 < 0) + { + iMinBlue1 = 0; + } + int iMaxBlue1 = iColor1Blue + (int)a_uiRadius; + if (iMaxBlue1 > 15) + { + iMinBlue1 = 15; + } + + int iColor2Red = m_frgbaOriginalColor2_TAndH.IntRed(15.0f); + int iColor2Green = m_frgbaOriginalColor2_TAndH.IntGreen(15.0f); + int iColor2Blue = m_frgbaOriginalColor2_TAndH.IntBlue(15.0f); + + int iMinRed2 = iColor2Red - (int)a_uiRadius; + if (iMinRed2 < 0) + { + iMinRed2 = 0; + } + int iMaxRed2 = iColor2Red + (int)a_uiRadius; + if (iMaxRed2 > 15) + { + iMinRed2 = 15; + } + + int iMinGreen2 = iColor2Green - (int)a_uiRadius; + if (iMinGreen2 < 0) + { + iMinGreen2 = 0; + } + int iMaxGreen2 = iColor2Green + (int)a_uiRadius; + if (iMaxGreen2 > 15) + { + iMinGreen2 = 15; + } + + int iMinBlue2 = iColor2Blue - (int)a_uiRadius; + if (iMinBlue2 < 0) + { + iMinBlue2 = 0; + } + int iMaxBlue2 = iColor2Blue + (int)a_uiRadius; + if (iMaxBlue2 > 15) + { + iMinBlue2 = 15; + } + + for (unsigned int uiDistance = 0; uiDistance < TH_DISTANCES; uiDistance++) + { + encodingTry.m_uiCW1 = uiDistance; + + // twiddle m_frgbaOriginalColor2_TAndH + // twiddle color2 first, since it affects 3 selectors, while color1 only affects one selector + // + for (int iRed2 = iMinRed2; iRed2 <= iMaxRed2; iRed2++) + { + for (int iGreen2 = iMinGreen2; iGreen2 <= iMaxGreen2; iGreen2++) + { + for (int iBlue2 = iMinBlue2; iBlue2 <= iMaxBlue2; iBlue2++) + { + for (unsigned int uiBaseColorSwaps = 0; uiBaseColorSwaps < 2; uiBaseColorSwaps++) + { + if (uiBaseColorSwaps == 0) + { + encodingTry.m_frgbaColor1 = m_frgbaOriginalColor1_TAndH; + encodingTry.m_frgbaColor2 = ColorFloatRGBA::ConvertFromRGB4((unsigned char)iRed2, (unsigned char)iGreen2, (unsigned char)iBlue2); + } + else + { + encodingTry.m_frgbaColor1 = ColorFloatRGBA::ConvertFromRGB4((unsigned char)iRed2, (unsigned char)iGreen2, (unsigned char)iBlue2); + encodingTry.m_frgbaColor2 = m_frgbaOriginalColor1_TAndH; + } + + encodingTry.TryT_BestSelectorCombination(); + + if (encodingTry.m_fError < m_fError) + { + m_mode = encodingTry.m_mode; + m_boolDiff = encodingTry.m_boolDiff; + m_boolFlip = encodingTry.m_boolFlip; + + m_frgbaColor1 = encodingTry.m_frgbaColor1; + m_frgbaColor2 = encodingTry.m_frgbaColor2; + m_uiCW1 = encodingTry.m_uiCW1; + + for (unsigned int uiPixel = 0; uiPixel < PIXELS; uiPixel++) + { + m_auiSelectors[uiPixel] = encodingTry.m_auiSelectors[uiPixel]; + m_afrgbaDecodedColors[uiPixel] = encodingTry.m_afrgbaDecodedColors[uiPixel]; + } + + m_fError = encodingTry.m_fError; + } + } + } + } + } + + // twiddle m_frgbaOriginalColor1_TAndH + for (int iRed1 = iMinRed1; iRed1 <= iMaxRed1; iRed1++) + { + for (int iGreen1 = iMinGreen1; iGreen1 <= iMaxGreen1; iGreen1++) + { + for (int iBlue1 = iMinBlue1; iBlue1 <= iMaxBlue1; iBlue1++) + { + for (unsigned int uiBaseColorSwaps = 0; uiBaseColorSwaps < 2; uiBaseColorSwaps++) + { + if (uiBaseColorSwaps == 0) + { + encodingTry.m_frgbaColor1 = ColorFloatRGBA::ConvertFromRGB4((unsigned char)iRed1, (unsigned char)iGreen1, (unsigned char)iBlue1); + encodingTry.m_frgbaColor2 = m_frgbaOriginalColor2_TAndH; + } + else + { + encodingTry.m_frgbaColor1 = m_frgbaOriginalColor2_TAndH; + encodingTry.m_frgbaColor2 = ColorFloatRGBA::ConvertFromRGB4((unsigned char)iRed1, (unsigned char)iGreen1, (unsigned char)iBlue1); + } + + encodingTry.TryT_BestSelectorCombination(); + + if (encodingTry.m_fError < m_fError) + { + m_mode = encodingTry.m_mode; + m_boolDiff = encodingTry.m_boolDiff; + m_boolFlip = encodingTry.m_boolFlip; + + m_frgbaColor1 = encodingTry.m_frgbaColor1; + m_frgbaColor2 = encodingTry.m_frgbaColor2; + m_uiCW1 = encodingTry.m_uiCW1; + + for (unsigned int uiPixel = 0; uiPixel < PIXELS; uiPixel++) + { + m_auiSelectors[uiPixel] = encodingTry.m_auiSelectors[uiPixel]; + m_afrgbaDecodedColors[uiPixel] = encodingTry.m_afrgbaDecodedColors[uiPixel]; + } + + m_fError = encodingTry.m_fError; + } + } + } + } + } + + } + + } + + // ---------------------------------------------------------------------------------------------------- + // find best selector combination for TryT + // called on an encodingTry + // + void Block4x4Encoding_RGB8A1::TryT_BestSelectorCombination(void) + { + + float fDistance = s_afTHDistanceTable[m_uiCW1]; + + unsigned int auiBestPixelSelectors[PIXELS]; + float afBestPixelErrors[PIXELS] = { FLT_MAX, FLT_MAX, FLT_MAX, FLT_MAX, FLT_MAX, FLT_MAX, FLT_MAX, FLT_MAX, + FLT_MAX, FLT_MAX, FLT_MAX, FLT_MAX, FLT_MAX, FLT_MAX, FLT_MAX, FLT_MAX }; + ColorFloatRGBA afrgbaBestDecodedPixels[PIXELS]; + ColorFloatRGBA afrgbaDecodedPixel[SELECTORS]; + + assert(SELECTORS == 4); + afrgbaDecodedPixel[0] = m_frgbaColor1; + afrgbaDecodedPixel[1] = (m_frgbaColor2 + fDistance).ClampRGB(); + afrgbaDecodedPixel[2] = ColorFloatRGBA(); + afrgbaDecodedPixel[3] = (m_frgbaColor2 - fDistance).ClampRGB(); + + // try each selector + for (unsigned int uiPixel = 0; uiPixel < PIXELS; uiPixel++) + { + unsigned int uiMinSelector = 0; + unsigned int uiMaxSelector = SELECTORS - 1; + + if (m_pafrgbaSource[uiPixel].fA < 0.5f) + { + uiMinSelector = 2; + uiMaxSelector = 2; + } + + for (unsigned int uiSelector = uiMinSelector; uiSelector <= uiMaxSelector; uiSelector++) + { + float fPixelError = CalcPixelError(afrgbaDecodedPixel[uiSelector], m_afDecodedAlphas[uiPixel], + m_pafrgbaSource[uiPixel]); + + if (fPixelError < afBestPixelErrors[uiPixel]) + { + afBestPixelErrors[uiPixel] = fPixelError; + auiBestPixelSelectors[uiPixel] = uiSelector; + afrgbaBestDecodedPixels[uiPixel] = afrgbaDecodedPixel[uiSelector]; + } + } + } + + + // add up all of the pixel errors + float fBlockError = 0.0f; + for (unsigned int uiPixel = 0; uiPixel < PIXELS; uiPixel++) + { + fBlockError += afBestPixelErrors[uiPixel]; + } + + if (fBlockError < m_fError) + { + m_fError = fBlockError; + + for (unsigned int uiPixel = 0; uiPixel < PIXELS; uiPixel++) + { + m_auiSelectors[uiPixel] = auiBestPixelSelectors[uiPixel]; + m_afrgbaDecodedColors[uiPixel] = afrgbaBestDecodedPixels[uiPixel]; + } + } + + } + + // ---------------------------------------------------------------------------------------------------- + // try encoding in H mode + // save this encoding if it improves the error + // + // since all pixels use the distance table, color1 and color2 can NOT be twiddled independently + // TWIDDLE_RADIUS of 2 is WAY too slow + // + void Block4x4Encoding_RGB8A1::TryH(unsigned int a_uiRadius) + { + Block4x4Encoding_RGB8A1 encodingTry = *this; + + // init "try" + { + encodingTry.m_mode = MODE_H; + encodingTry.m_boolDiff = true; + encodingTry.m_boolFlip = false; + encodingTry.m_fError = FLT_MAX; + } + + int iColor1Red = m_frgbaOriginalColor1_TAndH.IntRed(15.0f); + int iColor1Green = m_frgbaOriginalColor1_TAndH.IntGreen(15.0f); + int iColor1Blue = m_frgbaOriginalColor1_TAndH.IntBlue(15.0f); + + int iMinRed1 = iColor1Red - (int)a_uiRadius; + if (iMinRed1 < 0) + { + iMinRed1 = 0; + } + int iMaxRed1 = iColor1Red + (int)a_uiRadius; + if (iMaxRed1 > 15) + { + iMinRed1 = 15; + } + + int iMinGreen1 = iColor1Green - (int)a_uiRadius; + if (iMinGreen1 < 0) + { + iMinGreen1 = 0; + } + int iMaxGreen1 = iColor1Green + (int)a_uiRadius; + if (iMaxGreen1 > 15) + { + iMinGreen1 = 15; + } + + int iMinBlue1 = iColor1Blue - (int)a_uiRadius; + if (iMinBlue1 < 0) + { + iMinBlue1 = 0; + } + int iMaxBlue1 = iColor1Blue + (int)a_uiRadius; + if (iMaxBlue1 > 15) + { + iMinBlue1 = 15; + } + + int iColor2Red = m_frgbaOriginalColor2_TAndH.IntRed(15.0f); + int iColor2Green = m_frgbaOriginalColor2_TAndH.IntGreen(15.0f); + int iColor2Blue = m_frgbaOriginalColor2_TAndH.IntBlue(15.0f); + + int iMinRed2 = iColor2Red - (int)a_uiRadius; + if (iMinRed2 < 0) + { + iMinRed2 = 0; + } + int iMaxRed2 = iColor2Red + (int)a_uiRadius; + if (iMaxRed2 > 15) + { + iMinRed2 = 15; + } + + int iMinGreen2 = iColor2Green - (int)a_uiRadius; + if (iMinGreen2 < 0) + { + iMinGreen2 = 0; + } + int iMaxGreen2 = iColor2Green + (int)a_uiRadius; + if (iMaxGreen2 > 15) + { + iMinGreen2 = 15; + } + + int iMinBlue2 = iColor2Blue - (int)a_uiRadius; + if (iMinBlue2 < 0) + { + iMinBlue2 = 0; + } + int iMaxBlue2 = iColor2Blue + (int)a_uiRadius; + if (iMaxBlue2 > 15) + { + iMinBlue2 = 15; + } + + for (unsigned int uiDistance = 0; uiDistance < TH_DISTANCES; uiDistance++) + { + encodingTry.m_uiCW1 = uiDistance; + + // twiddle m_frgbaOriginalColor1_TAndH + for (int iRed1 = iMinRed1; iRed1 <= iMaxRed1; iRed1++) + { + for (int iGreen1 = iMinGreen1; iGreen1 <= iMaxGreen1; iGreen1++) + { + for (int iBlue1 = iMinBlue1; iBlue1 <= iMaxBlue1; iBlue1++) + { + encodingTry.m_frgbaColor1 = ColorFloatRGBA::ConvertFromRGB4((unsigned char)iRed1, (unsigned char)iGreen1, (unsigned char)iBlue1); + encodingTry.m_frgbaColor2 = m_frgbaOriginalColor2_TAndH; + + // if color1 == color2, H encoding issues can pop up, so abort + if (iRed1 == iColor2Red && iGreen1 == iColor2Green && iBlue1 == iColor2Blue) + { + continue; + } + + encodingTry.TryH_BestSelectorCombination(); + + if (encodingTry.m_fError < m_fError) + { + m_mode = encodingTry.m_mode; + m_boolDiff = encodingTry.m_boolDiff; + m_boolFlip = encodingTry.m_boolFlip; + + m_frgbaColor1 = encodingTry.m_frgbaColor1; + m_frgbaColor2 = encodingTry.m_frgbaColor2; + m_uiCW1 = encodingTry.m_uiCW1; + + for (unsigned int uiPixel = 0; uiPixel < PIXELS; uiPixel++) + { + m_auiSelectors[uiPixel] = encodingTry.m_auiSelectors[uiPixel]; + m_afrgbaDecodedColors[uiPixel] = encodingTry.m_afrgbaDecodedColors[uiPixel]; + } + + m_fError = encodingTry.m_fError; + } + } + } + } + + // twiddle m_frgbaOriginalColor2_TAndH + for (int iRed2 = iMinRed2; iRed2 <= iMaxRed2; iRed2++) + { + for (int iGreen2 = iMinGreen2; iGreen2 <= iMaxGreen2; iGreen2++) + { + for (int iBlue2 = iMinBlue2; iBlue2 <= iMaxBlue2; iBlue2++) + { + encodingTry.m_frgbaColor1 = m_frgbaOriginalColor1_TAndH; + encodingTry.m_frgbaColor2 = ColorFloatRGBA::ConvertFromRGB4((unsigned char)iRed2, (unsigned char)iGreen2, (unsigned char)iBlue2); + + // if color1 == color2, H encoding issues can pop up, so abort + if (iRed2 == iColor1Red && iGreen2 == iColor1Green && iBlue2 == iColor1Blue) + { + continue; + } + + encodingTry.TryH_BestSelectorCombination(); + + if (encodingTry.m_fError < m_fError) + { + m_mode = encodingTry.m_mode; + m_boolDiff = encodingTry.m_boolDiff; + m_boolFlip = encodingTry.m_boolFlip; + + m_frgbaColor1 = encodingTry.m_frgbaColor1; + m_frgbaColor2 = encodingTry.m_frgbaColor2; + m_uiCW1 = encodingTry.m_uiCW1; + + for (unsigned int uiPixel = 0; uiPixel < PIXELS; uiPixel++) + { + m_auiSelectors[uiPixel] = encodingTry.m_auiSelectors[uiPixel]; + m_afrgbaDecodedColors[uiPixel] = encodingTry.m_afrgbaDecodedColors[uiPixel]; + } + + m_fError = encodingTry.m_fError; + } + } + } + } + + } + + } + + // ---------------------------------------------------------------------------------------------------- + // find best selector combination for TryH + // called on an encodingTry + // + void Block4x4Encoding_RGB8A1::TryH_BestSelectorCombination(void) + { + + // abort if colors and CW will pose an encoding problem + { + unsigned int uiRed1 = (unsigned int)m_frgbaColor1.IntRed(255.0f); + unsigned int uiGreen1 = (unsigned int)m_frgbaColor1.IntGreen(255.0f); + unsigned int uiBlue1 = (unsigned int)m_frgbaColor1.IntBlue(255.0f); + unsigned int uiColorValue1 = (uiRed1 << 16) + (uiGreen1 << 8) + uiBlue1; + + unsigned int uiRed2 = (unsigned int)m_frgbaColor2.IntRed(255.0f); + unsigned int uiGreen2 = (unsigned int)m_frgbaColor2.IntGreen(255.0f); + unsigned int uiBlue2 = (unsigned int)m_frgbaColor2.IntBlue(255.0f); + unsigned int uiColorValue2 = (uiRed2 << 16) + (uiGreen2 << 8) + uiBlue2; + + unsigned int uiCWLsb = m_uiCW1 & 1; + + if ((uiColorValue1 >= (uiColorValue2 & uiCWLsb)) == 0 || + (uiColorValue1 < (uiColorValue2 & uiCWLsb)) == 1) + { + return; + } + } + + float fDistance = s_afTHDistanceTable[m_uiCW1]; + + unsigned int auiBestPixelSelectors[PIXELS]; + float afBestPixelErrors[PIXELS] = { FLT_MAX, FLT_MAX, FLT_MAX, FLT_MAX, FLT_MAX, FLT_MAX, FLT_MAX, FLT_MAX, + FLT_MAX, FLT_MAX, FLT_MAX, FLT_MAX, FLT_MAX, FLT_MAX, FLT_MAX, FLT_MAX }; + ColorFloatRGBA afrgbaBestDecodedPixels[PIXELS]; + ColorFloatRGBA afrgbaDecodedPixel[SELECTORS]; + + assert(SELECTORS == 4); + afrgbaDecodedPixel[0] = (m_frgbaColor1 + fDistance).ClampRGB(); + afrgbaDecodedPixel[1] = (m_frgbaColor1 - fDistance).ClampRGB(); + afrgbaDecodedPixel[2] = ColorFloatRGBA();; + afrgbaDecodedPixel[3] = (m_frgbaColor2 - fDistance).ClampRGB(); + + + // try each selector + for (unsigned int uiPixel = 0; uiPixel < PIXELS; uiPixel++) + { + unsigned int uiMinSelector = 0; + unsigned int uiMaxSelector = SELECTORS - 1; + + if (m_pafrgbaSource[uiPixel].fA < 0.5f) + { + uiMinSelector = 2; + uiMaxSelector = 2; + } + + for (unsigned int uiSelector = uiMinSelector; uiSelector <= uiMaxSelector; uiSelector++) + { + float fPixelError = CalcPixelError(afrgbaDecodedPixel[uiSelector], m_afDecodedAlphas[uiPixel], + m_pafrgbaSource[uiPixel]); + + if (fPixelError < afBestPixelErrors[uiPixel]) + { + afBestPixelErrors[uiPixel] = fPixelError; + auiBestPixelSelectors[uiPixel] = uiSelector; + afrgbaBestDecodedPixels[uiPixel] = afrgbaDecodedPixel[uiSelector]; + } + } + } + + + // add up all of the pixel errors + float fBlockError = 0.0f; + for (unsigned int uiPixel = 0; uiPixel < PIXELS; uiPixel++) + { + fBlockError += afBestPixelErrors[uiPixel]; + } + + if (fBlockError < m_fError) + { + m_fError = fBlockError; + + for (unsigned int uiPixel = 0; uiPixel < PIXELS; uiPixel++) + { + m_auiSelectors[uiPixel] = auiBestPixelSelectors[uiPixel]; + m_afrgbaDecodedColors[uiPixel] = afrgbaBestDecodedPixels[uiPixel]; + } + } + + } + + // ---------------------------------------------------------------------------------------------------- + // try version 1 of the degenerate search + // degenerate encodings use basecolor movement and a subset of the selectors to find useful encodings + // each subsequent version of the degenerate search uses more basecolor movement and is less likely to + // be successfull + // + void Block4x4Encoding_RGB8A1::TryDegenerates1(void) + { + + TryDifferential(m_boolMostLikelyFlip, 1, -2, 0); + TryDifferential(m_boolMostLikelyFlip, 1, 2, 0); + TryDifferential(m_boolMostLikelyFlip, 1, 0, 2); + TryDifferential(m_boolMostLikelyFlip, 1, 0, -2); + + } + + // ---------------------------------------------------------------------------------------------------- + // try version 2 of the degenerate search + // degenerate encodings use basecolor movement and a subset of the selectors to find useful encodings + // each subsequent version of the degenerate search uses more basecolor movement and is less likely to + // be successfull + // + void Block4x4Encoding_RGB8A1::TryDegenerates2(void) + { + + TryDifferential(!m_boolMostLikelyFlip, 1, -2, 0); + TryDifferential(!m_boolMostLikelyFlip, 1, 2, 0); + TryDifferential(!m_boolMostLikelyFlip, 1, 0, 2); + TryDifferential(!m_boolMostLikelyFlip, 1, 0, -2); + + } + + // ---------------------------------------------------------------------------------------------------- + // try version 3 of the degenerate search + // degenerate encodings use basecolor movement and a subset of the selectors to find useful encodings + // each subsequent version of the degenerate search uses more basecolor movement and is less likely to + // be successfull + // + void Block4x4Encoding_RGB8A1::TryDegenerates3(void) + { + + TryDifferential(m_boolMostLikelyFlip, 1, -2, -2); + TryDifferential(m_boolMostLikelyFlip, 1, -2, 2); + TryDifferential(m_boolMostLikelyFlip, 1, 2, -2); + TryDifferential(m_boolMostLikelyFlip, 1, 2, 2); + + } + + // ---------------------------------------------------------------------------------------------------- + // try version 4 of the degenerate search + // degenerate encodings use basecolor movement and a subset of the selectors to find useful encodings + // each subsequent version of the degenerate search uses more basecolor movement and is less likely to + // be successfull + // + void Block4x4Encoding_RGB8A1::TryDegenerates4(void) + { + + TryDifferential(m_boolMostLikelyFlip, 1, -4, 0); + TryDifferential(m_boolMostLikelyFlip, 1, 4, 0); + TryDifferential(m_boolMostLikelyFlip, 1, 0, 4); + TryDifferential(m_boolMostLikelyFlip, 1, 0, -4); + + } + + // ---------------------------------------------------------------------------------------------------- + // set the encoding bits based on encoding state + // + void Block4x4Encoding_RGB8A1::SetEncodingBits(void) + { + switch (m_mode) + { + case MODE_ETC1: + SetEncodingBits_ETC1(); + break; + + case MODE_T: + SetEncodingBits_T(); + break; + + case MODE_H: + SetEncodingBits_H(); + break; + + case MODE_PLANAR: + Block4x4Encoding_RGB8::SetEncodingBits_Planar(); + break; + + default: + assert(false); + } + } + + // ---------------------------------------------------------------------------------------------------- + // set the encoding bits based on encoding state if ETC1 mode + // + void Block4x4Encoding_RGB8A1::SetEncodingBits_ETC1(void) + { + + // there is no individual mode in RGB8A1 + assert(m_boolDiff); + + int iRed1 = m_frgbaColor1.IntRed(31.0f); + int iGreen1 = m_frgbaColor1.IntGreen(31.0f); + int iBlue1 = m_frgbaColor1.IntBlue(31.0f); + + int iRed2 = m_frgbaColor2.IntRed(31.0f); + int iGreen2 = m_frgbaColor2.IntGreen(31.0f); + int iBlue2 = m_frgbaColor2.IntBlue(31.0f); + + int iDRed2 = iRed2 - iRed1; + int iDGreen2 = iGreen2 - iGreen1; + int iDBlue2 = iBlue2 - iBlue1; + + assert(iDRed2 >= -4 && iDRed2 < 4); + assert(iDGreen2 >= -4 && iDGreen2 < 4); + assert(iDBlue2 >= -4 && iDBlue2 < 4); + + m_pencodingbitsRGB8->differential.red1 = iRed1; + m_pencodingbitsRGB8->differential.green1 = iGreen1; + m_pencodingbitsRGB8->differential.blue1 = iBlue1; + + m_pencodingbitsRGB8->differential.dred2 = iDRed2; + m_pencodingbitsRGB8->differential.dgreen2 = iDGreen2; + m_pencodingbitsRGB8->differential.dblue2 = iDBlue2; + + m_pencodingbitsRGB8->individual.cw1 = m_uiCW1; + m_pencodingbitsRGB8->individual.cw2 = m_uiCW2; + + SetEncodingBits_Selectors(); + + // in RGB8A1 encoding bits, opaque replaces differential + m_pencodingbitsRGB8->differential.diff = !m_boolPunchThroughPixels; + + m_pencodingbitsRGB8->individual.flip = m_boolFlip; + + } + + // ---------------------------------------------------------------------------------------------------- + // set the encoding bits based on encoding state if T mode + // + void Block4x4Encoding_RGB8A1::SetEncodingBits_T(void) + { + static const bool SANITY_CHECK = true; + + assert(m_mode == MODE_T); + assert(m_boolDiff == true); + + unsigned int uiRed1 = (unsigned int)m_frgbaColor1.IntRed(15.0f); + unsigned int uiGreen1 = (unsigned int)m_frgbaColor1.IntGreen(15.0f); + unsigned int uiBlue1 = (unsigned int)m_frgbaColor1.IntBlue(15.0f); + + unsigned int uiRed2 = (unsigned int)m_frgbaColor2.IntRed(15.0f); + unsigned int uiGreen2 = (unsigned int)m_frgbaColor2.IntGreen(15.0f); + unsigned int uiBlue2 = (unsigned int)m_frgbaColor2.IntBlue(15.0f); + + m_pencodingbitsRGB8->t.red1a = uiRed1 >> 2; + m_pencodingbitsRGB8->t.red1b = uiRed1; + m_pencodingbitsRGB8->t.green1 = uiGreen1; + m_pencodingbitsRGB8->t.blue1 = uiBlue1; + + m_pencodingbitsRGB8->t.red2 = uiRed2; + m_pencodingbitsRGB8->t.green2 = uiGreen2; + m_pencodingbitsRGB8->t.blue2 = uiBlue2; + + m_pencodingbitsRGB8->t.da = m_uiCW1 >> 1; + m_pencodingbitsRGB8->t.db = m_uiCW1; + + // in RGB8A1 encoding bits, opaque replaces differential + m_pencodingbitsRGB8->differential.diff = !m_boolPunchThroughPixels; + + Block4x4Encoding_ETC1::SetEncodingBits_Selectors(); + + // create an invalid R differential to trigger T mode + m_pencodingbitsRGB8->t.detect1 = 0; + m_pencodingbitsRGB8->t.detect2 = 0; + int iRed2 = (int)m_pencodingbitsRGB8->differential.red1 + (int)m_pencodingbitsRGB8->differential.dred2; + if (iRed2 >= 4) + { + m_pencodingbitsRGB8->t.detect1 = 7; + m_pencodingbitsRGB8->t.detect2 = 0; + } + else + { + m_pencodingbitsRGB8->t.detect1 = 0; + m_pencodingbitsRGB8->t.detect2 = 1; + } + + if (SANITY_CHECK) + { + iRed2 = (int)m_pencodingbitsRGB8->differential.red1 + (int)m_pencodingbitsRGB8->differential.dred2; + + // make sure red overflows + assert(iRed2 < 0 || iRed2 > 31); + } + + } + + // ---------------------------------------------------------------------------------------------------- + // set the encoding bits based on encoding state if H mode + // + // colors and selectors may need to swap in order to generate lsb of distance index + // + void Block4x4Encoding_RGB8A1::SetEncodingBits_H(void) + { + static const bool SANITY_CHECK = true; + + assert(m_mode == MODE_H); + assert(m_boolDiff == true); + + unsigned int uiRed1 = (unsigned int)m_frgbaColor1.IntRed(15.0f); + unsigned int uiGreen1 = (unsigned int)m_frgbaColor1.IntGreen(15.0f); + unsigned int uiBlue1 = (unsigned int)m_frgbaColor1.IntBlue(15.0f); + + unsigned int uiRed2 = (unsigned int)m_frgbaColor2.IntRed(15.0f); + unsigned int uiGreen2 = (unsigned int)m_frgbaColor2.IntGreen(15.0f); + unsigned int uiBlue2 = (unsigned int)m_frgbaColor2.IntBlue(15.0f); + + unsigned int uiColor1 = (uiRed1 << 16) + (uiGreen1 << 8) + uiBlue1; + unsigned int uiColor2 = (uiRed2 << 16) + (uiGreen2 << 8) + uiBlue2; + + bool boolOddDistance = m_uiCW1 & 1; + bool boolSwapColors = (uiColor1 < uiColor2) ^ !boolOddDistance; + + if (boolSwapColors) + { + m_pencodingbitsRGB8->h.red1 = uiRed2; + m_pencodingbitsRGB8->h.green1a = uiGreen2 >> 1; + m_pencodingbitsRGB8->h.green1b = uiGreen2; + m_pencodingbitsRGB8->h.blue1a = uiBlue2 >> 3; + m_pencodingbitsRGB8->h.blue1b = uiBlue2 >> 1; + m_pencodingbitsRGB8->h.blue1c = uiBlue2; + + m_pencodingbitsRGB8->h.red2 = uiRed1; + m_pencodingbitsRGB8->h.green2a = uiGreen1 >> 1; + m_pencodingbitsRGB8->h.green2b = uiGreen1; + m_pencodingbitsRGB8->h.blue2 = uiBlue1; + + m_pencodingbitsRGB8->h.da = m_uiCW1 >> 2; + m_pencodingbitsRGB8->h.db = m_uiCW1 >> 1; + } + else + { + m_pencodingbitsRGB8->h.red1 = uiRed1; + m_pencodingbitsRGB8->h.green1a = uiGreen1 >> 1; + m_pencodingbitsRGB8->h.green1b = uiGreen1; + m_pencodingbitsRGB8->h.blue1a = uiBlue1 >> 3; + m_pencodingbitsRGB8->h.blue1b = uiBlue1 >> 1; + m_pencodingbitsRGB8->h.blue1c = uiBlue1; + + m_pencodingbitsRGB8->h.red2 = uiRed2; + m_pencodingbitsRGB8->h.green2a = uiGreen2 >> 1; + m_pencodingbitsRGB8->h.green2b = uiGreen2; + m_pencodingbitsRGB8->h.blue2 = uiBlue2; + + m_pencodingbitsRGB8->h.da = m_uiCW1 >> 2; + m_pencodingbitsRGB8->h.db = m_uiCW1 >> 1; + } + + // in RGB8A1 encoding bits, opaque replaces differential + m_pencodingbitsRGB8->differential.diff = !m_boolPunchThroughPixels; + + Block4x4Encoding_ETC1::SetEncodingBits_Selectors(); + + if (boolSwapColors) + { + m_pencodingbitsRGB8->h.selectors ^= 0x0000FFFF; + } + + // create an invalid R differential to trigger T mode + m_pencodingbitsRGB8->h.detect1 = 0; + m_pencodingbitsRGB8->h.detect2 = 0; + m_pencodingbitsRGB8->h.detect3 = 0; + int iRed2 = (int)m_pencodingbitsRGB8->differential.red1 + (int)m_pencodingbitsRGB8->differential.dred2; + int iGreen2 = (int)m_pencodingbitsRGB8->differential.green1 + (int)m_pencodingbitsRGB8->differential.dgreen2; + if (iRed2 < 0 || iRed2 > 31) + { + m_pencodingbitsRGB8->h.detect1 = 1; + } + if (iGreen2 >= 4) + { + m_pencodingbitsRGB8->h.detect2 = 7; + m_pencodingbitsRGB8->h.detect3 = 0; + } + else + { + m_pencodingbitsRGB8->h.detect2 = 0; + m_pencodingbitsRGB8->h.detect3 = 1; + } + + if (SANITY_CHECK) + { + iRed2 = (int)m_pencodingbitsRGB8->differential.red1 + (int)m_pencodingbitsRGB8->differential.dred2; + iGreen2 = (int)m_pencodingbitsRGB8->differential.green1 + (int)m_pencodingbitsRGB8->differential.dgreen2; + + // make sure red doesn't overflow and green does + assert(iRed2 >= 0 && iRed2 <= 31); + assert(iGreen2 < 0 || iGreen2 > 31); + } + + } + + // #################################################################################################### + // Block4x4Encoding_RGB8A1_Opaque + // #################################################################################################### + + // ---------------------------------------------------------------------------------------------------- + // perform a single encoding iteration + // replace the encoding if a better encoding was found + // subsequent iterations generally take longer for each iteration + // set m_boolDone if encoding is perfect or encoding is finished based on a_fEffort + // + void Block4x4Encoding_RGB8A1_Opaque::PerformIteration(float a_fEffort) + { + assert(!m_boolPunchThroughPixels); + assert(!m_boolTransparent); + assert(!m_boolDone); + + switch (m_uiEncodingIterations) + { + case 0: + PerformFirstIteration(); + break; + + case 1: + Block4x4Encoding_ETC1::TryDifferential(m_boolMostLikelyFlip, 1, 0, 0); + break; + + case 2: + Block4x4Encoding_ETC1::TryDifferential(!m_boolMostLikelyFlip, 1, 0, 0); + break; + + case 3: + Block4x4Encoding_RGB8::TryPlanar(1); + break; + + case 4: + Block4x4Encoding_RGB8::TryTAndH(1); + if (a_fEffort <= 49.5f) + { + m_boolDone = true; + } + break; + + case 5: + Block4x4Encoding_ETC1::TryDegenerates1(); + if (a_fEffort <= 59.5f) + { + m_boolDone = true; + } + break; + + case 6: + Block4x4Encoding_ETC1::TryDegenerates2(); + if (a_fEffort <= 69.5f) + { + m_boolDone = true; + } + break; + + case 7: + Block4x4Encoding_ETC1::TryDegenerates3(); + if (a_fEffort <= 79.5f) + { + m_boolDone = true; + } + break; + + case 8: + Block4x4Encoding_ETC1::TryDegenerates4(); + m_boolDone = true; + break; + + default: + assert(0); + break; + } + + m_uiEncodingIterations++; + SetDoneIfPerfect(); + } + + // ---------------------------------------------------------------------------------------------------- + // find best initial encoding to ensure block has a valid encoding + // + void Block4x4Encoding_RGB8A1_Opaque::PerformFirstIteration(void) + { + + // set decoded alphas + // calculate alpha error + m_fError = 0.0f; + for (unsigned int uiPixel = 0; uiPixel < PIXELS; uiPixel++) + { + m_afDecodedAlphas[uiPixel] = 1.0f; + + float fDeltaA = 1.0f - m_pafrgbaSource[uiPixel].fA; + m_fError += fDeltaA * fDeltaA; + } + + CalculateMostLikelyFlip(); + + m_fError = FLT_MAX; + + Block4x4Encoding_ETC1::TryDifferential(m_boolMostLikelyFlip, 0, 0, 0); + SetDoneIfPerfect(); + if (m_boolDone) + { + return; + } + Block4x4Encoding_ETC1::TryDifferential(!m_boolMostLikelyFlip, 0, 0, 0); + SetDoneIfPerfect(); + if (m_boolDone) + { + return; + } + Block4x4Encoding_RGB8::TryPlanar(0); + SetDoneIfPerfect(); + if (m_boolDone) + { + return; + } + Block4x4Encoding_RGB8::TryTAndH(0); + SetDoneIfPerfect(); + } + + // #################################################################################################### + // Block4x4Encoding_RGB8A1_Transparent + // #################################################################################################### + + // ---------------------------------------------------------------------------------------------------- + // perform a single encoding iteration + // replace the encoding if a better encoding was found + // subsequent iterations generally take longer for each iteration + // set m_boolDone if encoding is perfect or encoding is finished based on a_fEffort + // + void Block4x4Encoding_RGB8A1_Transparent::PerformIteration(float ) + { + assert(!m_boolOpaque); + assert(m_boolTransparent); + assert(!m_boolDone); + assert(m_uiEncodingIterations == 0); + + m_mode = MODE_ETC1; + m_boolDiff = true; + m_boolFlip = false; + + m_uiCW1 = 0; + m_uiCW2 = 0; + + m_frgbaColor1 = ColorFloatRGBA(); + m_frgbaColor2 = ColorFloatRGBA(); + + for (unsigned int uiPixel = 0; uiPixel < PIXELS; uiPixel++) + { + m_auiSelectors[uiPixel] = TRANSPARENT_SELECTOR; + + m_afrgbaDecodedColors[uiPixel] = ColorFloatRGBA(); + m_afDecodedAlphas[uiPixel] = 0.0f; + } + + CalcBlockError(); + + m_boolDone = true; + m_uiEncodingIterations++; + + } + + // ---------------------------------------------------------------------------------------------------- + // +} diff --git a/EtcLib/EtcCodec/EtcBlock4x4Encoding_RGB8A1.h b/EtcLib/EtcCodec/EtcBlock4x4Encoding_RGB8A1.h new file mode 100644 index 0000000..91b9355 --- /dev/null +++ b/EtcLib/EtcCodec/EtcBlock4x4Encoding_RGB8A1.h @@ -0,0 +1,129 @@ +/* + * Copyright 2015 The Etc2Comp Authors. + * + * Licensed under the Apache License, Version 2.0 (the "License"); + * you may not use this file except in compliance with the License. + * You may obtain a copy of the License at + * + * http://www.apache.org/licenses/LICENSE-2.0 + * + * Unless required by applicable law or agreed to in writing, software + * distributed under the License is distributed on an "AS IS" BASIS, + * WITHOUT WARRANTIES OR CONDITIONS OF ANY KIND, either express or implied. + * See the License for the specific language governing permissions and + * limitations under the License. + */ + +#pragma once + +#include "EtcBlock4x4Encoding_RGB8.h" +#include "EtcErrorMetric.h" +#include "EtcBlock4x4EncodingBits.h" + +namespace Etc +{ + + // ################################################################################ + // Block4x4Encoding_RGB8A1 + // RGB8A1 if not completely opaque or transparent + // ################################################################################ + + class Block4x4Encoding_RGB8A1 : public Block4x4Encoding_RGB8 + { + public: + + static const unsigned int TRANSPARENT_SELECTOR = 2; + + Block4x4Encoding_RGB8A1(void); + virtual ~Block4x4Encoding_RGB8A1(void); + + virtual void InitFromSource(Block4x4 *a_pblockParent, + ColorFloatRGBA *a_pafrgbaSource, + unsigned char *a_paucEncodingBits, + ErrorMetric a_errormetric); + + virtual void InitFromEncodingBits(Block4x4 *a_pblockParent, + unsigned char *a_paucEncodingBits, + ColorFloatRGBA *a_pafrgbaSource, + ErrorMetric a_errormetric); + + virtual void PerformIteration(float a_fEffort); + + virtual void SetEncodingBits(void); + + void InitFromEncodingBits_ETC1(Block4x4 *a_pblockParent, + unsigned char *a_paucEncodingBits, + ColorFloatRGBA *a_pafrgbaSource, + ErrorMetric a_errormetric); + + void InitFromEncodingBits_T(void); + void InitFromEncodingBits_H(void); + + void PerformFirstIteration(void); + + void Decode_ETC1(void); + void DecodePixels_T(void); + void DecodePixels_H(void); + void SetEncodingBits_ETC1(void); + void SetEncodingBits_T(void); + void SetEncodingBits_H(void); + + protected: + + bool m_boolOpaque; // all source pixels have alpha >= 0.5 + bool m_boolTransparent; // all source pixels have alpha < 0.5 + bool m_boolPunchThroughPixels; // some source pixels have alpha < 0.5 + + static float s_aafCwOpaqueUnsetTable[CW_RANGES][SELECTORS]; + + private: + + void TryDifferential(bool a_boolFlip, unsigned int a_uiRadius, + int a_iGrayOffset1, int a_iGrayOffset2); + void TryDifferentialHalf(DifferentialTrys::Half *a_phalf); + + void TryT(unsigned int a_uiRadius); + void TryT_BestSelectorCombination(void); + void TryH(unsigned int a_uiRadius); + void TryH_BestSelectorCombination(void); + + void TryDegenerates1(void); + void TryDegenerates2(void); + void TryDegenerates3(void); + void TryDegenerates4(void); + + }; + + // ################################################################################ + // Block4x4Encoding_RGB8A1_Opaque + // RGB8A1 if all pixels have alpha==1 + // ################################################################################ + + class Block4x4Encoding_RGB8A1_Opaque : public Block4x4Encoding_RGB8A1 + { + public: + + virtual void PerformIteration(float a_fEffort); + + void PerformFirstIteration(void); + + private: + + }; + + // ################################################################################ + // Block4x4Encoding_RGB8A1_Transparent + // RGB8A1 if all pixels have alpha==0 + // ################################################################################ + + class Block4x4Encoding_RGB8A1_Transparent : public Block4x4Encoding_RGB8A1 + { + public: + + virtual void PerformIteration(float a_fEffort); + + private: + + }; + +} // namespace Etc diff --git a/EtcLib/EtcCodec/EtcBlock4x4Encoding_RGBA8.cpp b/EtcLib/EtcCodec/EtcBlock4x4Encoding_RGBA8.cpp new file mode 100644 index 0000000..0ca531d --- /dev/null +++ b/EtcLib/EtcCodec/EtcBlock4x4Encoding_RGBA8.cpp @@ -0,0 +1,474 @@ +/* + * Copyright 2015 The Etc2Comp Authors. + * + * Licensed under the Apache License, Version 2.0 (the "License"); + * you may not use this file except in compliance with the License. + * You may obtain a copy of the License at + * + * http://www.apache.org/licenses/LICENSE-2.0 + * + * Unless required by applicable law or agreed to in writing, software + * distributed under the License is distributed on an "AS IS" BASIS, + * WITHOUT WARRANTIES OR CONDITIONS OF ANY KIND, either express or implied. + * See the License for the specific language governing permissions and + * limitations under the License. + */ + +/* +EtcBlock4x4Encoding_RGBA8.cpp contains: + Block4x4Encoding_RGBA8 + Block4x4Encoding_RGBA8_Opaque + Block4x4Encoding_RGBA8_Transparent + +These encoders are used when targetting file format RGBA8. + +Block4x4Encoding_RGBA8_Opaque is used when all pixels in the 4x4 block are opaque +Block4x4Encoding_RGBA8_Transparent is used when all pixels in the 4x4 block are transparent +Block4x4Encoding_RGBA8 is used when there is a mixture of alphas in the 4x4 block + +*/ + +#include "EtcConfig.h" +#include "EtcBlock4x4Encoding_RGBA8.h" + +#include "EtcBlock4x4EncodingBits.h" +#include "EtcBlock4x4.h" + +#include +#include +#include +#include +#include + +namespace Etc +{ + + // #################################################################################################### + // Block4x4Encoding_RGBA8 + // #################################################################################################### + + float Block4x4Encoding_RGBA8::s_aafModifierTable[MODIFIER_TABLE_ENTRYS][ALPHA_SELECTORS] + { + { -3.0f / 255.0f, -6.0f / 255.0f, -9.0f / 255.0f, -15.0f / 255.0f, 2.0f / 255.0f, 5.0f / 255.0f, 8.0f / 255.0f, 14.0f / 255.0f }, + { -3.0f / 255.0f, -7.0f / 255.0f, -10.0f / 255.0f, -13.0f / 255.0f, 2.0f / 255.0f, 6.0f / 255.0f, 9.0f / 255.0f, 12.0f / 255.0f }, + { -2.0f / 255.0f, -5.0f / 255.0f, -8.0f / 255.0f, -13.0f / 255.0f, 1.0f / 255.0f, 4.0f / 255.0f, 7.0f / 255.0f, 12.0f / 255.0f }, + { -2.0f / 255.0f, -4.0f / 255.0f, -6.0f / 255.0f, -13.0f / 255.0f, 1.0f / 255.0f, 3.0f / 255.0f, 5.0f / 255.0f, 12.0f / 255.0f }, + + { -3.0f / 255.0f, -6.0f / 255.0f, -8.0f / 255.0f, -12.0f / 255.0f, 2.0f / 255.0f, 5.0f / 255.0f, 7.0f / 255.0f, 11.0f / 255.0f }, + { -3.0f / 255.0f, -7.0f / 255.0f, -9.0f / 255.0f, -11.0f / 255.0f, 2.0f / 255.0f, 6.0f / 255.0f, 8.0f / 255.0f, 10.0f / 255.0f }, + { -4.0f / 255.0f, -7.0f / 255.0f, -8.0f / 255.0f, -11.0f / 255.0f, 3.0f / 255.0f, 6.0f / 255.0f, 7.0f / 255.0f, 10.0f / 255.0f }, + { -3.0f / 255.0f, -5.0f / 255.0f, -8.0f / 255.0f, -11.0f / 255.0f, 2.0f / 255.0f, 4.0f / 255.0f, 7.0f / 255.0f, 10.0f / 255.0f }, + + { -2.0f / 255.0f, -6.0f / 255.0f, -8.0f / 255.0f, -10.0f / 255.0f, 1.0f / 255.0f, 5.0f / 255.0f, 7.0f / 255.0f, 9.0f / 255.0f }, + { -2.0f / 255.0f, -5.0f / 255.0f, -8.0f / 255.0f, -10.0f / 255.0f, 1.0f / 255.0f, 4.0f / 255.0f, 7.0f / 255.0f, 9.0f / 255.0f }, + { -2.0f / 255.0f, -4.0f / 255.0f, -8.0f / 255.0f, -10.0f / 255.0f, 1.0f / 255.0f, 3.0f / 255.0f, 7.0f / 255.0f, 9.0f / 255.0f }, + { -2.0f / 255.0f, -5.0f / 255.0f, -7.0f / 255.0f, -10.0f / 255.0f, 1.0f / 255.0f, 4.0f / 255.0f, 6.0f / 255.0f, 9.0f / 255.0f }, + + { -3.0f / 255.0f, -4.0f / 255.0f, -7.0f / 255.0f, -10.0f / 255.0f, 2.0f / 255.0f, 3.0f / 255.0f, 6.0f / 255.0f, 9.0f / 255.0f }, + { -1.0f / 255.0f, -2.0f / 255.0f, -3.0f / 255.0f, -10.0f / 255.0f, 0.0f / 255.0f, 1.0f / 255.0f, 2.0f / 255.0f, 9.0f / 255.0f }, + { -4.0f / 255.0f, -6.0f / 255.0f, -8.0f / 255.0f, -9.0f / 255.0f, 3.0f / 255.0f, 5.0f / 255.0f, 7.0f / 255.0f, 8.0f / 255.0f }, + { -3.0f / 255.0f, -5.0f / 255.0f, -7.0f / 255.0f, -9.0f / 255.0f, 2.0f / 255.0f, 4.0f / 255.0f, 6.0f / 255.0f, 8.0f / 255.0f } + }; + + // ---------------------------------------------------------------------------------------------------- + // + Block4x4Encoding_RGBA8::Block4x4Encoding_RGBA8(void) + { + + m_pencodingbitsA8 = nullptr; + + } + Block4x4Encoding_RGBA8::~Block4x4Encoding_RGBA8(void) {} + // ---------------------------------------------------------------------------------------------------- + // initialization prior to encoding + // a_pblockParent points to the block associated with this encoding + // a_errormetric is used to choose the best encoding + // a_pafrgbaSource points to a 4x4 block subset of the source image + // a_paucEncodingBits points to the final encoding bits + // + void Block4x4Encoding_RGBA8::InitFromSource(Block4x4 *a_pblockParent, + ColorFloatRGBA *a_pafrgbaSource, + unsigned char *a_paucEncodingBits, ErrorMetric a_errormetric) + { + Block4x4Encoding::Init(a_pblockParent, a_pafrgbaSource,a_errormetric); + + m_pencodingbitsA8 = (Block4x4EncodingBits_A8 *)a_paucEncodingBits; + m_pencodingbitsRGB8 = (Block4x4EncodingBits_RGB8 *)(a_paucEncodingBits + sizeof(Block4x4EncodingBits_A8)); + + } + + // ---------------------------------------------------------------------------------------------------- + // initialization from the encoding bits of a previous encoding + // a_pblockParent points to the block associated with this encoding + // a_errormetric is used to choose the best encoding + // a_pafrgbaSource points to a 4x4 block subset of the source image + // a_paucEncodingBits points to the final encoding bits of a previous encoding + // + void Block4x4Encoding_RGBA8::InitFromEncodingBits(Block4x4 *a_pblockParent, + unsigned char *a_paucEncodingBits, + ColorFloatRGBA *a_pafrgbaSource, + ErrorMetric a_errormetric) + { + + m_pencodingbitsA8 = (Block4x4EncodingBits_A8 *)a_paucEncodingBits; + m_pencodingbitsRGB8 = (Block4x4EncodingBits_RGB8 *)(a_paucEncodingBits + sizeof(Block4x4EncodingBits_A8)); + + // init RGB portion + Block4x4Encoding_RGB8::InitFromEncodingBits(a_pblockParent, + (unsigned char *) m_pencodingbitsRGB8, + a_pafrgbaSource, + a_errormetric); + + // init A8 portion + // has to be done after InitFromEncodingBits() + { + m_fBase = m_pencodingbitsA8->data.base / 255.0f; + m_fMultiplier = (float)m_pencodingbitsA8->data.multiplier; + m_uiModifierTableIndex = m_pencodingbitsA8->data.table; + + unsigned long long int ulliSelectorBits = 0; + ulliSelectorBits |= (unsigned long long int)m_pencodingbitsA8->data.selectors0 << 40; + ulliSelectorBits |= (unsigned long long int)m_pencodingbitsA8->data.selectors1 << 32; + ulliSelectorBits |= (unsigned long long int)m_pencodingbitsA8->data.selectors2 << 24; + ulliSelectorBits |= (unsigned long long int)m_pencodingbitsA8->data.selectors3 << 16; + ulliSelectorBits |= (unsigned long long int)m_pencodingbitsA8->data.selectors4 << 8; + ulliSelectorBits |= (unsigned long long int)m_pencodingbitsA8->data.selectors5; + for (unsigned int uiPixel = 0; uiPixel < PIXELS; uiPixel++) + { + unsigned int uiShift = 45 - (3 * uiPixel); + m_auiAlphaSelectors[uiPixel] = (ulliSelectorBits >> uiShift) & (ALPHA_SELECTORS - 1); + } + + // decode the alphas + // calc alpha error + m_fError = 0.0f; + for (unsigned int uiPixel = 0; uiPixel < PIXELS; uiPixel++) + { + m_afDecodedAlphas[uiPixel] = DecodePixelAlpha(m_fBase, m_fMultiplier, + m_uiModifierTableIndex, + m_auiAlphaSelectors[uiPixel]); + + float fDeltaAlpha = m_afDecodedAlphas[uiPixel] - m_pafrgbaSource[uiPixel].fA; + m_fError += fDeltaAlpha * fDeltaAlpha; + } + } + + // redo error calc to include alpha + CalcBlockError(); + + } + + // ---------------------------------------------------------------------------------------------------- + // perform a single encoding iteration + // replace the encoding if a better encoding was found + // subsequent iterations generally take longer for each iteration + // set m_boolDone if encoding is perfect or encoding is finished based on a_fEffort + // + // similar to Block4x4Encoding_RGB8_Base::Encode_RGB8(), but with alpha added + // + void Block4x4Encoding_RGBA8::PerformIteration(float a_fEffort) + { + assert(!m_boolDone); + + if (m_uiEncodingIterations == 0) + { + if (a_fEffort < 24.9f) + { + CalculateA8(0.0f); + } + else if (a_fEffort < 49.9f) + { + CalculateA8(1.0f); + } + else + { + CalculateA8(2.0f); + } + } + + Block4x4Encoding_RGB8::PerformIteration(a_fEffort); + + } + + // ---------------------------------------------------------------------------------------------------- + // find the best combination of base alpga, multiplier and selectors + // + // a_fRadius limits the range of base alpha to try + // + void Block4x4Encoding_RGBA8::CalculateA8(float a_fRadius) + { + + // find min/max alpha + float fMinAlpha = 1.0f; + float fMaxAlpha = 0.0f; + for (unsigned int uiPixel = 0; uiPixel < PIXELS; uiPixel++) + { + float fAlpha = m_pafrgbaSource[uiPixel].fA; + + // ignore border pixels + if (isnan(fAlpha)) + { + continue; + } + + if (fAlpha < fMinAlpha) + { + fMinAlpha = fAlpha; + } + if (fAlpha > fMaxAlpha) + { + fMaxAlpha = fAlpha; + } + } + assert(fMinAlpha <= fMaxAlpha); + + float fAlphaRange = fMaxAlpha - fMinAlpha; + + // try each modifier table entry + m_fError = FLT_MAX; // artificially high value + for (unsigned int uiTableEntry = 0; uiTableEntry < MODIFIER_TABLE_ENTRYS; uiTableEntry++) + { + static const unsigned int MIN_VALUE_SELECTOR = 3; + static const unsigned int MAX_VALUE_SELECTOR = 7; + + float fTableEntryCenter = -s_aafModifierTable[uiTableEntry][MIN_VALUE_SELECTOR]; + + float fTableEntryRange = s_aafModifierTable[uiTableEntry][MAX_VALUE_SELECTOR] - + s_aafModifierTable[uiTableEntry][MIN_VALUE_SELECTOR]; + + float fCenterRatio = fTableEntryCenter / fTableEntryRange; + + float fCenter = fMinAlpha + fCenterRatio*fAlphaRange; + fCenter = roundf(255.0f * fCenter) / 255.0f; + + float fMinBase = fCenter - (a_fRadius / 255.0f); + if (fMinBase < 0.0f) + { + fMinBase = 0.0f; + } + + float fMaxBase = fCenter + (a_fRadius / 255.0f); + if (fMaxBase > 1.0f) + { + fMaxBase = 1.0f; + } + + for (float fBase = fMinBase; fBase <= fMaxBase; fBase += (0.999999f / 255.0f)) + { + + float fRangeMultiplier = roundf(fAlphaRange / fTableEntryRange); + + float fMinMultiplier = fRangeMultiplier - a_fRadius; + if (fMinMultiplier < 1.0f) + { + fMinMultiplier = 1.0f; + } + else if (fMinMultiplier > 15.0f) + { + fMinMultiplier = 15.0f; + } + + float fMaxMultiplier = fRangeMultiplier + a_fRadius; + if (fMaxMultiplier < 1.0f) + { + fMaxMultiplier = 1.0f; + } + else if (fMaxMultiplier > 15.0f) + { + fMaxMultiplier = 15.0f; + } + + for (float fMultiplier = fMinMultiplier; fMultiplier <= fMaxMultiplier; fMultiplier += 1.0f) + { + // find best selector for each pixel + unsigned int auiBestSelectors[PIXELS]; + float afBestAlphaError[PIXELS]; + float afBestDecodedAlphas[PIXELS]; + for (unsigned int uiPixel = 0; uiPixel < PIXELS; uiPixel++) + { + float fBestPixelAlphaError = FLT_MAX; + for (unsigned int uiSelector = 0; uiSelector < ALPHA_SELECTORS; uiSelector++) + { + float fDecodedAlpha = DecodePixelAlpha(fBase, fMultiplier, uiTableEntry, uiSelector); + + // border pixels (NAN) should have zero error + float fPixelDeltaAlpha = isnan(m_pafrgbaSource[uiPixel].fA) ? + 0.0f : + fDecodedAlpha - m_pafrgbaSource[uiPixel].fA; + + float fPixelAlphaError = fPixelDeltaAlpha * fPixelDeltaAlpha; + + if (fPixelAlphaError < fBestPixelAlphaError) + { + fBestPixelAlphaError = fPixelAlphaError; + auiBestSelectors[uiPixel] = uiSelector; + afBestAlphaError[uiPixel] = fBestPixelAlphaError; + afBestDecodedAlphas[uiPixel] = fDecodedAlpha; + } + } + } + + float fBlockError = 0.0f; + for (unsigned int uiPixel = 0; uiPixel < PIXELS; uiPixel++) + { + fBlockError += afBestAlphaError[uiPixel]; + } + + if (fBlockError < m_fError) + { + m_fError = fBlockError; + + m_fBase = fBase; + m_fMultiplier = fMultiplier; + m_uiModifierTableIndex = uiTableEntry; + for (unsigned int uiPixel = 0; uiPixel < PIXELS; uiPixel++) + { + m_auiAlphaSelectors[uiPixel] = auiBestSelectors[uiPixel]; + m_afDecodedAlphas[uiPixel] = afBestDecodedAlphas[uiPixel]; + } + } + } + } + + } + + } + + // ---------------------------------------------------------------------------------------------------- + // set the encoding bits based on encoding state + // + void Block4x4Encoding_RGBA8::SetEncodingBits(void) + { + + // set the RGB8 portion + Block4x4Encoding_RGB8::SetEncodingBits(); + + // set the A8 portion + { + m_pencodingbitsA8->data.base = (unsigned char)roundf(255.0f * m_fBase); + m_pencodingbitsA8->data.table = m_uiModifierTableIndex; + m_pencodingbitsA8->data.multiplier = (unsigned char)roundf(m_fMultiplier); + + unsigned long long int ulliSelectorBits = 0; + for (unsigned int uiPixel = 0; uiPixel < PIXELS; uiPixel++) + { + unsigned int uiShift = 45 - (3 * uiPixel); + ulliSelectorBits |= ((unsigned long long int)m_auiAlphaSelectors[uiPixel]) << uiShift; + } + + m_pencodingbitsA8->data.selectors0 = ulliSelectorBits >> 40; + m_pencodingbitsA8->data.selectors1 = ulliSelectorBits >> 32; + m_pencodingbitsA8->data.selectors2 = ulliSelectorBits >> 24; + m_pencodingbitsA8->data.selectors3 = ulliSelectorBits >> 16; + m_pencodingbitsA8->data.selectors4 = ulliSelectorBits >> 8; + m_pencodingbitsA8->data.selectors5 = ulliSelectorBits; + } + + } + + // #################################################################################################### + // Block4x4Encoding_RGBA8_Opaque + // #################################################################################################### + + // ---------------------------------------------------------------------------------------------------- + // perform a single encoding iteration + // replace the encoding if a better encoding was found + // subsequent iterations generally take longer for each iteration + // set m_boolDone if encoding is perfect or encoding is finished based on a_fEffort + // + void Block4x4Encoding_RGBA8_Opaque::PerformIteration(float a_fEffort) + { + assert(!m_boolDone); + + if (m_uiEncodingIterations == 0) + { + m_fError = 0.0f; + + for (unsigned int uiPixel = 0; uiPixel < PIXELS; uiPixel++) + { + m_afDecodedAlphas[uiPixel] = 1.0f; + } + } + + Block4x4Encoding_RGB8::PerformIteration(a_fEffort); + + } + + // ---------------------------------------------------------------------------------------------------- + // set the encoding bits based on encoding state + // + void Block4x4Encoding_RGBA8_Opaque::SetEncodingBits(void) + { + + // set the RGB8 portion + Block4x4Encoding_RGB8::SetEncodingBits(); + + // set the A8 portion + m_pencodingbitsA8->data.base = 255; + m_pencodingbitsA8->data.table = 15; + m_pencodingbitsA8->data.multiplier = 15; + m_pencodingbitsA8->data.selectors0 = 0xFF; + m_pencodingbitsA8->data.selectors1 = 0xFF; + m_pencodingbitsA8->data.selectors2 = 0xFF; + m_pencodingbitsA8->data.selectors3 = 0xFF; + m_pencodingbitsA8->data.selectors4 = 0xFF; + m_pencodingbitsA8->data.selectors5 = 0xFF; + + } + + // #################################################################################################### + // Block4x4Encoding_RGBA8_Transparent + // #################################################################################################### + + // ---------------------------------------------------------------------------------------------------- + // perform a single encoding iteration + // replace the encoding if a better encoding was found + // subsequent iterations generally take longer for each iteration + // set m_boolDone if encoding is perfect or encoding is finished based on a_fEffort + // + void Block4x4Encoding_RGBA8_Transparent::PerformIteration(float ) + { + assert(!m_boolDone); + assert(m_uiEncodingIterations == 0); + + m_mode = MODE_ETC1; + m_boolDiff = true; + m_boolFlip = false; + + for (unsigned int uiPixel = 0; uiPixel < PIXELS; uiPixel++) + { + m_afrgbaDecodedColors[uiPixel] = ColorFloatRGBA(); + m_afDecodedAlphas[uiPixel] = 0.0f; + } + + m_fError = 0.0f; + + m_boolDone = true; + m_uiEncodingIterations++; + + } + + // ---------------------------------------------------------------------------------------------------- + // set the encoding bits based on encoding state + // + void Block4x4Encoding_RGBA8_Transparent::SetEncodingBits(void) + { + + Block4x4Encoding_RGB8::SetEncodingBits(); + + // set the A8 portion + m_pencodingbitsA8->data.base = 0; + m_pencodingbitsA8->data.table = 0; + m_pencodingbitsA8->data.multiplier = 1; + m_pencodingbitsA8->data.selectors0 = 0; + m_pencodingbitsA8->data.selectors1 = 0; + m_pencodingbitsA8->data.selectors2 = 0; + m_pencodingbitsA8->data.selectors3 = 0; + m_pencodingbitsA8->data.selectors4 = 0; + m_pencodingbitsA8->data.selectors5 = 0; + + } + + // ---------------------------------------------------------------------------------------------------- + // +} diff --git a/EtcLib/EtcCodec/EtcBlock4x4Encoding_RGBA8.h b/EtcLib/EtcCodec/EtcBlock4x4Encoding_RGBA8.h new file mode 100644 index 0000000..9d21e90 --- /dev/null +++ b/EtcLib/EtcCodec/EtcBlock4x4Encoding_RGBA8.h @@ -0,0 +1,121 @@ +/* + * Copyright 2015 The Etc2Comp Authors. + * + * Licensed under the Apache License, Version 2.0 (the "License"); + * you may not use this file except in compliance with the License. + * You may obtain a copy of the License at + * + * http://www.apache.org/licenses/LICENSE-2.0 + * + * Unless required by applicable law or agreed to in writing, software + * distributed under the License is distributed on an "AS IS" BASIS, + * WITHOUT WARRANTIES OR CONDITIONS OF ANY KIND, either express or implied. + * See the License for the specific language governing permissions and + * limitations under the License. + */ + +#pragma once + +#include "EtcBlock4x4Encoding_RGB8.h" + +namespace Etc +{ + class Block4x4EncodingBits_A8; + + // ################################################################################ + // Block4x4Encoding_RGBA8 + // RGBA8 if not completely opaque or transparent + // ################################################################################ + + class Block4x4Encoding_RGBA8 : public Block4x4Encoding_RGB8 + { + public: + + Block4x4Encoding_RGBA8(void); + virtual ~Block4x4Encoding_RGBA8(void); + + virtual void InitFromSource(Block4x4 *a_pblockParent, + ColorFloatRGBA *a_pafrgbaSource, + unsigned char *a_paucEncodingBits, ErrorMetric a_errormetric); + + virtual void InitFromEncodingBits(Block4x4 *a_pblockParent, + unsigned char *a_paucEncodingBits, + ColorFloatRGBA *a_pafrgbaSource, + ErrorMetric a_errormetric); + + virtual void PerformIteration(float a_fEffort); + + virtual void SetEncodingBits(void); + + protected: + + static const unsigned int MODIFIER_TABLE_ENTRYS = 16; + static const unsigned int ALPHA_SELECTOR_BITS = 3; + static const unsigned int ALPHA_SELECTORS = 1 << ALPHA_SELECTOR_BITS; + + static float s_aafModifierTable[MODIFIER_TABLE_ENTRYS][ALPHA_SELECTORS]; + + void CalculateA8(float a_fRadius); + + Block4x4EncodingBits_A8 *m_pencodingbitsA8; // A8 portion of Block4x4EncodingBits_RGBA8 + + float m_fBase; + float m_fMultiplier; + unsigned int m_uiModifierTableIndex; + unsigned int m_auiAlphaSelectors[PIXELS]; + + private: + + inline float DecodePixelAlpha(float a_fBase, float a_fMultiplier, + unsigned int a_uiTableIndex, unsigned int a_uiSelector) + { + float fPixelAlpha = a_fBase + + a_fMultiplier*s_aafModifierTable[a_uiTableIndex][a_uiSelector]; + if (fPixelAlpha < 0.0f) + { + fPixelAlpha = 0.0f; + } + else if (fPixelAlpha > 1.0f) + { + fPixelAlpha = 1.0f; + } + + return fPixelAlpha; + } + + }; + + // ################################################################################ + // Block4x4Encoding_RGBA8_Opaque + // RGBA8 if all pixels have alpha==1 + // ################################################################################ + + class Block4x4Encoding_RGBA8_Opaque : public Block4x4Encoding_RGBA8 + { + public: + + virtual void PerformIteration(float a_fEffort); + + virtual void SetEncodingBits(void); + + }; + + // ################################################################################ + // Block4x4Encoding_RGBA8_Transparent + // RGBA8 if all pixels have alpha==0 + // ################################################################################ + + class Block4x4Encoding_RGBA8_Transparent : public Block4x4Encoding_RGBA8 + { + public: + + virtual void PerformIteration(float a_fEffort); + + virtual void SetEncodingBits(void); + + }; + + // ---------------------------------------------------------------------------------------------------- + // + +} // namespace Etc diff --git a/EtcLib/EtcCodec/EtcDifferentialTrys.cpp b/EtcLib/EtcCodec/EtcDifferentialTrys.cpp new file mode 100644 index 0000000..0fcc550 --- /dev/null +++ b/EtcLib/EtcCodec/EtcDifferentialTrys.cpp @@ -0,0 +1,173 @@ +/* + * Copyright 2015 The Etc2Comp Authors. + * + * Licensed under the Apache License, Version 2.0 (the "License"); + * you may not use this file except in compliance with the License. + * You may obtain a copy of the License at + * + * http://www.apache.org/licenses/LICENSE-2.0 + * + * Unless required by applicable law or agreed to in writing, software + * distributed under the License is distributed on an "AS IS" BASIS, + * WITHOUT WARRANTIES OR CONDITIONS OF ANY KIND, either express or implied. + * See the License for the specific language governing permissions and + * limitations under the License. + */ + +/* +EtcDifferentialTrys.cpp + +Gathers the results of the various encoding trys for both halves of a 4x4 block for Differential mode + +*/ + +#include "EtcConfig.h" +#include "EtcDifferentialTrys.h" + +#include + +namespace Etc +{ + + // ---------------------------------------------------------------------------------------------------- + // construct a list of trys (encoding attempts) + // + // a_frgbaColor1 is the basecolor for the first half + // a_frgbaColor2 is the basecolor for the second half + // a_pauiPixelMapping1 is the pixel order for the first half + // a_pauiPixelMapping2 is the pixel order for the second half + // a_uiRadius is the amount to vary the base colors + // + DifferentialTrys::DifferentialTrys(ColorFloatRGBA a_frgbaColor1, ColorFloatRGBA a_frgbaColor2, + const unsigned int *a_pauiPixelMapping1, + const unsigned int *a_pauiPixelMapping2, + unsigned int a_uiRadius, + int a_iGrayOffset1, int a_iGrayOffset2) + { + assert(a_uiRadius <= MAX_RADIUS); + + m_boolSeverelyBentColors = false; + + ColorFloatRGBA frgbaQuantizedColor1 = a_frgbaColor1.QuantizeR5G5B5(); + ColorFloatRGBA frgbaQuantizedColor2 = a_frgbaColor2.QuantizeR5G5B5(); + + // quantize base colors + // ensure that trys with a_uiRadius don't overflow + int iRed1 = MoveAwayFromEdge(frgbaQuantizedColor1.IntRed(31.0f)+a_iGrayOffset1, a_uiRadius); + int iGreen1 = MoveAwayFromEdge(frgbaQuantizedColor1.IntGreen(31.0f) + a_iGrayOffset1, a_uiRadius); + int iBlue1 = MoveAwayFromEdge(frgbaQuantizedColor1.IntBlue(31.0f) + a_iGrayOffset1, a_uiRadius); + int iRed2 = MoveAwayFromEdge(frgbaQuantizedColor2.IntRed(31.0f) + a_iGrayOffset2, a_uiRadius); + int iGreen2 = MoveAwayFromEdge(frgbaQuantizedColor2.IntGreen(31.0f) + a_iGrayOffset2, a_uiRadius); + int iBlue2 = MoveAwayFromEdge(frgbaQuantizedColor2.IntBlue(31.0f) + a_iGrayOffset2, a_uiRadius); + + int iDeltaRed = iRed2 - iRed1; + int iDeltaGreen = iGreen2 - iGreen1; + int iDeltaBlue = iBlue2 - iBlue1; + + // make sure components are within range + { + if (iDeltaRed > 3) + { + if (iDeltaRed > 7) + { + m_boolSeverelyBentColors = true; + } + + iRed1 += (iDeltaRed - 3) / 2; + iRed2 = iRed1 + 3; + iDeltaRed = 3; + } + else if (iDeltaRed < -4) + { + if (iDeltaRed < -8) + { + m_boolSeverelyBentColors = true; + } + + iRed1 += (iDeltaRed + 4) / 2; + iRed2 = iRed1 - 4; + iDeltaRed = -4; + } + assert(iRed1 >= (signed)(0 + a_uiRadius) && iRed1 <= (signed)(31 - a_uiRadius)); + assert(iRed2 >= (signed)(0 + a_uiRadius) && iRed2 <= (signed)(31 - a_uiRadius)); + assert(iDeltaRed >= -4 && iDeltaRed <= 3); + + if (iDeltaGreen > 3) + { + if (iDeltaGreen > 7) + { + m_boolSeverelyBentColors = true; + } + + iGreen1 += (iDeltaGreen - 3) / 2; + iGreen2 = iGreen1 + 3; + iDeltaGreen = 3; + } + else if (iDeltaGreen < -4) + { + if (iDeltaGreen < -8) + { + m_boolSeverelyBentColors = true; + } + + iGreen1 += (iDeltaGreen + 4) / 2; + iGreen2 = iGreen1 - 4; + iDeltaGreen = -4; + } + assert(iGreen1 >= (signed)(0 + a_uiRadius) && iGreen1 <= (signed)(31 - a_uiRadius)); + assert(iGreen2 >= (signed)(0 + a_uiRadius) && iGreen2 <= (signed)(31 - a_uiRadius)); + assert(iDeltaGreen >= -4 && iDeltaGreen <= 3); + + if (iDeltaBlue > 3) + { + if (iDeltaBlue > 7) + { + m_boolSeverelyBentColors = true; + } + + iBlue1 += (iDeltaBlue - 3) / 2; + iBlue2 = iBlue1 + 3; + iDeltaBlue = 3; + } + else if (iDeltaBlue < -4) + { + if (iDeltaBlue < -8) + { + m_boolSeverelyBentColors = true; + } + + iBlue1 += (iDeltaBlue + 4) / 2; + iBlue2 = iBlue1 - 4; + iDeltaBlue = -4; + } + assert(iBlue1 >= (signed)(0+a_uiRadius) && iBlue1 <= (signed)(31 - a_uiRadius)); + assert(iBlue2 >= (signed)(0 + a_uiRadius) && iBlue2 <= (signed)(31 - a_uiRadius)); + assert(iDeltaBlue >= -4 && iDeltaBlue <= 3); + } + + m_half1.Init(iRed1, iGreen1, iBlue1, a_pauiPixelMapping1, a_uiRadius); + m_half2.Init(iRed2, iGreen2, iBlue2, a_pauiPixelMapping2, a_uiRadius); + + } + + // ---------------------------------------------------------------------------------------------------- + // + void DifferentialTrys::Half::Init(int a_iRed, int a_iGreen, int a_iBlue, + const unsigned int *a_pauiPixelMapping, unsigned int a_uiRadius) + { + + m_iRed = a_iRed; + m_iGreen = a_iGreen; + m_iBlue = a_iBlue; + + m_pauiPixelMapping = a_pauiPixelMapping; + m_uiRadius = a_uiRadius; + + m_uiTrys = 0; + + } + + // ---------------------------------------------------------------------------------------------------- + // + +} // namespace Etc diff --git a/EtcLib/EtcCodec/EtcDifferentialTrys.h b/EtcLib/EtcCodec/EtcDifferentialTrys.h new file mode 100644 index 0000000..6b1cd9c --- /dev/null +++ b/EtcLib/EtcCodec/EtcDifferentialTrys.h @@ -0,0 +1,97 @@ +/* + * Copyright 2015 The Etc2Comp Authors. + * + * Licensed under the Apache License, Version 2.0 (the "License"); + * you may not use this file except in compliance with the License. + * You may obtain a copy of the License at + * + * http://www.apache.org/licenses/LICENSE-2.0 + * + * Unless required by applicable law or agreed to in writing, software + * distributed under the License is distributed on an "AS IS" BASIS, + * WITHOUT WARRANTIES OR CONDITIONS OF ANY KIND, either express or implied. + * See the License for the specific language governing permissions and + * limitations under the License. + */ + +#pragma once + +#include "EtcColorFloatRGBA.h" + +namespace Etc +{ + + class DifferentialTrys + { + public: + + static const unsigned int MAX_RADIUS = 2; + + DifferentialTrys(ColorFloatRGBA a_frgbaColor1, + ColorFloatRGBA a_frgbaColor2, + const unsigned int *a_pauiPixelMapping1, + const unsigned int *a_pauiPixelMapping2, + unsigned int a_uiRadius, + int a_iGrayOffset1, int a_iGrayOffset2); + + inline static int MoveAwayFromEdge(int a_i, int a_iDistance) + { + if (a_i < (0+ a_iDistance)) + { + return (0 + a_iDistance); + } + else if (a_i > (31- a_iDistance)) + { + return (31 - a_iDistance); + } + + return a_i; + } + + class Try + { + public : + static const unsigned int SELECTORS = 8; // per half + + int m_iRed; + int m_iGreen; + int m_iBlue; + unsigned int m_uiCW; + unsigned int m_auiSelectors[SELECTORS]; + float m_fError; + }; + + class Half + { + public: + + static const unsigned int MAX_TRYS = 125; + + void Init(int a_iRed, int a_iGreen, int a_iBlue, + const unsigned int *a_pauiPixelMapping, + unsigned int a_uiRadius); + + // center of trys + int m_iRed; + int m_iGreen; + int m_iBlue; + + const unsigned int *m_pauiPixelMapping; + unsigned int m_uiRadius; + + unsigned int m_uiTrys; + Try m_atry[MAX_TRYS]; + + Try *m_ptryBest; + }; + + Half m_half1; + Half m_half2; + + bool m_boolSeverelyBentColors; + }; + + // ---------------------------------------------------------------------------------------------------- + // + +} // namespace Etc diff --git a/EtcLib/EtcCodec/EtcErrorMetric.h b/EtcLib/EtcCodec/EtcErrorMetric.h new file mode 100644 index 0000000..29bb33b --- /dev/null +++ b/EtcLib/EtcCodec/EtcErrorMetric.h @@ -0,0 +1,51 @@ +/* + * Copyright 2015 The Etc2Comp Authors. + * + * Licensed under the Apache License, Version 2.0 (the "License"); + * you may not use this file except in compliance with the License. + * You may obtain a copy of the License at + * + * http://www.apache.org/licenses/LICENSE-2.0 + * + * Unless required by applicable law or agreed to in writing, software + * distributed under the License is distributed on an "AS IS" BASIS, + * WITHOUT WARRANTIES OR CONDITIONS OF ANY KIND, either express or implied. + * See the License for the specific language governing permissions and + * limitations under the License. + */ + +#pragma once + +namespace Etc +{ + + enum ErrorMetric + { + RGBA, + REC709, + NUMERIC, + NORMALXYZ, + // + ERROR_METRICS, + // + BT709 = REC709 + }; + + inline const char *ErrorMetricToString(ErrorMetric errorMetric) + { + switch (errorMetric) + { + case RGBA: + return "RGBA"; + case REC709: + return "REC709"; + case NUMERIC: + return "NUMERIC"; + case NORMALXYZ: + return "NORMALXYZ"; + case ERROR_METRICS: + default: + return "UNKNOWN"; + } + } +} // namespace Etc diff --git a/EtcLib/EtcCodec/EtcIndividualTrys.cpp b/EtcLib/EtcCodec/EtcIndividualTrys.cpp new file mode 100644 index 0000000..951edec --- /dev/null +++ b/EtcLib/EtcCodec/EtcIndividualTrys.cpp @@ -0,0 +1,85 @@ +/* + * Copyright 2015 The Etc2Comp Authors. + * + * Licensed under the Apache License, Version 2.0 (the "License"); + * you may not use this file except in compliance with the License. + * You may obtain a copy of the License at + * + * http://www.apache.org/licenses/LICENSE-2.0 + * + * Unless required by applicable law or agreed to in writing, software + * distributed under the License is distributed on an "AS IS" BASIS, + * WITHOUT WARRANTIES OR CONDITIONS OF ANY KIND, either express or implied. + * See the License for the specific language governing permissions and + * limitations under the License. + */ + +/* +EtcIndividualTrys.cpp + +Gathers the results of the various encoding trys for both halves of a 4x4 block for Individual mode + +*/ + +#include "EtcConfig.h" +#include "EtcIndividualTrys.h" + +#include + +namespace Etc +{ + + // ---------------------------------------------------------------------------------------------------- + // construct a list of trys (encoding attempts) + // + // a_frgbaColor1 is the basecolor for the first half + // a_frgbaColor2 is the basecolor for the second half + // a_pauiPixelMapping1 is the pixel order for the first half + // a_pauiPixelMapping2 is the pixel order for the second half + // a_uiRadius is the amount to vary the base colors + // + IndividualTrys::IndividualTrys(ColorFloatRGBA a_frgbaColor1, ColorFloatRGBA a_frgbaColor2, + const unsigned int *a_pauiPixelMapping1, + const unsigned int *a_pauiPixelMapping2, + unsigned int a_uiRadius) + { + assert(a_uiRadius <= MAX_RADIUS); + + ColorFloatRGBA frgbaQuantizedColor1 = a_frgbaColor1.QuantizeR4G4B4(); + ColorFloatRGBA frgbaQuantizedColor2 = a_frgbaColor2.QuantizeR4G4B4(); + + // quantize base colors + // ensure that trys with a_uiRadius don't overflow + int iRed1 = MoveAwayFromEdge(frgbaQuantizedColor1.IntRed(15.0f), a_uiRadius); + int iGreen1 = MoveAwayFromEdge(frgbaQuantizedColor1.IntGreen(15.0f), a_uiRadius); + int iBlue1 = MoveAwayFromEdge(frgbaQuantizedColor1.IntBlue(15.0f), a_uiRadius); + int iRed2 = MoveAwayFromEdge(frgbaQuantizedColor2.IntRed(15.0f), a_uiRadius); + int iGreen2 = MoveAwayFromEdge(frgbaQuantizedColor2.IntGreen(15.0f), a_uiRadius); + int iBlue2 = MoveAwayFromEdge(frgbaQuantizedColor2.IntBlue(15.0f), a_uiRadius); + + m_half1.Init(iRed1, iGreen1, iBlue1, a_pauiPixelMapping1, a_uiRadius); + m_half2.Init(iRed2, iGreen2, iBlue2, a_pauiPixelMapping2, a_uiRadius); + + } + + // ---------------------------------------------------------------------------------------------------- + // + void IndividualTrys::Half::Init(int a_iRed, int a_iGreen, int a_iBlue, + const unsigned int *a_pauiPixelMapping, unsigned int a_uiRadius) + { + + m_iRed = a_iRed; + m_iGreen = a_iGreen; + m_iBlue = a_iBlue; + + m_pauiPixelMapping = a_pauiPixelMapping; + m_uiRadius = a_uiRadius; + + m_uiTrys = 0; + + } + + // ---------------------------------------------------------------------------------------------------- + // + +} // namespace Etc diff --git a/EtcLib/EtcCodec/EtcIndividualTrys.h b/EtcLib/EtcCodec/EtcIndividualTrys.h new file mode 100644 index 0000000..49170d4 --- /dev/null +++ b/EtcLib/EtcCodec/EtcIndividualTrys.h @@ -0,0 +1,95 @@ +/* + * Copyright 2015 The Etc2Comp Authors. + * + * Licensed under the Apache License, Version 2.0 (the "License"); + * you may not use this file except in compliance with the License. + * You may obtain a copy of the License at + * + * http://www.apache.org/licenses/LICENSE-2.0 + * + * Unless required by applicable law or agreed to in writing, software + * distributed under the License is distributed on an "AS IS" BASIS, + * WITHOUT WARRANTIES OR CONDITIONS OF ANY KIND, either express or implied. + * See the License for the specific language governing permissions and + * limitations under the License. + */ + +#pragma once + +#include "EtcColorFloatRGBA.h" + +namespace Etc +{ + + class IndividualTrys + { + public: + + static const unsigned int MAX_RADIUS = 1; + + IndividualTrys(ColorFloatRGBA a_frgbaColor1, + ColorFloatRGBA a_frgbaColor2, + const unsigned int *a_pauiPixelMapping1, + const unsigned int *a_pauiPixelMapping2, + unsigned int a_uiRadius); + + inline static int MoveAwayFromEdge(int a_i, int a_iDistance) + { + if (a_i < (0+ a_iDistance)) + { + return (0 + a_iDistance); + } + else if (a_i > (15- a_iDistance)) + { + return (15 - a_iDistance); + } + + return a_i; + } + + class Try + { + public : + static const unsigned int SELECTORS = 8; // per half + + int m_iRed; + int m_iGreen; + int m_iBlue; + unsigned int m_uiCW; + unsigned int m_auiSelectors[SELECTORS]; + float m_fError; + }; + + class Half + { + public: + + static const unsigned int MAX_TRYS = 27; + + void Init(int a_iRed, int a_iGreen, int a_iBlue, + const unsigned int *a_pauiPixelMapping, + unsigned int a_uiRadius); + + // center of trys + int m_iRed; + int m_iGreen; + int m_iBlue; + + const unsigned int *m_pauiPixelMapping; + unsigned int m_uiRadius; + + unsigned int m_uiTrys; + Try m_atry[MAX_TRYS]; + + Try *m_ptryBest; + }; + + Half m_half1; + Half m_half2; + + }; + + // ---------------------------------------------------------------------------------------------------- + // + +} // namespace Etc diff --git a/EtcLib/EtcCodec/EtcSortedBlockList.cpp b/EtcLib/EtcCodec/EtcSortedBlockList.cpp new file mode 100644 index 0000000..7f4f56e --- /dev/null +++ b/EtcLib/EtcCodec/EtcSortedBlockList.cpp @@ -0,0 +1,228 @@ +/* + * Copyright 2015 The Etc2Comp Authors. + * + * Licensed under the Apache License, Version 2.0 (the "License"); + * you may not use this file except in compliance with the License. + * You may obtain a copy of the License at + * + * http://www.apache.org/licenses/LICENSE-2.0 + * + * Unless required by applicable law or agreed to in writing, software + * distributed under the License is distributed on an "AS IS" BASIS, + * WITHOUT WARRANTIES OR CONDITIONS OF ANY KIND, either express or implied. + * See the License for the specific language governing permissions and + * limitations under the License. + */ + +/* +EtcSortedBlockList.cpp + +SortedBlockList is a list of 4x4 blocks that can be used by the "effort" system to prioritize +the encoding of the 4x4 blocks. + +The sorting is done with buckets, where each bucket is an indication of how much error each 4x4 block has + +*/ + +#include "EtcConfig.h" +#include "EtcSortedBlockList.h" + +#include "EtcBlock4x4.h" + +#include +#include +#include + +namespace Etc +{ + + // ---------------------------------------------------------------------------------------------------- + // construct an empty list + // + // allocate enough memory to add all of the image's 4x4 blocks later + // allocate enough buckets to sort the blocks + // + SortedBlockList::SortedBlockList(unsigned int a_uiImageBlocks, unsigned int a_uiBuckets) + { + m_uiImageBlocks = a_uiImageBlocks; + m_iBuckets = (int)a_uiBuckets; + + m_uiAddedBlocks = 0; + m_uiSortedBlocks = 0; + m_palinkPool = new Link[m_uiImageBlocks]; + m_pabucket = new Bucket[m_iBuckets]; + m_fMaxError = 0.0f; + + InitBuckets(); + + } + + // ---------------------------------------------------------------------------------------------------- + // + SortedBlockList::~SortedBlockList(void) + { + delete[] m_palinkPool; + delete[] m_pabucket; + } + + // ---------------------------------------------------------------------------------------------------- + // add a 4x4 block to the list + // the 4x4 block will be sorted later + // + void SortedBlockList::AddBlock(Block4x4 *a_pblock) + { + assert(m_uiAddedBlocks < m_uiImageBlocks); + Link *plink = &m_palinkPool[m_uiAddedBlocks++]; + plink->Init(a_pblock); + } + + // ---------------------------------------------------------------------------------------------------- + // sort all of the 4x4 blocks that have been added to the list + // + // first, determine the maximum error, then assign an error range to each bucket + // next, determine which bucket each 4x4 block belongs to based on the 4x4 block's error + // add the 4x4 block to the appropriate bucket + // lastly, walk thru the buckets and add each bucket to a sorted linked list + // + // the resultant sorting is an approximate sorting from most to least error + // + void SortedBlockList::Sort(void) + { + assert(m_uiAddedBlocks == m_uiImageBlocks); + InitBuckets(); + + // find max block error + m_fMaxError = -1.0f; + + for (unsigned int uiLink = 0; uiLink < m_uiAddedBlocks; uiLink++) + { + Link *plinkBlock = &m_palinkPool[uiLink]; + + float fBlockError = plinkBlock->GetBlock()->GetError(); + if (fBlockError > m_fMaxError) + { + m_fMaxError = fBlockError; + } + } + // prevent divide by zero or divide by negative + if (m_fMaxError <= 0.0f) + { + m_fMaxError = 1.0f; + } + //used for debugging + //int numDone = 0; + // put all of the blocks with unfinished encodings into the appropriate bucket + m_uiSortedBlocks = 0; + for (unsigned int uiLink = 0; uiLink < m_uiAddedBlocks; uiLink++) + { + Link *plinkBlock = &m_palinkPool[uiLink]; + + // if the encoding is done, don't add it to the list + if (plinkBlock->GetBlock()->GetEncoding()->IsDone()) + { + //numDone++; + continue; + } + + // calculate the appropriate sort bucket + float fBlockError = plinkBlock->GetBlock()->GetError(); + int iBucket = (int) floorf(m_iBuckets * fBlockError / m_fMaxError); + // clamp to bucket index + iBucket = iBucket < 0 ? 0 : iBucket >= m_iBuckets ? m_iBuckets - 1 : iBucket; + + // add block to bucket + { + Bucket *pbucket = &m_pabucket[iBucket]; + if (pbucket->plinkLast) + { + pbucket->plinkLast->SetNext(plinkBlock); + pbucket->plinkLast = plinkBlock; + } + else + { + pbucket->plinkFirst = pbucket->plinkLast = plinkBlock; + } + plinkBlock->SetNext(nullptr); + } + + m_uiSortedBlocks++; + + if (0) + { + printf("%u: e=%.3f\n", uiLink, fBlockError); + Print(); + printf("\n\n\n"); + } + } + //printf("num blocks already done: %d\n",numDone); + //link the blocks together across buckets + m_plinkFirst = nullptr; + m_plinkLast = nullptr; + for (int iBucket = m_iBuckets - 1; iBucket >= 0; iBucket--) + { + Bucket *pbucket = &m_pabucket[iBucket]; + + if (pbucket->plinkFirst) + { + if (m_plinkFirst == nullptr) + { + m_plinkFirst = pbucket->plinkFirst; + } + else + { + assert(pbucket->plinkLast->GetNext() == nullptr); + m_plinkLast->SetNext(pbucket->plinkFirst); + } + + m_plinkLast = pbucket->plinkLast; + } + } + + + } + + // ---------------------------------------------------------------------------------------------------- + // clear all of the buckets. normally done in preparation for a sort + // + void SortedBlockList::InitBuckets(void) + { + for (int iBucket = 0; iBucket < m_iBuckets; iBucket++) + { + Bucket *pbucket = &m_pabucket[iBucket]; + + pbucket->plinkFirst = 0; + pbucket->plinkLast = 0; + } + } + + // ---------------------------------------------------------------------------------------------------- + // print out the list of sorted 4x4 blocks + // normally used for debugging + // + void SortedBlockList::Print(void) + { + for (int iBucket = m_iBuckets-1; iBucket >= 0; iBucket--) + { + Bucket *pbucket = &m_pabucket[iBucket]; + + unsigned int uiBlocks = 0; + for (Link *plink = pbucket->plinkFirst; plink != nullptr; plink = plink->GetNext() ) + { + uiBlocks++; + + if (plink == pbucket->plinkLast) + { + break; + } + } + + float fBucketError = m_fMaxError * iBucket / m_iBuckets; + float fBucketRMS = sqrtf(fBucketError / (4.0f*16.0f) ); + printf("%3d: e=%.3f rms=%.6f %u\n", iBucket, fBucketError, fBucketRMS, uiBlocks); + } + } + + // ---------------------------------------------------------------------------------------------------- + // + +} // namespace Etc diff --git a/EtcLib/EtcCodec/EtcSortedBlockList.h b/EtcLib/EtcCodec/EtcSortedBlockList.h new file mode 100644 index 0000000..8ebb978 --- /dev/null +++ b/EtcLib/EtcCodec/EtcSortedBlockList.h @@ -0,0 +1,124 @@ +/* + * Copyright 2015 The Etc2Comp Authors. + * + * Licensed under the Apache License, Version 2.0 (the "License"); + * you may not use this file except in compliance with the License. + * You may obtain a copy of the License at + * + * http://www.apache.org/licenses/LICENSE-2.0 + * + * Unless required by applicable law or agreed to in writing, software + * distributed under the License is distributed on an "AS IS" BASIS, + * WITHOUT WARRANTIES OR CONDITIONS OF ANY KIND, either express or implied. + * See the License for the specific language governing permissions and + * limitations under the License. + */ + +#pragma once + +namespace Etc +{ + class Block4x4; + + class SortedBlockList + { + public: + + class Link + { + public: + + inline void Init(Block4x4 *a_pblock) + { + m_pblock = a_pblock; + m_plinkNext = nullptr; + } + + inline Block4x4 * GetBlock(void) + { + return m_pblock; + } + + inline void SetNext(Link *a_plinkNext) + { + m_plinkNext = a_plinkNext; + } + + inline Link * GetNext(void) + { + return m_plinkNext; + } + + inline Link * Advance(unsigned int a_uiSteps = 1) + { + Link *plink = this; + + for (unsigned int uiStep = 0; uiStep < a_uiSteps; uiStep++) + { + if (plink == nullptr) + { + break; + } + + plink = plink->m_plinkNext; + } + + return plink; + } + + private: + + Block4x4 *m_pblock; + Link *m_plinkNext; + }; + + SortedBlockList(unsigned int a_uiImageBlocks, unsigned int a_uiBuckets); + ~SortedBlockList(void); + + void AddBlock(Block4x4 *a_pblock); + + void Sort(void); + + inline Link * GetLinkToFirstBlock(void) + { + return m_plinkFirst; + } + + inline unsigned int GetNumberOfAddedBlocks(void) + { + return m_uiAddedBlocks; + } + + inline unsigned int GetNumberOfSortedBlocks(void) + { + return m_uiSortedBlocks; + } + + void Print(void); + + private: + + void InitBuckets(void); + + class Bucket + { + public: + Link *plinkFirst; + Link *plinkLast; + }; + + unsigned int m_uiImageBlocks; + int m_iBuckets; + + unsigned int m_uiAddedBlocks; + unsigned int m_uiSortedBlocks; + Link *m_palinkPool; + Bucket *m_pabucket; + float m_fMaxError; + + Link *m_plinkFirst; + Link *m_plinkLast; + + }; + +} // namespace Etc diff --git a/EtcTool/Args.txt b/EtcTool/Args.txt new file mode 100644 index 0000000..d69522b --- /dev/null +++ b/EtcTool/Args.txt @@ -0,0 +1,7 @@ +C:\Users\BSI\Desktop\etc2comp\googleTest.png +-format RGB8 +-errormetric rgba +-output ../../EncodedImages/googleTest.ktx +-analyze ../../Analysis/googleTest +-verbose +-effort 0 diff --git a/EtcTool/CMakeLists.txt b/EtcTool/CMakeLists.txt new file mode 100644 index 0000000..ef79cb0 --- /dev/null +++ b/EtcTool/CMakeLists.txt @@ -0,0 +1,28 @@ +# Copyright 2015 The Etc2Comp Authors. +# +# Licensed under the Apache License, Version 2.0 (the "License"); +# you may not use this file except in compliance with the License. +# You may obtain a copy of the License at +# +# http://www.apache.org/licenses/LICENSE-2.0 +# +# Unless required by applicable law or agreed to in writing, software +# distributed under the License is distributed on an "AS IS" BASIS, +# WITHOUT WARRANTIES OR CONDITIONS OF ANY KIND, either express or implied. +# See the License for the specific language governing permissions and +# limitations under the License. + +project(EtcTool) +include_directories(../EtcLib/Etc) +include_directories(../EtcLib/EtcCodec) +include_directories(../third_party/lodepng) + +file(GLOB SOURCES + ${PROJECT_SOURCE_DIR}/*.h + ${PROJECT_SOURCE_DIR}/*.cpp + ../third_party/lodepng/*.h + ../third_party/lodepng/*.cpp) +add_executable(EtcTool ${SOURCES}) + +target_link_libraries (EtcTool EtcLib) + diff --git a/EtcTool/EtcAnalysis.cpp b/EtcTool/EtcAnalysis.cpp new file mode 100644 index 0000000..4181e07 --- /dev/null +++ b/EtcTool/EtcAnalysis.cpp @@ -0,0 +1,410 @@ +/* + * Copyright 2015 The Etc2Comp Authors. + * + * Licensed under the Apache License, Version 2.0 (the "License"); + * you may not use this file except in compliance with the License. + * You may obtain a copy of the License at + * + * http://www.apache.org/licenses/LICENSE-2.0 + * + * Unless required by applicable law or agreed to in writing, software + * distributed under the License is distributed on an "AS IS" BASIS, + * WITHOUT WARRANTIES OR CONDITIONS OF ANY KIND, either express or implied. + * See the License for the specific language governing permissions and + * limitations under the License. + */ + +#ifdef _WIN32 +#define _CRT_SECURE_NO_WARNINGS (1) +#endif + +#include "EtcAnalysis.h" + +#include "EtcTool.h" +#include "EtcComparison.h" +#include "Etc.h" +#include "EtcFile.h" +#include "EtcMath.h" +#include "EtcImage.h" +#include "EtcBlock4x4.h" + +#include "lodepng.h" +#include +#include //sqrt fn() + +namespace Etc +{ + + // ---------------------------------------------------------------------------------------------------- + // + Analysis::Analysis(Image *a_pimage, const char *a_pstrOutputFolder) + { + + m_pimage = a_pimage; + m_pstrOutputFolder = a_pstrOutputFolder; + m_uiComparisons = 0; + + CreateNewDir(a_pstrOutputFolder); + + // write log file + { + char strFilename[200]; + sprintf(strFilename, "%s%cAnalysis.txt", m_pstrOutputFolder, ETC_PATH_SLASH); + + FILE *pfileTxt = fopen(strFilename, "wt"); + if (pfileTxt == nullptr) + { + printf("Error: couldn't create analysis log file (%s)\n", strFilename); + exit(1); + } + + float fImageError = m_pimage->GetError(); + unsigned int uiImagePixels = m_pimage->GetSourceWidth() * m_pimage->GetSourceHeight(); + + // output stats to both stdout and the analysis file + int numOutputs = 1; + FILE *apfile[2]; + apfile[0] = pfileTxt; + + if (m_pimage->m_bVerboseOutput) + { + apfile[1] = stdout; + numOutputs++; + } + + for (int i = 0; i < numOutputs; i++) + { + + if (a_pimage->GetFormat() == Image::Format::R11 || a_pimage->GetFormat() == Image::Format::SIGNED_R11) + { + fprintf(apfile[i], "PSNR(r) = %.4f\n", ConvertErrorToPSNR(fImageError, 1 * uiImagePixels)); + } + else if (a_pimage->GetFormat() == Image::Format::RG11 || a_pimage->GetFormat() == Image::Format::SIGNED_RG11) + { + fprintf(apfile[i], "PSNR(rg) = %.4f\n", ConvertErrorToPSNR(fImageError, 2 * uiImagePixels)); + } + else + { + int iComponents=3; + if (a_pimage->GetErrorMetric() == ErrorMetric::REC709) + { + iComponents = (int)Block4x4Encoding::LUMA_WEIGHT + 2; + } + fprintf(apfile[i], "PSNR(rgb) = %.4f\n", ConvertErrorToPSNR(fImageError, iComponents * uiImagePixels)); + fprintf(apfile[i], "PSNR(rgba) = %.4f\n", ConvertErrorToPSNR(fImageError, (iComponents+1) * uiImagePixels)); + } + + fprintf(apfile[i], "EncodeTime = %.3f seconds\n", (float)m_pimage->GetEncodingTimeMs() / 1000.0f); + } + + + fclose(pfileTxt); + } + + // scale == 1 + DrawImage(m_pimage, m_pstrOutputFolder, false); + + // scale == 2, with modes + DrawImage(m_pimage, m_pstrOutputFolder, true); + + } + + // ---------------------------------------------------------------------------------------------------- + // + void Analysis::Compare(const char *a_pstrFilename, int a_iPixelX, int a_iPixelY) + { + if (m_uiComparisons >= MAX_COMPARISONS) + { + printf("Error: too many comparisons\n"); + exit(1); + } + + m_apcomparison[m_uiComparisons] = new Comparison(this, m_uiComparisons, a_pstrFilename, a_iPixelX, a_iPixelY); + + m_uiComparisons++; + + } + + // ---------------------------------------------------------------------------------------------------- + // draw 2x image + // optionally display encoding modes + void Analysis::DrawImage(Image *a_pimage, const char *a_pstrOutputFolder, bool a_boolDrawModes) + { + unsigned int uiPngWidth = a_pimage->GetExtendedWidth(); + unsigned int uiPngHeight = a_pimage->GetExtendedHeight(); + if (a_boolDrawModes) + { + uiPngWidth *= 2; + uiPngHeight *= 2; + } + + unsigned char* paucPngPixels = new unsigned char[uiPngWidth * uiPngHeight * 4]; + assert(paucPngPixels); + + ColorR8G8B8A8 *pargba8PngPixels = (ColorR8G8B8A8 *)paucPngPixels; + + for (unsigned int uiBlock = 0; uiBlock < a_pimage->GetNumberOfBlocks(); uiBlock++) + { + Block4x4 *pblock = &a_pimage->GetBlocks()[uiBlock]; + if (a_boolDrawModes) + { + if (a_pimage->GetFormat() == Image::Format::R11 || a_pimage->GetFormat() == Image::Format::SIGNED_R11) + { + DrawBlockPixels(pblock, pargba8PngPixels, uiPngWidth, true, 2); + } + else + { + DrawBlockPixels(pblock, pargba8PngPixels, uiPngWidth, false, 2); + } + DrawBlockMode2x(pblock, pargba8PngPixels, uiPngWidth); + } + else + { + if (a_pimage->GetFormat() == Image::Format::R11 || a_pimage->GetFormat() == Image::Format::SIGNED_R11) + { + DrawBlockPixels(pblock, pargba8PngPixels, uiPngWidth, true); + } + else + { + DrawBlockPixels(pblock, pargba8PngPixels, uiPngWidth, false); + } + } + } + + char strFilename[200]; + if (a_boolDrawModes) + { + sprintf(strFilename, "%s%cModes.png", a_pstrOutputFolder, ETC_PATH_SLASH); + } + else + { + sprintf(strFilename, "%s%cDecoded.png", a_pstrOutputFolder, ETC_PATH_SLASH); + } + + unsigned iResult = lodepng_encode32_file(strFilename, paucPngPixels, uiPngWidth, uiPngHeight); + + if (iResult != 0) + { + if (a_boolDrawModes) + { + printf("Error couldn't write modes image (%s)\n", strFilename); + } + else + { + printf("Error couldn't write decoded image (%s)\n", strFilename); + } + + exit(1); + } + + } + + // ---------------------------------------------------------------------------------------------------- + // + void Analysis::DrawBlockPixels(Block4x4 *a_pblock, + ColorR8G8B8A8 *a_pargba8Output, + unsigned int a_uiOutputWidth, + bool a_bGrayscale, + unsigned int a_uiScale) + { + + // output pixel coord of upper left corner of block + unsigned int uiBlockH = a_uiScale * a_pblock->GetSourceH(); + unsigned int uiBlockV = a_uiScale * a_pblock->GetSourceV(); + + ColorR8G8B8A8 *pargba8Block = &a_pargba8Output[uiBlockV*a_uiOutputWidth + uiBlockH]; + ColorFloatRGBA *pafrgbaDecodedColor = a_pblock->GetDecodedColors(); + float *pafDecodedAlpha = a_pblock->GetDecodedAlphas(); + + for (unsigned int uiPixel = 0; uiPixel < Block4x4::PIXELS; uiPixel++) + { + ColorFloatRGBA *pfrgba = &pafrgbaDecodedColor[uiPixel]; + int iR; + int iG; + int iB; + if (a_bGrayscale) + { + iR = pfrgba->IntRed(255.0f); + iG = pfrgba->IntRed(255.0f); + iB = pfrgba->IntRed(255.0f); + } + else + { + iR = pfrgba->IntRed(255.0f); + iG = pfrgba->IntGreen(255.0f); + iB = pfrgba->IntBlue(255.0f); + } + + int iA = (int) roundf((255.0f*pafDecodedAlpha[uiPixel])); + + ColorR8G8B8A8 *pargba8ScaledPixel = &pargba8Block[(a_uiScale * (uiPixel % 4))*a_uiOutputWidth + + a_uiScale * (uiPixel / 4)]; + + // draw scaled pixel + for (unsigned int uiV = 0; uiV < a_uiScale; uiV++) + { + for (unsigned int uiH = 0; uiH < a_uiScale; uiH++) + { + ColorR8G8B8A8 *prgba8 = &pargba8ScaledPixel[uiV*a_uiOutputWidth + uiH]; + + prgba8->ucR = (unsigned char)iR; + prgba8->ucG = (unsigned char)iG; + prgba8->ucB = (unsigned char)iB; + prgba8->ucA = (unsigned char)iA; + } + } + } + + } + + // ---------------------------------------------------------------------------------------------------- + // + void Analysis::DrawBlockMode2x(Block4x4 *a_pblock, + ColorR8G8B8A8 *a_pargba8Output, + unsigned int a_uiOutputWidth) + { + static const unsigned int SCALE = 2; + + typedef struct + { + int iH; + int iV; + } PixelCoord; + + static const PixelCoord s_apixelcoordOutline[] = { + { 0,0 },{ 1,0 },{ 2,0 },{ 3,0 },{ 4,0 },{ 5,0 },{ 6,0 },{ 7,0 }, + { 0,1 },{ 0,2 },{ 0,3 },{ 0,4 },{ 0,5 },{ 0,6 },{ 0,7 },{ -1,-1 } + }; + + static const PixelCoord s_apixelcoordFlip0[] = + { + { 4,2 },{ 4,4 },{ 4,6 },{ -1,-1 } + }; + + static const PixelCoord s_apixelcoordFlip1[] = + { + { 2,4 },{ 4,4 },{ 6,4 },{ -1,-1 } + }; + + static const PixelCoord s_apixelcoordBentDiffFlip0[] = + { + { 4,2 },{ 4,3 },{ 4,5 },{ 4,6 },{ -1,-1 } + }; + + static const PixelCoord s_apixelcoordBentDiffFlip1[] = + { + { 2,4 },{ 3,4 },{ 5,4 },{ 6,4 },{ -1,-1 } + }; + + static const PixelCoord s_apixelcoordIndividual[] = + { + { 1,1 },{ 1,7 },{ 7,1 },{ 7,7 },{ -1,-1 } + }; + + static const PixelCoord s_apixelcoordT[] = + { + { 3,2 },{ 5,2 }, + { 4,2 },{ 4,3 },{ 4,4 },{ 4,5 },{ 4,6 },{ -1,-1 } + }; + + static const PixelCoord s_apixelcoordH[] = + { + { 4,4 }, + { 3,2 },{ 3,3 },{ 3,4 },{ 3,5 },{ 3,6 }, + { 5,2 },{ 5,3 },{ 5,4 },{ 5,5 },{ 5,6 },{ -1,-1 } + }; + + static const PixelCoord s_apixelcoordPlanar[] = + { + { 4,2 },{ 5,2 },{ 5,3 },{ 5,4 },{ 4,4 }, + { 3,2 },{ 3,3 },{ 3,4 },{ 3,5 },{ 3,6 },{ -1,-1 } + }; + + static const PixelCoord s_apixelcoordUnknown[] = + { + { 3,2 },{ 4,2 },{ 5,2 },{ 5,3 },{ 5,4 },{ 4,4 },{ 4,6 },{ -1,-1 } + }; + + // output pixel coord of upper left corner of block + unsigned int uiBlockH = SCALE * a_pblock->GetSourceH(); + unsigned int uiBlockV = SCALE * a_pblock->GetSourceV(); + + ColorR8G8B8A8 *pargba8Block = &a_pargba8Output[uiBlockV*a_uiOutputWidth + uiBlockH]; + ColorR8G8B8A8 rgba8Gray; + rgba8Gray.ucR = 128; + rgba8Gray.ucG = 128; + rgba8Gray.ucB = 128; + rgba8Gray.ucA = 255; + + // outline + for (const PixelCoord *pcoord = s_apixelcoordOutline; pcoord->iH >= 0; pcoord++) + { + ColorR8G8B8A8 *prgba8 = &pargba8Block[pcoord->iV*a_uiOutputWidth + pcoord->iH]; + *prgba8 = rgba8Gray; + } + + const PixelCoord *pacoordMode = nullptr; + + switch (a_pblock->GetEncodingMode()) + { + case Block4x4Encoding::MODE_ETC1: + + // H/V split + pacoordMode = a_pblock->GetFlip() ? s_apixelcoordFlip1 : s_apixelcoordFlip0; + + // individial + if (a_pblock->IsDifferential() == false) + { + for (const PixelCoord *pcoord = s_apixelcoordIndividual; pcoord->iH >= 0; pcoord++) + { + ColorR8G8B8A8 *prgba8 = &pargba8Block[pcoord->iV*a_uiOutputWidth + pcoord->iH]; + *prgba8 = rgba8Gray; + } + } + else if (a_pblock->GetEncoding()->HasSeverelyBentDifferentialColors()) + { + pacoordMode = a_pblock->GetFlip() ? s_apixelcoordBentDiffFlip1 : s_apixelcoordBentDiffFlip0; + } + break; + + case Block4x4Encoding::MODE_T: + pacoordMode = s_apixelcoordT; + break; + + case Block4x4Encoding::MODE_H: + pacoordMode = s_apixelcoordH; + break; + + case Block4x4Encoding::MODE_PLANAR: + pacoordMode = s_apixelcoordPlanar; + break; + + default: + pacoordMode = s_apixelcoordUnknown; + break; + } + + // draw mode + for (const PixelCoord *pcoord = pacoordMode; pcoord->iH >= 0; pcoord++) + { + ColorR8G8B8A8 *prgba8 = &pargba8Block[pcoord->iV*a_uiOutputWidth + pcoord->iH]; + *prgba8 = rgba8Gray; + } + + } + + // ---------------------------------------------------------------------------------------------------- + // + float Analysis::ConvertErrorToPSNR(float a_fError, unsigned int a_uiTotalComponents) + { + + float fMSE = a_fError / (float)a_uiTotalComponents; + float fPSNR = ConvertMSEToPSNR(fMSE); + + return fPSNR; + } + + // ---------------------------------------------------------------------------------------------------- + // + +} diff --git a/EtcTool/EtcAnalysis.h b/EtcTool/EtcAnalysis.h new file mode 100644 index 0000000..b814fd9 --- /dev/null +++ b/EtcTool/EtcAnalysis.h @@ -0,0 +1,71 @@ +/* + * Copyright 2015 The Etc2Comp Authors. + * + * Licensed under the Apache License, Version 2.0 (the "License"); + * you may not use this file except in compliance with the License. + * You may obtain a copy of the License at + * + * http://www.apache.org/licenses/LICENSE-2.0 + * + * Unless required by applicable law or agreed to in writing, software + * distributed under the License is distributed on an "AS IS" BASIS, + * WITHOUT WARRANTIES OR CONDITIONS OF ANY KIND, either express or implied. + * See the License for the specific language governing permissions and + * limitations under the License. + */ + +#pragma once + +#include "EtcColor.h" + +namespace Etc +{ + + class Comparison; + class Image; + class Block4x4; + + class Analysis + { + public: + + static const unsigned int MAX_COMPARISONS = 4; + + Analysis(Image *a_pimage, const char *a_pstrOutputFolder); + + void Compare(const char *a_pstrFilename, int a_iPixelX = -1, int a_iPixelY = -1); + + inline const char * GetOutputFolder(void) + { + return m_pstrOutputFolder; + } + + static void DrawImage(Image *a_pimage, const char *a_pstrOutputFolder, + bool a_boolDrawModes); + + inline Image * GetImage(void) + { + return m_pimage; + } + + static void DrawBlockPixels(Block4x4 *a_pblock, + ColorR8G8B8A8 *a_pargba8Output, + unsigned int a_uiOutputWidth, + bool a_bGrayscale, + unsigned int a_uiScale = 1); + + static float ConvertErrorToPSNR(float a_fError, unsigned int a_uiTotalComponents); + + private: + + static void DrawBlockMode2x(Block4x4 *a_pblock, + ColorR8G8B8A8 *a_pargba8Output, + unsigned int a_uiOutputWidth); + + Image *m_pimage; + const char *m_pstrOutputFolder; + unsigned int m_uiComparisons; + Comparison *m_apcomparison[MAX_COMPARISONS]; + }; + +} diff --git a/EtcTool/EtcComparison.cpp b/EtcTool/EtcComparison.cpp new file mode 100644 index 0000000..b94ccb5 --- /dev/null +++ b/EtcTool/EtcComparison.cpp @@ -0,0 +1,637 @@ +/* + * Copyright 2015 The Etc2Comp Authors. + * + * Licensed under the Apache License, Version 2.0 (the "License"); + * you may not use this file except in compliance with the License. + * You may obtain a copy of the License at + * + * http://www.apache.org/licenses/LICENSE-2.0 + * + * Unless required by applicable law or agreed to in writing, software + * distributed under the License is distributed on an "AS IS" BASIS, + * WITHOUT WARRANTIES OR CONDITIONS OF ANY KIND, either express or implied. + * See the License for the specific language governing permissions and + * limitations under the License. + */ + +#ifdef _WIN32 +#define _CRT_SECURE_NO_WARNINGS (1) +#endif + +#include "EtcConfig.h" + +#include "EtcComparison.h" + +#include "EtcAnalysis.h" +#include "EtcTool.h" +#include "Etc.h" +#include "EtcFile.h" +#include "EtcMath.h" +#include "EtcImage.h" +#include "EtcBlock4x4.h" +#include "EtcBlock4x4Encoding_ETC1.h" +#include "EtcBlock4x4Encoding_RGB8.h" +#include "EtcBlock4x4Encoding_R11.h" +#include "EtcBlock4x4Encoding_RG11.h" + +#include "lodepng.h" + +#include +#include +#include +#include + +namespace Etc +{ + const float Comparison::ERROR_EPSILON = 0.000001f; + + // ---------------------------------------------------------------------------------------------------- + // + Comparison::Comparison(Analysis *a_panalysisParent, unsigned int a_uiIndex, const char *a_pstrFilename, int a_iPixelX, int a_iPixelY) + { + + m_panalysisParent = a_panalysisParent; + m_uiIndex = a_uiIndex; + m_pstrFilename = new char[strlen(a_pstrFilename)+1]; + strcpy(m_pstrFilename, a_pstrFilename); + + SetName(); + + m_pstrOutputFolder = new char[256]; + sprintf(m_pstrOutputFolder, "%s%cComparison_%s", + m_panalysisParent->GetOutputFolder(), ETC_PATH_SLASH, m_pstrName); + + CreateNewDir(m_pstrOutputFolder); + + // read etc file + Etc::File etcfile(a_pstrFilename, Etc::File::Format::INFER_FROM_FILE_EXTENSION); + + etcfile.UseSingleBlock(a_iPixelX, a_iPixelY); + + // construct image with encoding bits + m_pimage = new Image(etcfile.GetImageFormat(), + etcfile.GetSourceWidth(), + etcfile.GetSourceHeight(), + etcfile.GetEncodingBits(), + etcfile.GetEncodingBitsBytes(), + a_panalysisParent->GetImage(), + a_panalysisParent->GetImage()->GetErrorMetric()); + + if (m_pimage->GetExtendedWidth() != a_panalysisParent->GetImage()->GetExtendedWidth() || + m_pimage->GetExtendedHeight() != a_panalysisParent->GetImage()->GetExtendedHeight()) + { + printf("Error: comparison image (%s) has different width or height\n", m_pstrFilename); + exit(1); + } + + // scale = 1 + Analysis::DrawImage(m_pimage, m_pstrOutputFolder, false); + + // scale = 2, with modes + Analysis::DrawImage(m_pimage, m_pstrOutputFolder, true); + + DrawImageComparison(m_panalysisParent->GetImage()); + + WriteLogFile(); + + } + + // ---------------------------------------------------------------------------------------------------- + // + void Comparison::DrawImageComparison(Image *a_pimageUnderTest) + { + + assert(m_pimage->GetExtendedWidth() == a_pimageUnderTest->GetExtendedWidth() && + m_pimage->GetExtendedHeight() == a_pimageUnderTest->GetExtendedHeight()); + + unsigned int uiPngWidth = 2 * m_pimage->GetExtendedWidth(); + unsigned int uiPngHeight = 2 * m_pimage->GetExtendedHeight(); + + unsigned char* paucPngPixels = new unsigned char[uiPngWidth * uiPngHeight * 4]; + assert(paucPngPixels); + + ColorR8G8B8A8 *pargba8PngPixels = (ColorR8G8B8A8 *)paucPngPixels; + + for (unsigned int uiBlock = 0; uiBlock < m_pimage->GetNumberOfBlocks(); uiBlock++) + { + Block4x4 *pblock = &m_pimage->GetBlocks()[uiBlock]; + Block4x4 *pblockUnderTest = &a_pimageUnderTest->GetBlocks()[uiBlock]; + + if (a_pimageUnderTest->GetFormat() == Image::Format::R11 || a_pimageUnderTest->GetFormat() == Image::Format::SIGNED_R11) + { + Analysis::DrawBlockPixels(pblock, pargba8PngPixels, uiPngWidth, true, 2); + } + else + { + Analysis::DrawBlockPixels(pblock, pargba8PngPixels, uiPngWidth, false, 2); + } + + float fErrorBlockReference = pblock->GetError(); + float fErrorBlockUnderTest = pblockUnderTest->GetError(); + + DrawBlockComparison2x(pblockUnderTest, pargba8PngPixels, uiPngWidth, + fErrorBlockUnderTest, fErrorBlockReference); + } + + char strFilename[200]; + sprintf(strFilename, "%s%cComparison.png", m_pstrOutputFolder, ETC_PATH_SLASH); + + unsigned iResult = lodepng_encode32_file(strFilename, paucPngPixels, uiPngWidth, uiPngHeight); + + if (iResult != 0) + { + printf("Error couldn't write modes image (%s)\n", strFilename); + } + + } + + // ---------------------------------------------------------------------------------------------------- + // + void Comparison::DrawBlockComparison2x(Block4x4 *a_pblockUnderTest, + ColorR8G8B8A8 *a_pargba8Output, + unsigned int a_uiOutputWidth, + float a_fErrorBlockUnderTest, + float a_fErrorBlockReference) + { + static const unsigned int SCALE = 2; + + typedef struct + { + int iH; + int iV; + } PixelCoord; + + static const PixelCoord s_apixelcoordDifference0[] = { + { 0,0 },{ 1,0 },{ 2,0 },{ 3,0 },{ 4,0 },{ 5,0 },{ 6,0 },{ 7,0 }, + { 0,1 },{ 0,2 },{ 0,3 },{ 0,4 },{ 0,5 },{ 0,6 },{ 0,7 }, + { 1,7 },{ 2,7 },{ 3,7 },{ 4,7 },{ 5,7 },{ 6,7 },{ 7,7 }, + { 7,1 },{ 7,2 },{ 7,3 },{ 7,4 },{ 7,5 },{ 7,6 },{ -1,-1 } + }; + + static const PixelCoord s_apixelcoordDifference5[] = { + { 3,3 }, + { 0,0 },{ 1,0 },{ 2,0 },{ 3,0 },{ 4,0 },{ 5,0 },{ 6,0 },{ 7,0 }, + { 0,1 },{ 0,2 },{ 0,3 },{ 0,4 },{ 0,5 },{ 0,6 },{ 0,7 }, + { 1,7 },{ 2,7 },{ 3,7 },{ 4,7 },{ 5,7 },{ 6,7 },{ 7,7 }, + { 7,1 },{ 7,2 },{ 7,3 },{ 7,4 },{ 7,5 },{ 7,6 },{ -1,-1 } + }; + + static const PixelCoord s_apixelcoordDifference10[] = { + { 3,3 },{ 3,4 },{ 4,3 },{ 4,4 }, + { 0,0 },{ 1,0 },{ 2,0 },{ 3,0 },{ 4,0 },{ 5,0 },{ 6,0 },{ 7,0 }, + { 0,1 },{ 0,2 },{ 0,3 },{ 0,4 },{ 0,5 },{ 0,6 },{ 0,7 }, + { 1,7 },{ 2,7 },{ 3,7 },{ 4,7 },{ 5,7 },{ 6,7 },{ 7,7 }, + { 7,1 },{ 7,2 },{ 7,3 },{ 7,4 },{ 7,5 },{ 7,6 },{ -1,-1 } + }; + + static const PixelCoord s_apixelcoordDifference20[] = { + { 2,2 },{ 2,3 },{ 3,2 },{ 3,3 },{ 4,4 },{ 4,5 },{ 5,4 },{ 5,5 }, + { 0,0 },{ 1,0 },{ 2,0 },{ 3,0 },{ 4,0 },{ 5,0 },{ 6,0 },{ 7,0 }, + { 0,1 },{ 0,2 },{ 0,3 },{ 0,4 },{ 0,5 },{ 0,6 },{ 0,7 }, + { 1,7 },{ 2,7 },{ 3,7 },{ 4,7 },{ 5,7 },{ 6,7 },{ 7,7 }, + { 7,1 },{ 7,2 },{ 7,3 },{ 7,4 },{ 7,5 },{ 7,6 },{ -1,-1 } + }; + + static const PixelCoord s_apixelcoordCorners[] = { + { 0,0 },{ 7,0 },{ 0,7 },{ 7,7 },{ -1,-1 } + }; + + // output pixel coord of upper left corner of block + unsigned int uiBlockH = SCALE * a_pblockUnderTest->GetSourceH(); + unsigned int uiBlockV = SCALE * a_pblockUnderTest->GetSourceV(); + + ColorR8G8B8A8 *pargba8Block = &a_pargba8Output[uiBlockV*a_uiOutputWidth + uiBlockH]; + + float fRelativeError = 0.0f; + if (a_fErrorBlockUnderTest != a_fErrorBlockReference) + { + fRelativeError = fabs((a_fErrorBlockUnderTest - a_fErrorBlockReference) / + (a_fErrorBlockUnderTest + a_fErrorBlockReference)); + + if (fabsf(a_fErrorBlockUnderTest - a_fErrorBlockReference) < ERROR_EPSILON) + { + fRelativeError = 0.0f; + } + } + + ColorR8G8B8A8 rgb8Draw; + rgb8Draw.ucA = 255; + + // equal + if (fRelativeError == 0.0f) + { + rgb8Draw.ucR = 128; + rgb8Draw.ucG = 128; + rgb8Draw.ucB = 128; + } + // better tthan reference + else if (a_fErrorBlockUnderTest < a_fErrorBlockReference) + { + rgb8Draw.ucR = 0; + rgb8Draw.ucG = 255; + rgb8Draw.ucB = 0; + } + // worse than reference + else if (a_fErrorBlockUnderTest > a_fErrorBlockReference) + { + rgb8Draw.ucR = 255; + rgb8Draw.ucG = 0; + rgb8Draw.ucB = 0; + } + + const PixelCoord *papixelcoordDraw = s_apixelcoordCorners; + + // if 20% worse + if (fRelativeError >= 1.44f) + { + papixelcoordDraw = s_apixelcoordDifference20; + } + // if 10% worse + else if (fRelativeError >= 1.21f) + { + papixelcoordDraw = s_apixelcoordDifference10; + } + // if 5% worse + else if (fRelativeError >= 1.1025f) + { + papixelcoordDraw = s_apixelcoordDifference5; + } + else if (fRelativeError > 0.0f) + { + papixelcoordDraw = s_apixelcoordDifference0; + } + + // outline + for (const PixelCoord *pcoord = papixelcoordDraw; pcoord->iH >= 0; pcoord++) + { + ColorR8G8B8A8 *prgba8 = &pargba8Block[pcoord->iV*a_uiOutputWidth + pcoord->iH]; + + *prgba8 = rgb8Draw; + } + + } + + // ---------------------------------------------------------------------------------------------------- + // + void Comparison::SetName(void) + { + // alloc memory for name + unsigned int uiNameBytes = (unsigned int)strlen(m_pstrFilename) + 1; + if (uiNameBytes < 4) + { + uiNameBytes = 4; + } + m_pstrName = new char[uiNameBytes]; + m_pstrName[0] = 0; + + // first, try to find folder name of comparison image + { + char *pstr = new char[strlen(m_pstrFilename) + 1]; + strcpy(pstr, m_pstrFilename); + + int iLastSlash; + int iPenultimateSlash; + + // find last slash + for (iLastSlash = (int)strlen(pstr) - 1; iLastSlash >= 0; iLastSlash--) + { + if (pstr[iLastSlash] == ETC_PATH_SLASH) + { + break; + } + } + + // find penultimate slash + for (iPenultimateSlash = iLastSlash - 1; iPenultimateSlash >= 0; iPenultimateSlash--) + { + if (pstr[iPenultimateSlash] == ETC_PATH_SLASH) + { + break; + } + } + + if (iLastSlash > 0) + { + pstr[iLastSlash] = 0; + strcpy(m_pstrName, &pstr[iPenultimateSlash] + 1); + assert(strlen(m_pstrName) > 0); + } + + delete[] pstr; + } + + // otherwise use index as name + if (m_pstrName[0] == 0) + { + sprintf(m_pstrName, "%u", m_uiIndex); + } + + } + + // ---------------------------------------------------------------------------------------------------- + // + void Comparison::WriteLogFile(void) + { + + char strPath[200]; + sprintf(strPath, "%s%cComparison.txt", m_pstrOutputFolder, ETC_PATH_SLASH); + + FILE *pfile = fopen(strPath, "wt"); + if (pfile == nullptr) + { + printf("Error couldn't open comparison log file (%s)\n", strPath); + exit(1); + } + + float fImageError = m_panalysisParent->GetImage()->GetError(); + unsigned int uiImagePixels = m_panalysisParent->GetImage()->GetSourceWidth() * + m_panalysisParent->GetImage()->GetSourceHeight(); + + if (m_pimage->GetFormat() == Image::Format::R11 || m_pimage->GetFormat() == Image::Format::SIGNED_R11) + { + fprintf(pfile, "PSNR(r) = %.4f\n", Analysis::ConvertErrorToPSNR(fImageError, 1 * uiImagePixels)); + fprintf(pfile, "\n"); + } + else if (m_pimage->GetFormat() == Image::Format::RG11 || m_pimage->GetFormat() == Image::Format::SIGNED_RG11) + { + fprintf(pfile, "PSNR(rg) = %.4f\n", Analysis::ConvertErrorToPSNR(fImageError, 2 * uiImagePixels)); + fprintf(pfile, "\n"); + } + else + { + fprintf(pfile, "PSNR(rgb) = %.4f\n", Analysis::ConvertErrorToPSNR(fImageError, 3 * uiImagePixels)); + fprintf(pfile, "PSNR(rgba) = %.4f\n", Analysis::ConvertErrorToPSNR(fImageError, 4 * uiImagePixels)); + fprintf(pfile, "\n"); + } + float fReferenceImageError = m_pimage->GetError(); + + fprintf(pfile, "reference image %s\n", m_pstrFilename); + if (m_pimage->GetFormat() == Image::Format::R11 || m_pimage->GetFormat() == Image::Format::SIGNED_R11) + { + fprintf(pfile, "reference PSNR(r) = %.4f\n", + Analysis::ConvertErrorToPSNR(fReferenceImageError, 1 * uiImagePixels)); + } + else if (m_pimage->GetFormat() == Image::Format::RG11 || m_pimage->GetFormat() == Image::Format::SIGNED_RG11) + { + fprintf(pfile, "reference PSNR(rg) = %.4f\n", + Analysis::ConvertErrorToPSNR(fReferenceImageError, 2 * uiImagePixels)); + } + else + { + fprintf(pfile, "reference PSNR(rgb) = %.4f\n", + Analysis::ConvertErrorToPSNR(fReferenceImageError, 3 * uiImagePixels)); + fprintf(pfile, "reference PSNR(rgba) = %.4f\n", + Analysis::ConvertErrorToPSNR(fReferenceImageError, 4 * uiImagePixels)); + } + WriteBetterOrWorseBlocks(pfile, false); + WriteBetterOrWorseBlocks(pfile, true); + + fclose(pfile); + + } + + // ---------------------------------------------------------------------------------------------------- + // + void Comparison::WriteBetterOrWorseBlocks(FILE *a_pfile, bool a_boolWorse) + { + + fprintf(a_pfile, "\n"); + if (a_boolWorse) + { + fprintf(a_pfile, "Worse Blocks\n"); + } + else + { + fprintf(a_pfile, "Better Blocks\n"); + } + + for (unsigned int uiBlock = 0; uiBlock < m_pimage->GetNumberOfBlocks(); uiBlock++) + { + Block4x4 *pblockReference = &m_pimage->GetBlocks()[uiBlock]; + Block4x4 *pblockUnderTest = &m_panalysisParent->GetImage()->GetBlocks()[uiBlock]; + + float fErrorBlockReference = pblockReference->GetError(); + float fErrorBlockUnderTest = pblockUnderTest->GetError(); + + if (fabsf(fErrorBlockUnderTest - fErrorBlockReference) >= ERROR_EPSILON) + { + if ((a_boolWorse && (fErrorBlockUnderTest > fErrorBlockReference)) || + (!a_boolWorse && (fErrorBlockUnderTest < fErrorBlockReference))) + { + fprintf(a_pfile, "HV=%u,%u\n", + pblockUnderTest->GetSourceH(), pblockUnderTest->GetSourceV()); + + WriteBlockInfo(a_pfile, pblockUnderTest, "bsi"); + WriteBlockInfo(a_pfile, pblockReference, "ref"); + + } + } + + } + + } + + // ---------------------------------------------------------------------------------------------------- + // + void Comparison::WriteBlockInfo(FILE *a_pfile, Block4x4 *a_pblock, const char *a_pstrEncoder) + { + + Block4x4Encoding *pencoding = a_pblock->GetEncoding(); + + Block4x4Encoding_R11 *pencoding_r11 = static_cast(pencoding); + Block4x4Encoding_RG11 *pencoding_rg11 = static_cast(pencoding); + Block4x4Encoding_RGB8 *pencoding_rgb8 = static_cast(pencoding); + + fprintf(a_pfile, " %s:", a_pstrEncoder); + + float fPSNR = 0.0f; + if (m_pimage->GetFormat() == Image::Format::R11 || m_pimage->GetFormat() == Image::Format::SIGNED_R11) + { + fPSNR = Analysis::ConvertErrorToPSNR(a_pblock->GetError(), 1 * 16); + } + else if (m_pimage->GetFormat() == Image::Format::RG11 || m_pimage->GetFormat() == Image::Format::SIGNED_RG11) + { + fPSNR = Analysis::ConvertErrorToPSNR(a_pblock->GetError(), 2 * 16); + } + else + { + fPSNR = Analysis::ConvertErrorToPSNR(a_pblock->GetError(), 3 * 16); + } + if (a_pblock->GetEncodingMode() == Block4x4Encoding::MODE_RG11) + { + fprintf(a_pfile, "psnr=%6.3f\n", fPSNR); + } + else + { + fprintf(a_pfile, "psnr=%6.3f ", fPSNR); + } + bool boolWriteSelectors = false; + + if (m_pimage->GetFormat() == Image::Format::RGBA8) + { + fprintf(a_pfile, "RGBA8 "); + fprintf(a_pfile, "alpha base(%4.0f) ", pencoding_r11->GetRedBase()); + fprintf(a_pfile, "alpha multiplier(%3.0f) ", pencoding_r11->GetRedMultiplier()); + fprintf(a_pfile, "alpha table index(%2d) ", pencoding_r11->GetRedTableIndex()); + } + if (a_pblock->GetEncodingMode() == Block4x4Encoding::MODE_ETC1) + { + fprintf(a_pfile, "ETC1(%c,%c) ", + a_pblock->IsDifferential() ? 'D' : 'I', + a_pblock->GetFlip() ? 'V' : 'H' ); + + if (a_pblock->IsDifferential()) + { + fprintf(a_pfile, "color1(%2d,%2d,%2d) ", + pencoding_rgb8->GetColor1().IntRed(31.0f), + pencoding_rgb8->GetColor1().IntGreen(31.0f), + pencoding_rgb8->GetColor1().IntBlue(31.0f) ); + fprintf(a_pfile, "color2(%2d,%2d,%2d) ", + pencoding_rgb8->GetColor2().IntRed(31.0f), + pencoding_rgb8->GetColor2().IntGreen(31.0f), + pencoding_rgb8->GetColor2().IntBlue(31.0f)); + } + else + { + fprintf(a_pfile, "color1(%2d,%2d,%2d) ", + pencoding_rgb8->GetColor1().IntRed(15.0f), + pencoding_rgb8->GetColor1().IntGreen(15.0f), + pencoding_rgb8->GetColor1().IntBlue(15.0f)); + fprintf(a_pfile, "color2(%2d,%2d,%2d) ", + pencoding_rgb8->GetColor2().IntRed(15.0f), + pencoding_rgb8->GetColor2().IntGreen(15.0f), + pencoding_rgb8->GetColor2().IntBlue(15.0f)); + } + + fprintf(a_pfile, "cw1(%d) cw2(%d) ", pencoding_rgb8->GetCW1(), pencoding_rgb8->GetCW2()); + + boolWriteSelectors = true; + } + else if (a_pblock->GetEncodingMode() == Block4x4Encoding::MODE_T || + a_pblock->GetEncodingMode() == Block4x4Encoding::MODE_H) + { + fprintf(a_pfile, "%-9s ", a_pblock->GetEncodingModeName()); + + fprintf(a_pfile, "color1(%2d,%2d,%2d) ", + pencoding_rgb8->GetColor1().IntRed(15.0f), + pencoding_rgb8->GetColor1().IntGreen(15.0f), + pencoding_rgb8->GetColor1().IntBlue(15.0f)); + fprintf(a_pfile, "color2(%2d,%2d,%2d) ", + pencoding_rgb8->GetColor2().IntRed(15.0f), + pencoding_rgb8->GetColor2().IntGreen(15.0f), + pencoding_rgb8->GetColor2().IntBlue(15.0f)); + + fprintf(a_pfile, "cw1(%d) ", pencoding_rgb8->GetCW1() ); + + boolWriteSelectors = true; + } + else if (a_pblock->GetEncodingMode() == Block4x4Encoding::MODE_PLANAR) + { + fprintf(a_pfile, "%-9s ", a_pblock->GetEncodingModeName()); + + fprintf(a_pfile, "color1(%2d,%2d,%2d) ", + pencoding_rgb8->GetColor1().IntRed(63.0f), + pencoding_rgb8->GetColor1().IntGreen(127.0f), + pencoding_rgb8->GetColor1().IntBlue(63.0f)); + fprintf(a_pfile, "color2(%2d,%2d,%2d) ", + pencoding_rgb8->GetColor2().IntRed(63.0f), + pencoding_rgb8->GetColor2().IntGreen(127.0f), + pencoding_rgb8->GetColor2().IntBlue(63.0f)); + fprintf(a_pfile, "color3(%2d,%2d,%2d) ", + pencoding_rgb8->GetColor3().IntRed(63.0f), + pencoding_rgb8->GetColor3().IntGreen(127.0f), + pencoding_rgb8->GetColor3().IntBlue(63.0f)); + } + else if (a_pblock->GetEncodingMode() == Block4x4Encoding::MODE_R11) + { + fprintf(a_pfile, "R11 "); + fprintf(a_pfile, "base(%4.0f) ", pencoding_r11->GetRedBase()); + fprintf(a_pfile, "multiplier(%3.0f) ", pencoding_r11->GetRedMultiplier()); + fprintf(a_pfile, "table index(%2d) ", pencoding_r11->GetRedTableIndex()); + + boolWriteSelectors = true; + } + else if (a_pblock->GetEncodingMode() == Block4x4Encoding::MODE_RG11) + { + fprintf(a_pfile, "RG11 image: %s\n", a_pstrEncoder); + fprintf(a_pfile, "red base(%4.0f)\n", pencoding_rg11->GetRedBase()); + fprintf(a_pfile, "grn base(%4.0f)\n\n", pencoding_rg11->GetGrnBase()); + + fprintf(a_pfile, "red multiplier(%3.0f)\n", pencoding_rg11->GetRedMultiplier()); + fprintf(a_pfile, "grn multiplier(%3.0f)\n\n", pencoding_rg11->GetGrnMultiplier()); + + fprintf(a_pfile, "red table index(%2d)\n", pencoding_rg11->GetRedTableIndex()); + fprintf(a_pfile, "grn table index(%2d)\n\n", pencoding_rg11->GetGrnTableIndex()); + + boolWriteSelectors = true; + } + else + { + assert(0); + } + + if (boolWriteSelectors) + { + size_t selectorStringSize = Block4x4Encoding::PIXELS*4; + char *redSelectors = new char[selectorStringSize]; + char *grnSelectors = new char[selectorStringSize]; + memset(&redSelectors[0], 0, selectorStringSize); + memset(&grnSelectors[0], 0, selectorStringSize); + + if (a_pblock->GetEncodingMode() != Block4x4Encoding::MODE_RG11) + { + fprintf(a_pfile, "selectors("); + } + + for (unsigned int uiPixel = 0; uiPixel < Block4x4Encoding::PIXELS; uiPixel++) + { + if (a_pblock->GetEncodingMode() == Block4x4Encoding::MODE_R11) + { + fprintf(a_pfile, "%u", pencoding_r11->GetRedSelectors()[uiPixel]); + } + else if (a_pblock->GetEncodingMode() == Block4x4Encoding::MODE_RG11) + { + char temp[2]; + sprintf(temp,"%d",pencoding_rg11->GetRedSelectors()[uiPixel]); + strcat(redSelectors,temp); + sprintf(temp, "%d", pencoding_rg11->GetGrnSelectors()[uiPixel]); + strcat(grnSelectors, temp); + + strcat(redSelectors, ",\0"); + strcat(grnSelectors, ",\0"); + + fprintf(a_pfile, "pixel[%d]:(%f,%f,%f)\n", uiPixel, + pencoding->GetDecodedColors()[uiPixel].fR, + pencoding->GetDecodedColors()[uiPixel].fG, + pencoding->GetDecodedColors()[uiPixel].fB); + + } + else + { + fprintf(a_pfile, "%u", pencoding_rgb8->GetSelectors()[uiPixel]); + } + } + if (a_pblock->GetEncodingMode() == Block4x4Encoding::MODE_RG11) + { + redSelectors[selectorStringSize - 1] = '\0'; + grnSelectors[selectorStringSize - 1] = '\0'; + fprintf(a_pfile, "selectors(red: %s)\n", redSelectors); + fprintf(a_pfile, "selectors(grn: %s)\n,", grnSelectors); + delete[] redSelectors; + redSelectors = NULL; + delete[] grnSelectors; + grnSelectors = NULL; + } + else + { + fprintf(a_pfile, ")"); + } + } + + fprintf(a_pfile, "\n"); + + } + + // ---------------------------------------------------------------------------------------------------- + // +} diff --git a/EtcTool/EtcComparison.h b/EtcTool/EtcComparison.h new file mode 100644 index 0000000..52b39f0 --- /dev/null +++ b/EtcTool/EtcComparison.h @@ -0,0 +1,61 @@ +/* + * Copyright 2015 The Etc2Comp Authors. + * + * Licensed under the Apache License, Version 2.0 (the "License"); + * you may not use this file except in compliance with the License. + * You may obtain a copy of the License at + * + * http://www.apache.org/licenses/LICENSE-2.0 + * + * Unless required by applicable law or agreed to in writing, software + * distributed under the License is distributed on an "AS IS" BASIS, + * WITHOUT WARRANTIES OR CONDITIONS OF ANY KIND, either express or implied. + * See the License for the specific language governing permissions and + * limitations under the License. + */ + +#pragma once + +#include "EtcColor.h" +#include + +namespace Etc +{ + class Analysis; + class Image; + class Block4x4; + + class Comparison + { + public: + + Comparison(Analysis *a_panalysisParent, unsigned int a_uiIndex, const char *a_pstrFilename, int a_iPixelX = -1, int a_iPixelY = -1); + + private: + + static const float ERROR_EPSILON; + + void SetName(void); + + void DrawImageComparison(Image *a_pimageUnderTest); + + void DrawBlockComparison2x(Block4x4 *a_pblockUnderTest, + ColorR8G8B8A8 *a_pargba8Output, + unsigned int a_uiOutputWidth, + float a_fErrorBlockUnderTest, + float a_fErrorBlockReference); + + void WriteLogFile(void); + void WriteBetterOrWorseBlocks(FILE *a_pfile, bool a_boolWorse); + void WriteBlockInfo(FILE *a_pfile, Block4x4 *a_pblock, const char *a_pstrEncoder); + + Analysis *m_panalysisParent; + unsigned int m_uiIndex; + char *m_pstrFilename; + Image *m_pimage; + char *m_pstrName; + char *m_pstrOutputFolder; + }; + + +} diff --git a/EtcTool/EtcFile.cpp b/EtcTool/EtcFile.cpp new file mode 100644 index 0000000..f57a0b9 --- /dev/null +++ b/EtcTool/EtcFile.cpp @@ -0,0 +1,311 @@ +/* + * Copyright 2015 The Etc2Comp Authors. + * + * Licensed under the Apache License, Version 2.0 (the "License"); + * you may not use this file except in compliance with the License. + * You may obtain a copy of the License at + * + * http://www.apache.org/licenses/LICENSE-2.0 + * + * Unless required by applicable law or agreed to in writing, software + * distributed under the License is distributed on an "AS IS" BASIS, + * WITHOUT WARRANTIES OR CONDITIONS OF ANY KIND, either express or implied. + * See the License for the specific language governing permissions and + * limitations under the License. + */ + +#ifdef _WIN32 +#define _CRT_SECURE_NO_WARNINGS (1) +#endif + +#include "EtcConfig.h" + + +#include "EtcFile.h" + +#include "EtcFileHeader.h" +#include "EtcColor.h" +#include "Etc.h" +#include "EtcBlock4x4EncodingBits.h" + +#include +#include +#include +#include + +using namespace Etc; + +// ---------------------------------------------------------------------------------------------------- +// +File::File(const char *a_pstrFilename, Format a_fileformat, Image::Format a_imageformat, + unsigned char *a_paucEncodingBits, unsigned int a_uiEncodingBitsBytes, + unsigned int a_uiSourceWidth, unsigned int a_uiSourceHeight, + unsigned int a_uiExtendedWidth, unsigned int a_uiExtendedHeight) +{ + if (a_pstrFilename == nullptr) + { + m_pstrFilename = const_cast(""); + } + else + { + m_pstrFilename = new char[strlen(a_pstrFilename) + 1]; + strcpy(m_pstrFilename, a_pstrFilename); + } + + m_fileformat = a_fileformat; + if (m_fileformat == Format::INFER_FROM_FILE_EXTENSION) + { + // ***** TODO: add this later ***** + m_fileformat = Format::KTX; + } + + m_imageformat = a_imageformat; + + m_paucEncodingBits = a_paucEncodingBits; + m_uiEncodingBitsBytes = a_uiEncodingBitsBytes; + m_uiSourceWidth = a_uiSourceWidth; + m_uiSourceHeight = a_uiSourceHeight; + m_uiExtendedWidth = a_uiExtendedWidth; + m_uiExtendedHeight = a_uiExtendedHeight; + + switch (m_fileformat) + { + case Format::PKM: + m_pheader = new FileHeader_Pkm(this); + break; + + case Format::KTX: + m_pheader = new FileHeader_Ktx(this); + break; + + default: + assert(0); + break; + } + +} + + +// ---------------------------------------------------------------------------------------------------- +// +File::File(const char *a_pstrFilename, Format a_fileformat) +{ + if (a_pstrFilename == nullptr) + { + return; + } + else + { + m_pstrFilename = new char[strlen(a_pstrFilename) + 1]; + strcpy(m_pstrFilename, a_pstrFilename); + } + + m_fileformat = a_fileformat; + if (m_fileformat == Format::INFER_FROM_FILE_EXTENSION) + { + // ***** TODO: add this later ***** + m_fileformat = Format::KTX; + } + + FILE *pfile = fopen(m_pstrFilename, "rb"); + if (pfile == nullptr) + { + printf("ERROR: Couldn't open %s", m_pstrFilename); + exit(1); + } + fseek(pfile, 0, SEEK_END); + unsigned int fileSize = ftell(pfile); + fseek(pfile, 0, SEEK_SET); + size_t szResult; + + m_pheader = new FileHeader_Ktx(this); + szResult = fread( ((FileHeader_Ktx*)m_pheader)->GetData(), 1, sizeof(FileHeader_Ktx::Data), pfile); + assert(szResult > 0); + + if (((FileHeader_Ktx*)m_pheader)->GetData()->m_u32BytesOfKeyValueData > 0) + fseek(pfile, ((FileHeader_Ktx*)m_pheader)->GetData()->m_u32BytesOfKeyValueData, SEEK_CUR); + szResult = fread(&m_uiEncodingBitsBytes, 1, sizeof(unsigned int), pfile); + assert(szResult > 0); + + m_paucEncodingBits = new unsigned char[m_uiEncodingBitsBytes]; + assert(ftell(pfile) + m_uiEncodingBitsBytes <= fileSize); + szResult = fread(m_paucEncodingBits, 1, m_uiEncodingBitsBytes, pfile); + assert(szResult == m_uiEncodingBitsBytes); + + uint32_t uiInternalFormat = ((FileHeader_Ktx*)m_pheader)->GetData()->m_u32GlInternalFormat; + uint32_t uiBaseInternalFormat = ((FileHeader_Ktx*)m_pheader)->GetData()->m_u32GlBaseInternalFormat; + + if (uiInternalFormat == (uint32_t)FileHeader_Ktx::InternalFormat::ETC1_RGB8 && uiBaseInternalFormat == (uint32_t)FileHeader_Ktx::BaseInternalFormat::ETC1_RGB8) + { + m_imageformat = Image::Format::ETC1; + } + else if (uiInternalFormat == (uint32_t)FileHeader_Ktx::InternalFormat::ETC2_RGB8 && uiBaseInternalFormat == (uint32_t)FileHeader_Ktx::BaseInternalFormat::ETC2_RGB8) + { + m_imageformat = Image::Format::RGB8; + } + else if (uiInternalFormat == (uint32_t)FileHeader_Ktx::InternalFormat::ETC2_RGB8A1 && uiBaseInternalFormat == (uint32_t)FileHeader_Ktx::BaseInternalFormat::ETC2_RGB8A1) + { + m_imageformat = Image::Format::RGB8A1; + } + else if (uiInternalFormat == (uint32_t)FileHeader_Ktx::InternalFormat::ETC2_RGBA8 && uiBaseInternalFormat == (uint32_t)FileHeader_Ktx::BaseInternalFormat::ETC2_RGBA8) + { + m_imageformat = Image::Format::RGBA8; + } + else if (uiInternalFormat == (uint32_t)FileHeader_Ktx::InternalFormat::ETC2_R11 && uiBaseInternalFormat == (uint32_t)FileHeader_Ktx::BaseInternalFormat::ETC2_R11) + { + m_imageformat = Image::Format::R11; + } + else if (uiInternalFormat == (uint32_t)FileHeader_Ktx::InternalFormat::ETC2_SIGNED_R11 && uiBaseInternalFormat == (uint32_t)FileHeader_Ktx::BaseInternalFormat::ETC2_R11) + { + m_imageformat = Image::Format::SIGNED_R11; + } + else if (uiInternalFormat == (uint32_t)FileHeader_Ktx::InternalFormat::ETC2_RG11 && uiBaseInternalFormat == (uint32_t)FileHeader_Ktx::BaseInternalFormat::ETC2_RG11) + { + m_imageformat = Image::Format::RG11; + } + else if (uiInternalFormat == (uint32_t)FileHeader_Ktx::InternalFormat::ETC2_SIGNED_RG11 && uiBaseInternalFormat == (uint32_t)FileHeader_Ktx::BaseInternalFormat::ETC2_RG11) + { + m_imageformat = Image::Format::SIGNED_RG11; + } + else + { + m_imageformat = Image::Format::UNKNOWN; + } + + m_uiSourceWidth = ((FileHeader_Ktx*)m_pheader)->GetData()->m_u32PixelWidth; + m_uiSourceHeight = ((FileHeader_Ktx*)m_pheader)->GetData()->m_u32PixelHeight; + m_uiExtendedWidth = Image::CalcExtendedDimension((unsigned short)m_uiSourceWidth); + m_uiExtendedHeight = Image::CalcExtendedDimension((unsigned short)m_uiSourceHeight); + + unsigned int uiBlocks = m_uiExtendedWidth * m_uiExtendedHeight / 16; + Block4x4EncodingBits::Format encodingbitsformat = Image::DetermineEncodingBitsFormat(m_imageformat); + unsigned int expectedbytes = uiBlocks * Block4x4EncodingBits::GetBytesPerBlock(encodingbitsformat); + assert(expectedbytes == m_uiEncodingBitsBytes); + + fclose(pfile); +} + +File::~File() +{ + + if (m_paucEncodingBits != nullptr) + { + delete[] m_paucEncodingBits; + m_paucEncodingBits = nullptr; + } + if(m_pstrFilename != nullptr) + { + delete[] m_pstrFilename; + m_pstrFilename = nullptr; + } + + if (m_pheader != nullptr) + { + delete m_pheader; + m_pheader = nullptr; + } +} + +void File::UseSingleBlock(int a_iPixelX, int a_iPixelY) +{ + if (a_iPixelX <= -1 || a_iPixelY <= -1) + return; + if (a_iPixelX >(int) m_uiSourceWidth) + { + //if we are using a ktx thats the size of a single block or less + //then make sure we use the 4x4 image as the single block + if (m_uiSourceWidth <= 4) + { + a_iPixelX = 0; + } + else + { + printf("blockAtHV: H coordinate out of range, capped to image width\n"); + a_iPixelX = m_uiSourceWidth - 1; + } + } + if (a_iPixelY >(int) m_uiSourceHeight) + { + //if we are using a ktx thats the size of a single block or less + //then make sure we use the 4x4 image as the single block + if (m_uiSourceHeight <= 4) + { + a_iPixelY= 0; + } + else + { + printf("blockAtHV: V coordinate out of range, capped to image height\n"); + a_iPixelY = m_uiSourceHeight - 1; + } + } + + unsigned int origWidth = m_uiSourceWidth; + unsigned int origHeight = m_uiSourceHeight; + + m_uiSourceWidth = 4; + m_uiSourceHeight = 4; + m_uiExtendedWidth = Image::CalcExtendedDimension((unsigned short)m_uiSourceWidth); + m_uiExtendedHeight = Image::CalcExtendedDimension((unsigned short)m_uiSourceHeight); + + //block position in pixels + // remove the bottom 2 bits to get the block coordinates + unsigned int iBlockPosX = (a_iPixelX & 0xFFFFFFFC); + unsigned int iBlockPosY = (a_iPixelY & 0xFFFFFFFC); + + int numXBlocks = (origWidth / 4); + int numYBlocks = (origHeight / 4); + + Block4x4EncodingBits::Format encodingbitsformat = Image::DetermineEncodingBitsFormat(m_imageformat); + unsigned int uiEncodingBitsBytesPerBlock = Block4x4EncodingBits::GetBytesPerBlock(encodingbitsformat); + + // block location + //int iBlockX = (a_iPixelX % 4) == 0 ? a_iPixelX / 4.0f : (a_iPixelX / 4) + 1; + //int iBlockY = (a_iPixelY % 4) == 0 ? a_iPixelY / 4.0f : (a_iPixelY / 4) + 1; + //m_paucEncodingBits += ((iBlockY * numXBlocks) + iBlockX) * uiEncodingBitsBytesPerBlock; + + + unsigned int num = numXBlocks*numYBlocks; + unsigned int uiH = 0, uiV = 0; + for (unsigned int uiBlock = 0; uiBlock < num; uiBlock++) + { + if (uiH == iBlockPosX && uiV == iBlockPosY) + { + break; + } + m_paucEncodingBits += uiEncodingBitsBytesPerBlock; + uiH += 4; + + if (uiH >= origWidth) + { + uiH = 0; + uiV += 4; + } + } +} +// ---------------------------------------------------------------------------------------------------- +// +void File::Write() +{ + + FILE *pfile = fopen(m_pstrFilename, "wb"); + if (pfile == nullptr) + { + printf("Error: couldn't open Etc file (%s)\n", m_pstrFilename); + exit(1); + } + + m_pheader->Write(pfile); + unsigned int iResult = (int)fwrite(m_paucEncodingBits, 1, m_uiEncodingBitsBytes, pfile); + if (iResult != m_uiEncodingBitsBytes) + { + printf("Error: couldn't write Etc file (%s)\n", m_pstrFilename); + exit(1); + } + + fclose(pfile); + +} + +// ---------------------------------------------------------------------------------------------------- +// + diff --git a/EtcTool/EtcFile.h b/EtcTool/EtcFile.h new file mode 100644 index 0000000..d55b050 --- /dev/null +++ b/EtcTool/EtcFile.h @@ -0,0 +1,99 @@ +/* + * Copyright 2015 The Etc2Comp Authors. + * + * Licensed under the Apache License, Version 2.0 (the "License"); + * you may not use this file except in compliance with the License. + * You may obtain a copy of the License at + * + * http://www.apache.org/licenses/LICENSE-2.0 + * + * Unless required by applicable law or agreed to in writing, software + * distributed under the License is distributed on an "AS IS" BASIS, + * WITHOUT WARRANTIES OR CONDITIONS OF ANY KIND, either express or implied. + * See the License for the specific language governing permissions and + * limitations under the License. + */ + +#pragma once + +#include "EtcColorFloatRGBA.h" +#include "EtcImage.h" + +namespace Etc +{ + class FileHeader; + class SourceImage; + + class File + { + public: + + enum class Format + { + INFER_FROM_FILE_EXTENSION, + PKM, + KTX, + }; + + File(const char *a_pstrFilename, Format a_fileformat, Image::Format a_imageformat, + unsigned char *a_paucEncodingBits, unsigned int a_uiEncodingBitsBytes, + unsigned int a_uiSourceWidth, unsigned int a_uiSourceHeight, + unsigned int a_uiExtendedWidth, unsigned int a_uiExtendedHeight); + + File(const char *a_pstrFilename, Format a_fileformat); + ~File(); + const char *GetFilename(void) { return m_pstrFilename; } + + void Read(const char *a_pstrFilename); + void Write(void); + + inline unsigned int GetSourceWidth(void) + { + return m_uiSourceWidth; + } + + inline unsigned int GetSourceHeight(void) + { + return m_uiSourceHeight; + } + + inline unsigned int GetExtendedWidth(void) + { + return m_uiExtendedWidth; + } + + inline unsigned int GetExtendedHeight(void) + { + return m_uiExtendedHeight; + } + + inline Image::Format GetImageFormat() + { + return m_imageformat; + } + + inline unsigned int GetEncodingBitsBytes() + { + return m_uiEncodingBitsBytes; + } + + inline unsigned char * GetEncodingBits() + { + return m_paucEncodingBits; + } + void UseSingleBlock(int a_iPixelX = -1, int a_iPixelY = -1); + private: + + char *m_pstrFilename; // includes directory path and file extension + Format m_fileformat; + Image::Format m_imageformat; + FileHeader *m_pheader; + unsigned char *m_paucEncodingBits; + unsigned int m_uiEncodingBitsBytes; + unsigned int m_uiSourceWidth; + unsigned int m_uiSourceHeight; + unsigned int m_uiExtendedWidth; + unsigned int m_uiExtendedHeight; + }; + +} diff --git a/EtcTool/EtcFileHeader.cpp b/EtcTool/EtcFileHeader.cpp new file mode 100644 index 0000000..19f0662 --- /dev/null +++ b/EtcTool/EtcFileHeader.cpp @@ -0,0 +1,192 @@ +/* + * Copyright 2015 The Etc2Comp Authors. + * + * Licensed under the Apache License, Version 2.0 (the "License"); + * you may not use this file except in compliance with the License. + * You may obtain a copy of the License at + * + * http://www.apache.org/licenses/LICENSE-2.0 + * + * Unless required by applicable law or agreed to in writing, software + * distributed under the License is distributed on an "AS IS" BASIS, + * WITHOUT WARRANTIES OR CONDITIONS OF ANY KIND, either express or implied. + * See the License for the specific language governing permissions and + * limitations under the License. + */ + +#include "EtcFileHeader.h" + +#include "EtcBlock4x4EncodingBits.h" + +#include + +namespace Etc +{ + + // ---------------------------------------------------------------------------------------------------- + // + FileHeader_Pkm::FileHeader_Pkm(File *a_pfile) + { + m_pfile = a_pfile; + + static const char s_acMagicNumberData[4] = { 'P', 'K', 'M', ' ' }; + static const char s_acVersionData[2] = { '1', '0' }; + + for (unsigned int ui = 0; ui < sizeof(s_acMagicNumberData); ui++) + { + m_data.m_acMagicNumber[ui] = s_acMagicNumberData[ui]; + } + + for (unsigned int ui = 0; ui < sizeof(s_acVersionData); ui++) + { + m_data.m_acVersion[ui] = s_acVersionData[ui]; + } + + m_data.m_ucDataType_msb = 0; // ETC1_RGB_NO_MIPMAPS + m_data.m_ucDataType_lsb = 0; + + m_data.m_ucOriginalWidth_msb = (unsigned char)(m_pfile->GetSourceWidth() >> 8); + m_data.m_ucOriginalWidth_lsb = m_pfile->GetSourceWidth() & 0xFF; + m_data.m_ucOriginalHeight_msb = (unsigned char)(m_pfile->GetSourceHeight() >> 8); + m_data.m_ucOriginalHeight_lsb = m_pfile->GetSourceHeight() & 0xFF; + + m_data.m_ucExtendedWidth_msb = (unsigned char)(m_pfile->GetExtendedWidth() >> 8); + m_data.m_ucExtendedWidth_lsb = m_pfile->GetExtendedWidth() & 0xFF; + m_data.m_ucExtendedHeight_msb = (unsigned char)(m_pfile->GetExtendedHeight() >> 8); + m_data.m_ucExtendedHeight_lsb = m_pfile->GetExtendedHeight() & 0xFF; + + } + + // ---------------------------------------------------------------------------------------------------- + // + void FileHeader_Pkm::Write(FILE *a_pfile) + { + + fwrite(&m_data, sizeof(Data), 1, a_pfile); + + } + + // ---------------------------------------------------------------------------------------------------- + // + FileHeader_Ktx::FileHeader_Ktx(File *a_pfile) + { + m_pfile = a_pfile; + + static const uint8_t s_au8Itentfier[12] = + { + 0xAB, 0x4B, 0x54, 0x58, // first four bytes of Byte[12] identifier + 0x20, 0x31, 0x31, 0xBB, // next four bytes of Byte[12] identifier + 0x0D, 0x0A, 0x1A, 0x0A // final four bytes of Byte[12] identifier + }; + + for (unsigned int ui = 0; ui < sizeof(s_au8Itentfier); ui++) + { + m_data.m_au8Identifier[ui] = s_au8Itentfier[ui]; + } + + m_data.m_u32Endianness = 0x04030201; + m_data.m_u32GlType = 0; + m_data.m_u32GlTypeSize = 1; + m_data.m_u32GlFormat = 0; + + switch (m_pfile->GetImageFormat()) + { + case Image::Format::RGB8: + case Image::Format::SRGB8: + m_data.m_u32GlInternalFormat = (unsigned int)InternalFormat::ETC2_RGB8; + m_data.m_u32GlBaseInternalFormat = (unsigned int)BaseInternalFormat::ETC2_RGB8; + break; + + case Image::Format::RGBA8: + case Image::Format::SRGBA8: + m_data.m_u32GlInternalFormat = (unsigned int)InternalFormat::ETC2_RGBA8; + m_data.m_u32GlBaseInternalFormat = (unsigned int)BaseInternalFormat::ETC2_RGBA8; + break; + + case Image::Format::RGB8A1: + case Image::Format::SRGB8A1: + m_data.m_u32GlInternalFormat = (unsigned int)InternalFormat::ETC2_RGB8A1; + m_data.m_u32GlBaseInternalFormat = (unsigned int)BaseInternalFormat::ETC2_RGB8A1; + break; + + case Image::Format::R11: + m_data.m_u32GlInternalFormat = (unsigned int)InternalFormat::ETC2_R11; + m_data.m_u32GlBaseInternalFormat = (unsigned int)BaseInternalFormat::ETC2_R11; + break; + + case Image::Format::SIGNED_R11: + m_data.m_u32GlInternalFormat = (unsigned int)InternalFormat::ETC2_SIGNED_R11; + m_data.m_u32GlBaseInternalFormat = (unsigned int)BaseInternalFormat::ETC2_R11; + break; + + case Image::Format::RG11: + m_data.m_u32GlInternalFormat = (unsigned int)InternalFormat::ETC2_RG11; + m_data.m_u32GlBaseInternalFormat = (unsigned int)BaseInternalFormat::ETC2_RG11; + break; + + case Image::Format::SIGNED_RG11: + m_data.m_u32GlInternalFormat = (unsigned int)InternalFormat::ETC2_SIGNED_RG11; + m_data.m_u32GlBaseInternalFormat = (unsigned int)BaseInternalFormat::ETC2_RG11; + break; + + default: + m_data.m_u32GlInternalFormat = (unsigned int)InternalFormat::ETC1_RGB8; + m_data.m_u32GlBaseInternalFormat = (unsigned int)BaseInternalFormat::ETC1_RGB8; + break; + } + + m_data.m_u32PixelWidth = 0; + m_data.m_u32PixelHeight = 0; + m_data.m_u32PixelDepth = 0; + m_data.m_u32NumberOfArrayElements = 0; + m_data.m_u32NumberOfFaces = 0; + m_data.m_u32NumberOfMipmapLevels = 0; + m_data.m_u32BytesOfKeyValueData = 0; + + m_pkeyvaluepair = nullptr; + + m_u32Images = 0; + m_u32KeyValuePairs = 0; + + m_data.m_u32PixelWidth = m_pfile->GetSourceWidth(); + m_data.m_u32PixelHeight = m_pfile->GetSourceHeight(); + m_data.m_u32PixelDepth = 0; + m_data.m_u32NumberOfArrayElements = 0; + m_data.m_u32NumberOfFaces = 1; + m_data.m_u32NumberOfMipmapLevels = 1; + + } + + // ---------------------------------------------------------------------------------------------------- + // + void FileHeader_Ktx::Write(FILE *a_pfile) + { + size_t szBytesWritten; + + // Write header + szBytesWritten = fwrite(&m_data, 1, sizeof(Data), a_pfile); + assert(szBytesWritten == sizeof(Data)); + + // Write KeyAndValuePairs + if (m_u32KeyValuePairs) + { + fwrite(m_pkeyvaluepair, m_pkeyvaluepair->u32KeyAndValueByteSize, 1, a_pfile); + } + + // Write u32 image size + uint32_t u32ImageSize = m_pfile->GetEncodingBitsBytes(); + szBytesWritten = fwrite(&u32ImageSize, 1, sizeof(u32ImageSize), a_pfile); + assert(szBytesWritten == sizeof(u32ImageSize)); + + } + + // ---------------------------------------------------------------------------------------------------- + // + FileHeader_Ktx::Data *FileHeader_Ktx::GetData() + { + return &m_data; + } + + // ---------------------------------------------------------------------------------------------------- + // +} // namespace Etc diff --git a/EtcTool/EtcFileHeader.h b/EtcTool/EtcFileHeader.h new file mode 100644 index 0000000..4356afe --- /dev/null +++ b/EtcTool/EtcFileHeader.h @@ -0,0 +1,146 @@ +/* + * Copyright 2015 The Etc2Comp Authors. + * + * Licensed under the Apache License, Version 2.0 (the "License"); + * you may not use this file except in compliance with the License. + * You may obtain a copy of the License at + * + * http://www.apache.org/licenses/LICENSE-2.0 + * + * Unless required by applicable law or agreed to in writing, software + * distributed under the License is distributed on an "AS IS" BASIS, + * WITHOUT WARRANTIES OR CONDITIONS OF ANY KIND, either express or implied. + * See the License for the specific language governing permissions and + * limitations under the License. + */ + +#pragma once + +#include "EtcFile.h" +#include +#include + +namespace Etc +{ + + class Image; + + class FileHeader + { + public: + + virtual void Write(FILE *a_pfile) = 0; + File GetFile(); + virtual ~FileHeader(void) {} + protected: + + File *m_pfile; + }; + + // ---------------------------------------------------------------------------------------------------- + // + class FileHeader_Pkm : public FileHeader + { + public: + + FileHeader_Pkm(File *a_pfile); + + virtual void Write(FILE *a_pfile); + virtual ~FileHeader_Pkm(void) {} + private: + + typedef struct + { + char m_acMagicNumber[4]; + char m_acVersion[2]; + unsigned char m_ucDataType_msb; // e.g. ETC1_RGB_NO_MIPMAPS + unsigned char m_ucDataType_lsb; + unsigned char m_ucExtendedWidth_msb; // padded to 4x4 blocks + unsigned char m_ucExtendedWidth_lsb; + unsigned char m_ucExtendedHeight_msb; // padded to 4x4 blocks + unsigned char m_ucExtendedHeight_lsb; + unsigned char m_ucOriginalWidth_msb; + unsigned char m_ucOriginalWidth_lsb; + unsigned char m_ucOriginalHeight_msb; + unsigned char m_ucOriginalHeight_lsb; + } Data; + + Data m_data; + }; + + // ---------------------------------------------------------------------------------------------------- + // + class FileHeader_Ktx : public FileHeader + { + public: + + typedef struct + { + uint32_t u32KeyAndValueByteSize; + } KeyValuePair; + + typedef struct + { + uint8_t m_au8Identifier[12]; + uint32_t m_u32Endianness; + uint32_t m_u32GlType; + uint32_t m_u32GlTypeSize; + uint32_t m_u32GlFormat; + uint32_t m_u32GlInternalFormat; + uint32_t m_u32GlBaseInternalFormat; + uint32_t m_u32PixelWidth; + uint32_t m_u32PixelHeight; + uint32_t m_u32PixelDepth; + uint32_t m_u32NumberOfArrayElements; + uint32_t m_u32NumberOfFaces; + uint32_t m_u32NumberOfMipmapLevels; + uint32_t m_u32BytesOfKeyValueData; + } Data; + + enum class InternalFormat + { + ETC1_RGB8 = 0x8D64, + ETC1_ALPHA8 = ETC1_RGB8, + // + ETC2_R11 = 0x9270, + ETC2_SIGNED_R11 = 0x9271, + ETC2_RG11 = 0x9272, + ETC2_SIGNED_RG11 = 0x9273, + ETC2_RGB8 = 0x9274, + ETC2_SRGB8 = 0x9275, + ETC2_RGB8A1 = 0x9276, + ETC2_SRGB8_PUNCHTHROUGH_ALPHA1 = 0x9277, + ETC2_RGBA8 = 0x9278 + }; + + enum class BaseInternalFormat + { + ETC2_R11 = 0x1903, + ETC2_RG11 = 0x8227, + ETC1_RGB8 = 0x1907, + ETC1_ALPHA8 = ETC1_RGB8, + // + ETC2_RGB8 = 0x1907, + ETC2_RGB8A1 = 0x1908, + ETC2_RGBA8 = 0x1908, + }; + + FileHeader_Ktx(File *a_pfile); + + virtual void Write(FILE *a_pfile); + virtual ~FileHeader_Ktx(void) {} + + void AddKeyAndValue(KeyValuePair *a_pkeyvaluepair); + + Data* GetData(); + + private: + + Data m_data; + KeyValuePair *m_pkeyvaluepair; + + uint32_t m_u32Images; + uint32_t m_u32KeyValuePairs; + }; + +} // namespace Etc diff --git a/EtcTool/EtcMemTest.cpp b/EtcTool/EtcMemTest.cpp new file mode 100644 index 0000000..855f1be --- /dev/null +++ b/EtcTool/EtcMemTest.cpp @@ -0,0 +1,258 @@ +/* + * Copyright 2015 The Etc2Comp Authors. + * + * Licensed under the Apache License, Version 2.0 (the "License"); + * you may not use this file except in compliance with the License. + * You may obtain a copy of the License at + * + * http://www.apache.org/licenses/LICENSE-2.0 + * + * Unless required by applicable law or agreed to in writing, software + * distributed under the License is distributed on an "AS IS" BASIS, + * WITHOUT WARRANTIES OR CONDITIONS OF ANY KIND, either express or implied. + * See the License for the specific language governing permissions and + * limitations under the License. + */ + +#include "EtcMemTest.h" +#include +#include +#include +#include +#include +using namespace std; + +#if defined(_WIN32) +#include +#include +// Use to convert bytes to KB +#define DIV 1024 + +// Specify the width of the field in which to print the numbers. +// The asterisk in the format specifier "%*I64d" takes an integer +// argument and uses it to pad and right justify the number. +#define WIDTH 7 + +size_t GetMemoryUsageAmount() +{ + //MEMORYSTATUS status; // if 64 bit version isnt working, use 32bit version + //GlobalMemoryStatus(&status); + /*MEMORYSTATUSEX status; + status.dwLength = sizeof(status); + GlobalMemoryStatusEx(&status); + return (size_t)(status.ullAvailPhys / DIV);*/ + + PROCESS_MEMORY_COUNTERS_EX pmc; + DWORD ret = GetProcessMemoryInfo(GetCurrentProcess(), (PROCESS_MEMORY_COUNTERS*)&pmc, sizeof(pmc)); + if (ret == 0) + { + printf("GetProcessMemoryInfo failed with code %d\n", GetLastError()); + } + + return pmc.PrivateUsage; +} + +void PrintWindowsMemUsage() +{ + MEMORYSTATUSEX statex; + + statex.dwLength = sizeof(statex); + + GlobalMemoryStatusEx(&statex); + + printf("There is %*ld percent of memory in use.\n", + WIDTH, statex.dwMemoryLoad); + printf("There are %*I64d total KB of physical memory.\n", + WIDTH, statex.ullTotalPhys / DIV); + printf("There are %*I64d free KB of physical memory.\n", + WIDTH, statex.ullAvailPhys / DIV); + printf("There are %*I64d total KB of paging file.\n", + WIDTH, statex.ullTotalPageFile / DIV); + printf("There are %*I64d free KB of paging file.\n", + WIDTH, statex.ullAvailPageFile / DIV); + printf("There are %*I64d total KB of virtual memory.\n", + WIDTH, statex.ullTotalVirtual / DIV); + printf("There are %*I64d free KB of virtual memory.\n", + WIDTH, statex.ullAvailVirtual / DIV); + + // Show the amount of extended memory available. + + printf("There are %*I64d free KB of extended memory.\n", + WIDTH, statex.ullAvailExtendedVirtual / DIV); +} + + + + +int RunMemTest(bool verboseOutput, size_t numTestIterations) +{ + WIN32_FIND_DATA ffd; + srand(static_cast(time(NULL))); + HANDLE hFind = INVALID_HANDLE_VALUE; + + + vector allImages; + + string imagesDir = "D:\\source\\\etc2Sourcetree\\TestImages\\"; + string outputDir = "C:\\Users\\BSI\\Desktop\\etc2comp\\"; + + hFind = FindFirstFile((imagesDir+"*").c_str(), &ffd); + + if (INVALID_HANDLE_VALUE == hFind) + { + printf ("image dir doesnt exist: %s\n", imagesDir.c_str()); + return -1; + } + + do + { + if (ffd.dwFileAttributes & FILE_ATTRIBUTE_DIRECTORY) + { + continue; + } + else + { + string filename = ffd.cFileName; + size_t found = filename.find_last_of("."); + string ext = filename.substr(found + 1); + if(ext != "png") + continue; + + allImages.push_back(filename); + } + } while (FindNextFile(hFind, &ffd) != 0); + FindClose(hFind); + + size_t oldMemSize = 0; + size_t curMemSize = 0; + size_t fileIndex = 0; + int encodingType = 2;//0 = c , 1= both c and c++ encoding, 2 = c++ encoding + bool encodeRandomImage = true; + for (size_t i = 0; i < numTestIterations; i++) + { + encodingType = rand()%3; + + printf("--------------iteration %zu-----------\n",i+1); + int randomNumber = 1 + rand() % 11; + //Image::Format format = (Image::Format)randomNumber; + Image::Format format = Image::Format::RGBA8; + + randomNumber = rand() % 4; + ErrorMetric e_ErrMetric = (ErrorMetric)randomNumber; + + randomNumber = rand() % 100; + float curEffort = (float)randomNumber; + + int jobs = 1 + rand() % 8; + long memDiff = 0; + + if (verboseOutput) + { + printf(" effort = %.f\n", curEffort); + printf(" encoding = %s\n", Image::EncodingFormatToString(format)); + printf(" error metric: %s\n", ErrorMetricToString(e_ErrMetric)); + printf("jobs: %d\n", jobs); + } + + oldMemSize = curMemSize; + curMemSize = GetMemoryUsageAmount(); + memDiff = (long)curMemSize - (long)oldMemSize; + printf("mem diff: %ukb\n", memDiff); + + string curFile = imagesDir + allImages[fileIndex]; + string outFile = outputDir + allImages[fileIndex] +".ktx"; + printf("file: %s\n", curFile.c_str()); + + if (!encodeRandomImage) + { + if (fileIndex >= allImages.size() - 1) + fileIndex = 0; + else + fileIndex++; + } + else + { + fileIndex = rand()%allImages.size(); + } + SourceImage sourceimage(curFile.c_str(), -1, -1); + + unsigned int uiSourceWidth = sourceimage.GetWidth(); + unsigned int uiSourceHeight = sourceimage.GetHeight(); + //char *imgonnaleak = new char[1000]; + + unsigned char *paucEncodingBits; + unsigned int uiEncodingBitsBytes; + unsigned int uiExtendedWidth; + unsigned int uiExtendedHeight; + int iEncodingTime_ms; + //////////////C INTERFACE FIRST////////////////////// + if (encodingType < 2) + { + if (verboseOutput) + { + printf("start C interface Encoding:\n"); + } + + Etc::Encode((float *)sourceimage.GetPixels(), + uiSourceWidth, uiSourceHeight, + format, + e_ErrMetric, + curEffort, + jobs, + MAX_JOBS, + &paucEncodingBits, &uiEncodingBitsBytes, + &uiExtendedWidth, &uiExtendedHeight, + &iEncodingTime_ms, verboseOutput); + if (verboseOutput) + { + printf(" encode time = %dms\n", iEncodingTime_ms); + printf("EncodedImage: %s\n", outFile.c_str()); + } + Etc::File C_interfaceEtcfile(outFile.c_str(), Etc::File::Format::INFER_FROM_FILE_EXTENSION, + format, + paucEncodingBits, uiEncodingBitsBytes, + uiSourceWidth, uiSourceHeight, + uiExtendedWidth, uiExtendedHeight); + C_interfaceEtcfile.Write(); + + oldMemSize = curMemSize; + curMemSize = GetMemoryUsageAmount(); + memDiff = (long)curMemSize - (long)oldMemSize; + printf("mem diff again: %ukb\n", memDiff); + } + //////////////C++ INTERFACE /////////////////////////// + if (encodingType > 0) + { + if (verboseOutput) + { + printf("start C++ Encoding:\n"); + } + Etc::Image image((float *)sourceimage.GetPixels(), + uiSourceWidth, uiSourceHeight, + e_ErrMetric); + image.m_bVerboseOutput = verboseOutput; + + Etc::Image::EncodingStatus encStatus = Etc::Image::EncodingStatus::SUCCESS; + + encStatus = image.Encode(format, e_ErrMetric, curEffort, jobs, MAX_JOBS); + if (verboseOutput) + { + printf(" encode time = %dms\n", image.GetEncodingTimeMs()); + printf("EncodedImage: %s\n", outFile.c_str()); + printf("status bitfield: %u\n", encStatus); + } + Etc::File etcfile(outFile.c_str(), Etc::File::Format::INFER_FROM_FILE_EXTENSION, + format, + image.GetEncodingBits(), image.GetEncodingBitsBytes(), + image.GetSourceWidth(), image.GetSourceHeight(), + image.GetExtendedWidth(), image.GetExtendedHeight()); + + etcfile.Write(); + } + + } + + return 0; +} + +#endif diff --git a/EtcTool/EtcMemTest.h b/EtcTool/EtcMemTest.h new file mode 100644 index 0000000..f225159 --- /dev/null +++ b/EtcTool/EtcMemTest.h @@ -0,0 +1,61 @@ +/* + * Copyright 2015 The Etc2Comp Authors. + * + * Licensed under the Apache License, Version 2.0 (the "License"); + * you may not use this file except in compliance with the License. + * You may obtain a copy of the License at + * + * http://www.apache.org/licenses/LICENSE-2.0 + * + * Unless required by applicable law or agreed to in writing, software + * distributed under the License is distributed on an "AS IS" BASIS, + * WITHOUT WARRANTIES OR CONDITIONS OF ANY KIND, either express or implied. + * See the License for the specific language governing permissions and + * limitations under the License. + */ + +#pragma once + +#include "EtcConfig.h" + +#include "Etc.h" + +#include "EtcSourceImage.h" +#include "EtcFile.h" +#include "EtcMath.h" +#include "EtcImage.h" +#include "EtcErrorMetric.h" +#include "EtcBlock4x4EncodingBits.h" +#include "EtcFileHeader.h" + +#include +#include +#include + +#define MAX_JOBS 1024 +using namespace Etc; + +int RunMemTest(bool verboseOutput, size_t numTestIterations); + + + + + +/*int getMem() +{ + +int tSize = 0, resident = 0, share = 0; +ifstream buffer("/proc/self/statm"); +buffer >> tSize >> resident >> share; +buffer.close(); + +long page_size_kb = sysconf(_SC_PAGE_SIZE) / 1024; // in case x86-64 is configured to use 2MB pages +double rss = resident * page_size_kb; +cout << "RSS - " << rss << " kB\n"; + +double shared_mem = share * page_size_kb; +cout << "Shared Memory - " << shared_mem << " kB\n"; + +cout << "Private Memory - " << rss - shared_mem << "kB\n"; +return 0; +}*/ diff --git a/EtcTool/EtcSourceImage.cpp b/EtcTool/EtcSourceImage.cpp new file mode 100644 index 0000000..0c64fea --- /dev/null +++ b/EtcTool/EtcSourceImage.cpp @@ -0,0 +1,292 @@ +/* + * Copyright 2015 The Etc2Comp Authors. + * + * Licensed under the Apache License, Version 2.0 (the "License"); + * you may not use this file except in compliance with the License. + * You may obtain a copy of the License at + * + * http://www.apache.org/licenses/LICENSE-2.0 + * + * Unless required by applicable law or agreed to in writing, software + * distributed under the License is distributed on an "AS IS" BASIS, + * WITHOUT WARRANTIES OR CONDITIONS OF ANY KIND, either express or implied. + * See the License for the specific language governing permissions and + * limitations under the License. + */ + +#ifdef _WIN32 +#define _CRT_SECURE_NO_WARNINGS (1) +#endif + +#include "EtcConfig.h" +#include "EtcSourceImage.h" +#include "Etc.h" + +#if USE_STB_IMAGE_LOAD +#include "stb_image.h" +#endif + +#include +#include +#include +#include + +#include "lodepng.h" + +namespace Etc +{ + + // ---------------------------------------------------------------------------------------------------- + // + SourceImage::SourceImage(const char *a_pstrFilename, int a_iPixelX, int a_iPixelY) + { + m_pstrFilename = nullptr; + m_pstrName = nullptr; + m_pstrFileExtension = nullptr; + m_uiWidth = 0; + m_uiHeight = 0; + m_pafrgbaPixels = nullptr; + + SetName(a_pstrFilename); + + Read(a_iPixelX, a_iPixelY); + } + + // ---------------------------------------------------------------------------------------------------- + // + SourceImage::SourceImage(ColorFloatRGBA *a_pafrgbaSource, + unsigned int a_uiSourceWidth, + unsigned int a_uiSourceHeight) + { + m_pstrFilename = nullptr; + m_pstrName = nullptr; + m_pstrFileExtension = nullptr; + m_uiWidth = a_uiSourceWidth; + m_uiHeight = a_uiSourceHeight; + m_pafrgbaPixels = a_pafrgbaSource; + + } + // ---------------------------------------------------------------------------------------------------- + // + SourceImage::~SourceImage() + { + if (m_pstrFilename != nullptr) + { + delete[] m_pstrFilename; + m_pstrFilename = nullptr; + } + + if (m_pstrName != nullptr) + { + delete[] m_pstrName; + m_pstrName = nullptr; + m_pstrFileExtension = nullptr; + } + + if (m_pafrgbaPixels != nullptr) + { + delete[] m_pafrgbaPixels; + m_pafrgbaPixels = nullptr; + } + m_uiWidth = 0; + m_uiHeight = 0; + } + // ---------------------------------------------------------------------------------------------------- + // + void SourceImage::Read(int a_iPixelX, int a_iPixelY) + { + unsigned char* paucPixels = nullptr; + + int iWidth; + int iHeight; + int iBitsPerPixel; + bool bool16BitImage = false; + +#if USE_STB_IMAGE_LOAD + //if stb_iamge is available, only use it to load files other than png + char *fileExt = strrchr(m_pstrFilename, '.'); + + if (strcmp(fileExt, ".png") != 0) + { + paucPixels = stbi_load(m_pstrFilename, &iWidth, &iHeight, &iBitsPerPixel, 4); + + if (paucPixels == nullptr) + { + printf("stb_image error %s\n", stbi_failure_reason()); + assert(0); + exit(1); + } + } +#endif + + if (paucPixels == nullptr) + { + //we can load 8 or 16 bit pngs + int iBitDepth = 16; + int error = lodepng_decode_file(&paucPixels, + (unsigned int*)&iWidth, (unsigned int*)&iHeight, + m_pstrFilename, + LCT_RGBA, iBitDepth); + + bool16BitImage = (iBitDepth == 16) ? true : false; + if (error) + { + printf("lodePNG error %u: %s\n", error, lodepng_error_text(error)); + assert(0); + exit(1); + } + } + + //the pixel cords for the top left corner of the block + int iBlockX = 0; + int iBlockY = 0; + if (a_iPixelX > -1 && a_iPixelY > -1) + { + // in 1 block mode, we basically will have an img thats 4x4 + m_uiWidth = 4; + m_uiHeight = 4; + + if(a_iPixelX > iWidth) + a_iPixelX = iWidth; + if (a_iPixelY > iHeight) + a_iPixelY = iHeight; + + // remove the bottom 2 bits to get the block coordinates + iBlockX = (a_iPixelX & 0xFFFFFFFC); + iBlockY = (a_iPixelY & 0xFFFFFFFC); + } + else + { + m_uiWidth = iWidth; + m_uiHeight = iHeight; + } + + m_pafrgbaPixels = new ColorFloatRGBA[m_uiWidth * m_uiHeight]; + assert(m_pafrgbaPixels); + + int iBytesPerPixel = bool16BitImage ? 8 : 4; + unsigned char *pucPixel; // = &paucPixels[(iBlockY * iWidth + iBlockX) * iBytesPerPixel]; + ColorFloatRGBA *pfrgbaPixel = m_pafrgbaPixels; + + // convert pixels from RGBA* to ColorFloatRGBA + for (unsigned int uiV = iBlockY; uiV < (iBlockY+m_uiHeight); ++uiV) + { + // reset coordinate for each row + pucPixel = &paucPixels[(uiV * iWidth + iBlockX) * iBytesPerPixel]; + + // read each row + for (unsigned int uiH = iBlockX; uiH < (iBlockX+m_uiWidth); ++uiH) + { + if (bool16BitImage) + { + unsigned short ushR = (pucPixel[0]<<8) + pucPixel[1]; + unsigned short ushG = (pucPixel[2]<<8) + pucPixel[3]; + unsigned short ushB = (pucPixel[4]<<8) + pucPixel[5]; + unsigned short ushA = (pucPixel[6]<<8) + pucPixel[7]; + + *pfrgbaPixel++ = ColorFloatRGBA((float)ushR / 65535.0f, + (float)ushG / 65535.0f, + (float)ushB / 65535.0f, + (float)ushA / 65535.0f); + } + else + { + *pfrgbaPixel++ = ColorFloatRGBA::ConvertFromRGBA8(pucPixel[0], pucPixel[1], + pucPixel[2], pucPixel[3]); + } + + pucPixel += iBytesPerPixel; + } + } + +#if USE_STB_IMAGE_LOAD + stbi_image_free(paucPixels); +#else + free(paucPixels); +#endif + } + + // ---------------------------------------------------------------------------------------------------- + // sets m_pstrFilename, m_pstrName and m_pstrFileExtension + // + void SourceImage::SetName(const char *a_pstrFilename) + { + if (a_pstrFilename == nullptr) + { + return; + } + + m_pstrFilename = new char[strlen(a_pstrFilename) + 1]; + strcpy(m_pstrFilename, a_pstrFilename); + + m_pstrName = new char[strlen(m_pstrFilename) + 1]; + + // ignore directory path + char *pcLastSlash = strrchr(m_pstrFilename, '/'); + char *pcLastBackSlash = strrchr(m_pstrFilename, '\\'); + if (pcLastSlash == nullptr && pcLastBackSlash == nullptr) + { + strcpy(m_pstrName, m_pstrFilename); + } + else if (pcLastSlash > pcLastBackSlash) + { + strcpy(m_pstrName, pcLastSlash + 1); + } + else + { + strcpy(m_pstrName, pcLastBackSlash + 1); + } + + // find file extension and remove it from image name + char *pcLastPeriod = strrchr(m_pstrName, '.'); + if (pcLastPeriod != nullptr) + { + m_pstrFileExtension = pcLastPeriod + 1; + *strrchr(m_pstrName, '.') = 0; + } + + } + + // ---------------------------------------------------------------------------------------------------- + // + void SourceImage::NormalizeXYZ(void) + { + int iPixels = m_uiWidth * m_uiHeight; + + ColorFloatRGBA *pfrgbaPixel = m_pafrgbaPixels; + for (int iPixel = 0; iPixel < iPixels; iPixel++) + { + float fX = 2.0f*pfrgbaPixel->fR - 1.0f; + float fY = 2.0f*pfrgbaPixel->fG - 1.0f; + float fZ = 2.0f*pfrgbaPixel->fB - 1.0f; + + float fLength2 = fX*fX + fY*fY + fZ*fZ; + + if (fLength2 == 0.0f) + { + pfrgbaPixel->fR = 1.0f; + pfrgbaPixel->fG = 0.0f; + pfrgbaPixel->fB = 0.0f; + } + else + { + float fLength = sqrtf(fLength2); + + float fNormalizedX = fX / fLength; + float fNormalizedY = fY / fLength; + float fNormalizedZ = fZ / fLength; + + pfrgbaPixel->fR = 0.5f * (fNormalizedX + 1.0f); + pfrgbaPixel->fG = 0.5f * (fNormalizedY + 1.0f); + pfrgbaPixel->fB = 0.5f * (fNormalizedZ + 1.0f); + } + + pfrgbaPixel++; + } + + } + + // ---------------------------------------------------------------------------------------------------- + // + +} // namespace Etc diff --git a/EtcTool/EtcSourceImage.h b/EtcTool/EtcSourceImage.h new file mode 100644 index 0000000..0e57e6e --- /dev/null +++ b/EtcTool/EtcSourceImage.h @@ -0,0 +1,93 @@ +/* + * Copyright 2015 The Etc2Comp Authors. + * + * Licensed under the Apache License, Version 2.0 (the "License"); + * you may not use this file except in compliance with the License. + * You may obtain a copy of the License at + * + * http://www.apache.org/licenses/LICENSE-2.0 + * + * Unless required by applicable law or agreed to in writing, software + * distributed under the License is distributed on an "AS IS" BASIS, + * WITHOUT WARRANTIES OR CONDITIONS OF ANY KIND, either express or implied. + * See the License for the specific language governing permissions and + * limitations under the License. + */ + +#pragma once + +#include "EtcColorFloatRGBA.h" + +namespace Etc +{ + + class SourceImage + { + public: + + SourceImage(const char *a_pstrFilename, int a_iPixelX = -1, int a_iPixelY = -1); + + SourceImage(ColorFloatRGBA *a_pafrgbaSource, + unsigned int a_uiSourceWidth, + unsigned int a_uiSourceHeight); + + ~SourceImage(); + + void SetName(const char *a_pstrFilename); + + void NormalizeXYZ(void); + + inline const char *GetFilename(void) + { + return m_pstrFilename; + } + + inline const char *GetName(void) + { + return m_pstrName; + } + + inline const char *GetFileExtension(void) + { + return m_pstrFileExtension; + } + + inline unsigned int GetWidth(void) + { + return m_uiWidth; + } + + inline unsigned int GetHeight(void) + { + return m_uiHeight; + } + + inline ColorFloatRGBA * GetPixels(void) + { + return m_pafrgbaPixels; + } + + inline ColorFloatRGBA * GetPixel(unsigned int a_uiColumn, unsigned int a_uiRow) + { + if (m_pafrgbaPixels == nullptr) + { + return nullptr; + } + + return &m_pafrgbaPixels[a_uiRow*m_uiWidth + a_uiColumn]; + } + + private: + + void Read(int a_iPixelX = -1, int a_iPixelY = -1); + + char *m_pstrFilename; // includes directory path and file extension + char *m_pstrName; // file name with directory path and file extension removed + char *m_pstrFileExtension; + unsigned int m_uiWidth; // not necessarily block aligned + unsigned int m_uiHeight; // not necessarily block aligned + ColorFloatRGBA *m_pafrgbaPixels; + + }; + +} // namespace Etc diff --git a/EtcTool/EtcTool.cpp b/EtcTool/EtcTool.cpp new file mode 100644 index 0000000..bc478fc --- /dev/null +++ b/EtcTool/EtcTool.cpp @@ -0,0 +1,719 @@ +/* + * Copyright 2015 The Etc2Comp Authors. + * + * Licensed under the Apache License, Version 2.0 (the "License"); + * you may not use this file except in compliance with the License. + * You may obtain a copy of the License at + * + * http://www.apache.org/licenses/LICENSE-2.0 + * + * Unless required by applicable law or agreed to in writing, software + * distributed under the License is distributed on an "AS IS" BASIS, + * WITHOUT WARRANTIES OR CONDITIONS OF ANY KIND, either express or implied. + * See the License for the specific language governing permissions and + * limitations under the License. + */ + +#ifdef _WIN32 +#define _CRT_SECURE_NO_WARNINGS (1) +#endif +/* +since this code will be used on a wide varity of platforms and configurations +its important to have some sort of sanity check for the amount of threads that can be used. +change this macro to suit your configuration. This will be the maximum amount of threads +that can be created. +*/ +#define MAX_JOBS 1024 + +#define RUN_MEM_TEST 0 + +#include "EtcConfig.h" + +#include "Etc.h" + +#include "EtcTool.h" +#include "EtcSourceImage.h" +#include "EtcFile.h" +#include "EtcMath.h" +#include "EtcImage.h" +#include "EtcErrorMetric.h" +#include "EtcBlock4x4EncodingBits.h" + +#include "EtcAnalysis.h" + +#include +#include +#include + +using namespace Etc; + +#if ETC_WINDOWS +const char *ETC_MKDIR_COMMAND = "mkdir"; + +int strcasecmp(const char *s1, const char *s2) +{ + return _stricmp(s1, s2); +} +#else +const char *ETC_MKDIR_COMMAND = "mkdir -p"; +#endif + +#if RUN_MEM_TEST +#include "EtcMemTest.h" +#endif + +class Commands +{ +public: + + static const unsigned int MIN_JOBS = 8; + + Commands(void) + { + pstrSourceFilename = nullptr; + pstrOutputFilename = nullptr; + format = Image::Format::DEFAULT; + pstrAnalysisDirectory = nullptr; + uiComparisons = 0; + for (unsigned int uiComparison = 0; uiComparison < Analysis::MAX_COMPARISONS; uiComparison++) + { + apstrCompareFilename[uiComparison] = nullptr; + } + fEffort = 40.0f; + //Rec. 709 or BT.709...the default + e_ErrMetric = ErrorMetric::BT709; + uiJobs = MIN_JOBS; + + //these are ignored if they are < 0 + i_hPixel = -1; + i_vPixel = -1; + verboseOutput = false; + boolNormalizeXYZ = false; + } + + bool ProcessCommandLineArguments(int a_iArgs, const char *a_apstrArgs[]); + void PrintUsageMessage(void); + static void FixSlashes(char *a_pstr); + + char *pstrSourceFilename; + char *pstrOutputFilename; + + Image::Format format; + char *pstrAnalysisDirectory; + char *formatType; + unsigned int uiComparisons; + char *apstrCompareFilename[Analysis::MAX_COMPARISONS]; + float fEffort; + ErrorMetric e_ErrMetric; + unsigned int uiJobs; // for threading + bool verboseOutput; + //when both of these are >= 0 then single block mode is on + int i_hPixel; + int i_vPixel; + bool boolNormalizeXYZ; +}; + +#include "EtcFileHeader.h" + +// ---------------------------------------------------------------------------------------------------- +// +int main(int argc, const char * argv[]) +{ + + static const bool USE_C_INTERFACE = false; + + // this code tests for memory leaks +#if RUN_MEM_TEST + RunMemTest(true, 100); + printf("an extra line to see how the memory is free'd\n"); + printf("all done!\n"); + exit(0); +#endif + + Commands commands; + bool boolPrintUsage = commands.ProcessCommandLineArguments(argc, argv); + if (boolPrintUsage) + { + commands.PrintUsageMessage(); + exit(1); + } + + if (commands.verboseOutput) + { + printf("SourceImage: %s\n", commands.pstrSourceFilename); + } + SourceImage sourceimage(commands.pstrSourceFilename, commands.i_hPixel, commands.i_vPixel); + if (commands.boolNormalizeXYZ) + { + sourceimage.NormalizeXYZ(); + } + + unsigned int uiSourceWidth = sourceimage.GetWidth(); + unsigned int uiSourceHeight = sourceimage.GetHeight(); + + if (USE_C_INTERFACE) + { + unsigned char *paucEncodingBits; + unsigned int uiEncodingBitsBytes; + unsigned int uiExtendedWidth; + unsigned int uiExtendedHeight; + int iEncodingTime_ms; + + if (commands.verboseOutput) + { + printf("Encoding:\n"); + printf(" effort = %.f\n", commands.fEffort); + printf(" encoding = %s\n", Image::EncodingFormatToString(commands.format)); + printf(" error metric: %s\n", ErrorMetricToString(commands.e_ErrMetric)); + } + Etc::Encode((float *)sourceimage.GetPixels(), + uiSourceWidth, uiSourceHeight, + commands.format, + commands.e_ErrMetric, + commands.fEffort, + commands.uiJobs, + MAX_JOBS, + &paucEncodingBits, &uiEncodingBitsBytes, + &uiExtendedWidth, &uiExtendedHeight, + &iEncodingTime_ms); + if (commands.verboseOutput) + { + printf(" encode time = %dms\n", iEncodingTime_ms); + printf("EncodedImage: %s\n", commands.pstrOutputFilename); + } + Etc::File etcfile(commands.pstrOutputFilename, Etc::File::Format::INFER_FROM_FILE_EXTENSION, + commands.format, + paucEncodingBits, uiEncodingBitsBytes, + uiSourceWidth, uiSourceHeight, + uiExtendedWidth, uiExtendedHeight); + etcfile.Write(); + } + else + { + if (commands.verboseOutput) + { + printf("Encoding:\n"); + printf(" effort = %.f%%\n", commands.fEffort); + printf(" encoding = %s\n", Image::EncodingFormatToString(commands.format)); + printf(" error metric: %s\n", ErrorMetricToString(commands.e_ErrMetric)); + } + Etc::Image image((float *)sourceimage.GetPixels(), + uiSourceWidth, uiSourceHeight, + commands.e_ErrMetric); + image.m_bVerboseOutput = commands.verboseOutput; + Etc::Image::EncodingStatus encStatus = Etc::Image::EncodingStatus::SUCCESS; + + encStatus = image.Encode(commands.format, commands.e_ErrMetric, commands.fEffort, commands.uiJobs,MAX_JOBS); + if (commands.verboseOutput) + { + printf(" encode time = %dms\n", image.GetEncodingTimeMs()); + printf("EncodedImage: %s\n", commands.pstrOutputFilename); + printf("status bitfield: %u\n", encStatus); + } + Etc::File etcfile(commands.pstrOutputFilename, Etc::File::Format::INFER_FROM_FILE_EXTENSION, + commands.format, + image.GetEncodingBits(), image.GetEncodingBitsBytes(), + image.GetSourceWidth(), image.GetSourceHeight(), + image.GetExtendedWidth(), image.GetExtendedHeight()); + + etcfile.Write(); + + if (commands.pstrAnalysisDirectory) + { + if (commands.verboseOutput) + { + printf("Analysis: %s\n", commands.pstrAnalysisDirectory); + } + Analysis analysis(&image, commands.pstrAnalysisDirectory); + + for (unsigned int uiComparison = 0; uiComparison < commands.uiComparisons; uiComparison++) + { + analysis.Compare(commands.apstrCompareFilename[uiComparison], commands.i_hPixel, commands.i_vPixel); + } + } + + } + + return 0; +} + +// ---------------------------------------------------------------------------------------------------- +// return true if usage message should be printed +// +bool Commands::ProcessCommandLineArguments(int a_iArgs, const char *a_apstrArgs[]) +{ + static const bool DEBUG_PRINT = false; + + if (a_iArgs == 1) + { + printf("Error: missing arguments\n"); + return true; + } + + for (int iArg = 1; iArg < a_iArgs; iArg++) + { + if (DEBUG_PRINT) + { + printf("%s: %u %s\n", a_apstrArgs[0], iArg, a_apstrArgs[iArg]); + } + + if (strcmp(a_apstrArgs[iArg], "-analyze") == 0) + { + ++iArg; + + if (iArg >= (a_iArgs)) + { + printf("Error: missing folder parameter for -analyze\n"); + return true; + } + else + { + pstrAnalysisDirectory = new char[strlen(a_apstrArgs[iArg]) + 1]; + strcpy(pstrAnalysisDirectory, a_apstrArgs[iArg]); + FixSlashes(pstrAnalysisDirectory); + } + } + else if (strcmp(a_apstrArgs[iArg], "-argfile") == 0) + { + static const unsigned int MAX_LINE_CHARS = 1000; + static const unsigned int MAX_ARGFILE_ARGS = 100; + + ++iArg; + + if (iArg >= (a_iArgs)) + { + printf("Error: missing file parameter for -argfile\n"); + return true; + } + else + { + FILE *pfile = fopen(a_apstrArgs[iArg], "rt"); + if (pfile == nullptr) + { + printf("Error: couldn't open argfile (%s)\n", a_apstrArgs[iArg]); + return true; + } + + char **apstrArgs = new char *[MAX_ARGFILE_ARGS]; + assert(apstrArgs); + + // add null executable name + apstrArgs[0] = const_cast(""); + int iArgs = 1; + + // read in tokens + { + char *pcTokens = new char[MAX_LINE_CHARS + 1]; + assert(pcTokens); + char *pcToken = nullptr; + + // read in each line + while (fgets(pcTokens, MAX_LINE_CHARS, pfile)) + { + // skip over lines with '#' in first char + if (pcTokens[0] == '#') + { + continue; + } + + // abort remainder of argfile with '%' in first char + if (pcTokens[0] == '%') + { + break; + } + pcToken = strtok(pcTokens, " \n\r"); + + if (pcToken != nullptr) + { + apstrArgs[iArgs] = new char[strlen(pcToken) + 1]; + strcpy(apstrArgs[iArgs], pcToken); + iArgs++; + } + + while (pcToken != nullptr) + { + pcToken = strtok(nullptr, " \n"); + if (pcToken != nullptr) + { + apstrArgs[iArgs] = new char[strlen(pcToken) + 1]; + strcpy(apstrArgs[iArgs], pcToken); + iArgs++; + } + } + } + + delete[] pcTokens; + } + + fclose(pfile); + + bool boolErrors = ProcessCommandLineArguments(iArgs, const_cast(apstrArgs)); + + for (iArg = 1; iArg < iArgs; iArg++) + { + delete[] apstrArgs[iArg]; + } + delete[] apstrArgs; + + if (boolErrors) + { + return true; + } + } + } + //used for debugging...select a single block to encode + //supply the horiz and very pos of the block + else if (strcmp(a_apstrArgs[iArg], "-blockAtHV") == 0) + { + ++iArg; + + //make sure we have two more args after -block + if (iArg + 1 >= (a_iArgs)) + { + printf("Error: missing horiz and vert position of pixel for single block mode \n"); + return true; + } + i_hPixel = atoi(a_apstrArgs[iArg]); + ++iArg; + i_vPixel = atoi(a_apstrArgs[iArg]); + } + else if (strcmp(a_apstrArgs[iArg], "-compare") == 0) + { + ++iArg; + + if (iArg >= (a_iArgs)) + { + printf("Error: missing comprison_image parameter for -compare\n"); + return true; + } + else + { + if (uiComparisons >= Analysis::MAX_COMPARISONS) + { + printf("Error: too many comparisons\n"); + return true; + } + + char **ppstrCompareFilename = &apstrCompareFilename[uiComparisons++]; + + *ppstrCompareFilename = new char[strlen(a_apstrArgs[iArg]) + 1]; + strcpy(*ppstrCompareFilename, a_apstrArgs[iArg]); + FixSlashes(*ppstrCompareFilename); + } + } + else if (strcmp(a_apstrArgs[iArg], "-effort") == 0) + { + ++iArg; + + if (iArg >= (a_iArgs)) + { + printf("Error: missing amount parameter for -effort\n"); + return true; + } + else + { + float f; + int iScans = sscanf(a_apstrArgs[iArg], "%f", &f); + + if (iScans != 1) + { + printf("Error: couldn't parse amount for -effort (%s)\n", a_apstrArgs[iArg]); + return true; + } + else + { + fEffort = f; + } + } + } + else if (strcmp(a_apstrArgs[iArg], "-errormetric") == 0) + { + ++iArg; + + if (iArg >= (a_iArgs)) + { + printf("Error: missing error metric type %s\n", a_apstrArgs[iArg]); + return true; + } + else + { + if (strcmp(a_apstrArgs[iArg], "rgba") == 0) + { + e_ErrMetric = ErrorMetric::RGBA; + } + else if (strcmp(a_apstrArgs[iArg], "rec709") == 0) + { + e_ErrMetric = ErrorMetric::REC709; + } + else if (strcmp(a_apstrArgs[iArg], "numeric") == 0) + { + e_ErrMetric = ErrorMetric::NUMERIC; + } + else if (strcmp(a_apstrArgs[iArg], "normalxyz") == 0 || + strcmp(a_apstrArgs[iArg], "normalXYZ") == 0) + { + e_ErrMetric = ErrorMetric::NORMALXYZ; + } + else + { + printf("unrecognized error metric (%s), using numeric\n", a_apstrArgs[iArg]); + e_ErrMetric = ErrorMetric::NUMERIC; + } + } + } + else if (strcmp(a_apstrArgs[iArg], "-format") == 0) + { + ++iArg; + + if (iArg >= (a_iArgs)) + { + printf("Error: missing etc_format parameter for -format\n"); + return true; + } + else + { + formatType = new char[strlen(a_apstrArgs[iArg])+1]; + strcpy(formatType,a_apstrArgs[iArg]); + if (strcmp(a_apstrArgs[iArg], "ETC1") == 0) + { + format = Image::Format::ETC1; + } + else if (strcmp(a_apstrArgs[iArg], "RGB8") == 0) + { + format = Image::Format::RGB8; + } + else if (strcmp(a_apstrArgs[iArg], "SRGB8") == 0) + { + format = Image::Format::SRGB8; + } + else if (strcmp(a_apstrArgs[iArg], "RGBA8") == 0) + { + format = Image::Format::RGBA8; + } + else if (strcmp(a_apstrArgs[iArg], "SRGBA8") == 0) + { + format = Image::Format::SRGBA8; + } + else if (strcmp(a_apstrArgs[iArg], "R11") == 0) + { + format = Image::Format::R11; + } + else if (strcmp(a_apstrArgs[iArg], "SIGNED_R11") == 0) + { + format = Image::Format::SIGNED_R11; + } + else if (strcmp(a_apstrArgs[iArg], "RG11") == 0) + { + format = Image::Format::RG11; + } + else if (strcmp(a_apstrArgs[iArg], "SIGNED_RG11") == 0) + { + format = Image::Format::SIGNED_RG11; + } + else if (strcmp(a_apstrArgs[iArg], "RGB8A1") == 0) + { + format = Image::Format::RGB8A1; + } + else if (strcmp(a_apstrArgs[iArg], "SRGB8A1") == 0) + { + format = Image::Format::SRGB8A1; + } + else + { + printf("Error: unknown etc_format parameter for -format\n"); + format = Image::Format::UNKNOWN; + return true; + } + } + } + else if (strcmp(a_apstrArgs[iArg], "-help") == 0) + { + return true; + } + else if (strcmp(a_apstrArgs[iArg], "-j") == 0 || + strcmp(a_apstrArgs[iArg], "-jobs") == 0) + { + ++iArg; + + if (iArg >= (a_iArgs)) + { + printf("Error: missing job count for %s\n", a_apstrArgs[iArg]); + return true; + } + else + { + unsigned int ui; + int iScans = sscanf(a_apstrArgs[iArg], "%u", &ui); + + if (iScans != 1) + { + printf("Error: couldn't parse job count for %s (%s)\n", a_apstrArgs[iArg-1], a_apstrArgs[iArg]); + return true; + } + else + { + if (ui < MIN_JOBS) + { + ui = MIN_JOBS; + } + + uiJobs = ui; + } + } + } + else if (strcmp(a_apstrArgs[iArg], "-normalizexyz") == 0 || + strcmp(a_apstrArgs[iArg], "-normalizeXYZ") == 0) + { + boolNormalizeXYZ = true; + } + else if (strcmp(a_apstrArgs[iArg], "-output") == 0) + { + ++iArg; + + if (iArg >= (a_iArgs)) + { + printf("Error: missing encoded_image parameter for -output\n"); + return true; + } + else + { + pstrOutputFilename = new char[strlen(a_apstrArgs[iArg]) + 1]; + strcpy(pstrOutputFilename, a_apstrArgs[iArg]); + //take the output file name and extract the directory path so we can create the directory if nescacary + char *ptrOutputDir = nullptr; + + FixSlashes(pstrOutputFilename); + for (int c = (int)strlen(pstrOutputFilename); c > 0; c--) + { + //find the last slash, to get the name of the directory + if (pstrOutputFilename[c] == ETC_PATH_SLASH) + { + c++; + ptrOutputDir = new char[c]; + strncpy(ptrOutputDir, pstrOutputFilename, c); + ptrOutputDir[c] = '\0'; + CreateNewDir(ptrOutputDir); + break; + } + } + + if (ptrOutputDir == nullptr) + { + printf("couldnt find a place to put converted images\n"); + exit(1); + } + } + } + else if (strcmp(a_apstrArgs[iArg], "-verbose") == 0 || + strcmp(a_apstrArgs[iArg], "-v") == 0) + { + verboseOutput = true; + } + + else if (a_apstrArgs[iArg][0] == '-') + { + printf("Error: unknown option (%s)\n", a_apstrArgs[iArg]); + return true; + } + else if (a_apstrArgs[iArg][0] == '\r') + { + continue; + } + else + { + if (pstrSourceFilename != nullptr) + { + printf("Error: only support one source_image (%s)\n", a_apstrArgs[iArg]); + return true; + } + + pstrSourceFilename = new char[strlen(a_apstrArgs[iArg])+1]; + strcpy(pstrSourceFilename, a_apstrArgs[iArg]); + } + } + + if (pstrSourceFilename == nullptr) + { + printf("Error: missing source_image\n"); + return true; + } + + if (pstrOutputFilename == nullptr) + { + printf("Error: missing -output encoded_image\n"); + return true; + } + + if (uiComparisons > 0 && pstrAnalysisDirectory == nullptr) + { + printf("Error: -compare is only valid with -analyze\n"); + return true; + } + + return false; +} + +// ---------------------------------------------------------------------------------------------------- +// +void Commands::FixSlashes(char *a_pstr) +{ + while (*a_pstr) + { + if (*a_pstr == ETC_BAD_PATH_SLASH) + { + *a_pstr = ETC_PATH_SLASH; + } + + a_pstr++; + } +} + +// ---------------------------------------------------------------------------------------------------- +// print usage message and exit +// +void Commands::PrintUsageMessage(void) +{ + printf("Usage: etctool.exe source_image [options ...] -output \n"); + printf("Options:\n"); + printf(" -analyze \n"); + printf(" -argfile additional command line arguments\n"); + printf(" -blockAtHV encodes a single block that contains the\n"); + printf(" pixel specified by the H V coordinates\n"); + printf(" -compare compares source_image to comparison_image\n"); + printf(" -effort number between 0 and 100\n"); + printf(" -errormetric specify the error metric, the options are\n"); + printf(" rgba, rec709, numeric and normalxyz\n"); + printf(" -format ETC1, RGB8, SRGB8, RGBA8, SRGB8, RGB8A1,\n"); + printf(" SRGB8A1 or R11\n"); + printf(" -help prints this message\n"); + printf(" -jobs or -j specifies the number of threads (default=1)\n"); + printf(" -normalizexyz normalize RGB to have a length of 1\n"); + printf(" -verbose or -v shows status information during the encoding\n"); + printf(" process\n"); + printf("\n"); + + exit(1); +} + + // ---------------------------------------------------------------------------------------------------- + // + void CreateNewDir(const char *path) + { + char strCommand[300]; + +#if ETC_WINDOWS + sprintf_s(strCommand, "if not exist %s %s %s", path, ETC_MKDIR_COMMAND, path); +#else + sprintf(strCommand, "%s %s", ETC_MKDIR_COMMAND, path); +#endif + int iResult = system(strCommand); + if (iResult != 0) + { + printf("Error: couldn't create directory (%s)\n", path); + exit(0); + } + + } + + // ---------------------------------------------------------------------------------------------------- + // diff --git a/EtcTool/EtcTool.h b/EtcTool/EtcTool.h new file mode 100644 index 0000000..67cfccd --- /dev/null +++ b/EtcTool/EtcTool.h @@ -0,0 +1,36 @@ +/* + * Copyright 2015 The Etc2Comp Authors. + * + * Licensed under the Apache License, Version 2.0 (the "License"); + * you may not use this file except in compliance with the License. + * You may obtain a copy of the License at + * + * http://www.apache.org/licenses/LICENSE-2.0 + * + * Unless required by applicable law or agreed to in writing, software + * distributed under the License is distributed on an "AS IS" BASIS, + * WITHOUT WARRANTIES OR CONDITIONS OF ANY KIND, either express or implied. + * See the License for the specific language governing permissions and + * limitations under the License. + */ + +#pragma once + +#include "EtcConfig.h" + +#if ETC_WINDOWS + const char ETC_PATH_SLASH = '\\'; + const char ETC_BAD_PATH_SLASH = '/'; + + extern const char *ETC_MKDIR_COMMAND; + extern const char *ETC_IF_DIR_NOT_EXIST_COMMAND; + + int strcasecmp(const char *s1, const char *s2); +#else + const char ETC_PATH_SLASH = '/'; + const char ETC_BAD_PATH_SLASH = '\\'; + + extern const char *ETC_MKDIR_COMMAND; +#endif + + void CreateNewDir(const char *path); diff --git a/LICENSE b/LICENSE index d645695..75b5248 100644 --- a/LICENSE +++ b/LICENSE @@ -1,202 +1,202 @@ - - Apache License - Version 2.0, January 2004 - http://www.apache.org/licenses/ - - TERMS AND CONDITIONS FOR USE, REPRODUCTION, AND DISTRIBUTION - - 1. Definitions. - - "License" shall mean the terms and conditions for use, reproduction, - and distribution as defined by Sections 1 through 9 of this document. - - "Licensor" shall mean the copyright owner or entity authorized by - the copyright owner that is granting the License. - - "Legal Entity" shall mean the union of the acting entity and all - other entities that control, are controlled by, or are under common - control with that entity. For the purposes of this definition, - "control" means (i) the power, direct or indirect, to cause the - direction or management of such entity, whether by contract or - otherwise, or (ii) ownership of fifty percent (50%) or more of the - outstanding shares, or (iii) beneficial ownership of such entity. - - "You" (or "Your") shall mean an individual or Legal Entity - exercising permissions granted by this License. - - "Source" form shall mean the preferred form for making modifications, - including but not limited to software source code, documentation - source, and configuration files. - - "Object" form shall mean any form resulting from mechanical - transformation or translation of a Source form, including but - not limited to compiled object code, generated documentation, - and conversions to other media types. - - "Work" shall mean the work of authorship, whether in Source or - Object form, made available under the License, as indicated by a - copyright notice that is included in or attached to the work - (an example is provided in the Appendix below). - - "Derivative Works" shall mean any work, whether in Source or Object - form, that is based on (or derived from) the Work and for which the - editorial revisions, annotations, elaborations, or other modifications - represent, as a whole, an original work of authorship. For the purposes - of this License, Derivative Works shall not include works that remain - separable from, or merely link (or bind by name) to the interfaces of, - the Work and Derivative Works thereof. - - "Contribution" shall mean any work of authorship, including - the original version of the Work and any modifications or additions - to that Work or Derivative Works thereof, that is intentionally - submitted to Licensor for inclusion in the Work by the copyright owner - or by an individual or Legal Entity authorized to submit on behalf of - the copyright owner. For the purposes of this definition, "submitted" - means any form of electronic, verbal, or written communication sent - to the Licensor or its representatives, including but not limited to - communication on electronic mailing lists, source code control systems, - and issue tracking systems that are managed by, or on behalf of, the - Licensor for the purpose of discussing and improving the Work, but - excluding communication that is conspicuously marked or otherwise - designated in writing by the copyright owner as "Not a Contribution." - - "Contributor" shall mean Licensor and any individual or Legal Entity - on behalf of whom a Contribution has been received by Licensor and - subsequently incorporated within the Work. - - 2. Grant of Copyright License. Subject to the terms and conditions of - this License, each Contributor hereby grants to You a perpetual, - worldwide, non-exclusive, no-charge, royalty-free, irrevocable - copyright license to reproduce, prepare Derivative Works of, - publicly display, publicly perform, sublicense, and distribute the - Work and such Derivative Works in Source or Object form. - - 3. Grant of Patent License. Subject to the terms and conditions of - this License, each Contributor hereby grants to You a perpetual, - worldwide, non-exclusive, no-charge, royalty-free, irrevocable - (except as stated in this section) patent license to make, have made, - use, offer to sell, sell, import, and otherwise transfer the Work, - where such license applies only to those patent claims licensable - by such Contributor that are necessarily infringed by their - Contribution(s) alone or by combination of their Contribution(s) - with the Work to which such Contribution(s) was submitted. If You - institute patent litigation against any entity (including a - cross-claim or counterclaim in a lawsuit) alleging that the Work - or a Contribution incorporated within the Work constitutes direct - or contributory patent infringement, then any patent licenses - granted to You under this License for that Work shall terminate - as of the date such litigation is filed. - - 4. Redistribution. You may reproduce and distribute copies of the - Work or Derivative Works thereof in any medium, with or without - modifications, and in Source or Object form, provided that You - meet the following conditions: - - (a) You must give any other recipients of the Work or - Derivative Works a copy of this License; and - - (b) You must cause any modified files to carry prominent notices - stating that You changed the files; and - - (c) You must retain, in the Source form of any Derivative Works - that You distribute, all copyright, patent, trademark, and - attribution notices from the Source form of the Work, - excluding those notices that do not pertain to any part of - the Derivative Works; and - - (d) If the Work includes a "NOTICE" text file as part of its - distribution, then any Derivative Works that You distribute must - include a readable copy of the attribution notices contained - within such NOTICE file, excluding those notices that do not - pertain to any part of the Derivative Works, in at least one - of the following places: within a NOTICE text file distributed - as part of the Derivative Works; within the Source form or - documentation, if provided along with the Derivative Works; or, - within a display generated by the Derivative Works, if and - wherever such third-party notices normally appear. The contents - of the NOTICE file are for informational purposes only and - do not modify the License. You may add Your own attribution - notices within Derivative Works that You distribute, alongside - or as an addendum to the NOTICE text from the Work, provided - that such additional attribution notices cannot be construed - as modifying the License. - - You may add Your own copyright statement to Your modifications and - may provide additional or different license terms and conditions - for use, reproduction, or distribution of Your modifications, or - for any such Derivative Works as a whole, provided Your use, - reproduction, and distribution of the Work otherwise complies with - the conditions stated in this License. - - 5. Submission of Contributions. Unless You explicitly state otherwise, - any Contribution intentionally submitted for inclusion in the Work - by You to the Licensor shall be under the terms and conditions of - this License, without any additional terms or conditions. - Notwithstanding the above, nothing herein shall supersede or modify - the terms of any separate license agreement you may have executed - with Licensor regarding such Contributions. - - 6. Trademarks. This License does not grant permission to use the trade - names, trademarks, service marks, or product names of the Licensor, - except as required for reasonable and customary use in describing the - origin of the Work and reproducing the content of the NOTICE file. - - 7. Disclaimer of Warranty. Unless required by applicable law or - agreed to in writing, Licensor provides the Work (and each - Contributor provides its Contributions) on an "AS IS" BASIS, - WITHOUT WARRANTIES OR CONDITIONS OF ANY KIND, either express or - implied, including, without limitation, any warranties or conditions - of TITLE, NON-INFRINGEMENT, MERCHANTABILITY, or FITNESS FOR A - PARTICULAR PURPOSE. You are solely responsible for determining the - appropriateness of using or redistributing the Work and assume any - risks associated with Your exercise of permissions under this License. - - 8. Limitation of Liability. In no event and under no legal theory, - whether in tort (including negligence), contract, or otherwise, - unless required by applicable law (such as deliberate and grossly - negligent acts) or agreed to in writing, shall any Contributor be - liable to You for damages, including any direct, indirect, special, - incidental, or consequential damages of any character arising as a - result of this License or out of the use or inability to use the - Work (including but not limited to damages for loss of goodwill, - work stoppage, computer failure or malfunction, or any and all - other commercial damages or losses), even if such Contributor - has been advised of the possibility of such damages. - - 9. Accepting Warranty or Additional Liability. While redistributing - the Work or Derivative Works thereof, You may choose to offer, - and charge a fee for, acceptance of support, warranty, indemnity, - or other liability obligations and/or rights consistent with this - License. However, in accepting such obligations, You may act only - on Your own behalf and on Your sole responsibility, not on behalf - of any other Contributor, and only if You agree to indemnify, - defend, and hold each Contributor harmless for any liability - incurred by, or claims asserted against, such Contributor by reason - of your accepting any such warranty or additional liability. - - END OF TERMS AND CONDITIONS - - APPENDIX: How to apply the Apache License to your work. - - To apply the Apache License to your work, attach the following - boilerplate notice, with the fields enclosed by brackets "[]" - replaced with your own identifying information. (Don't include - the brackets!) The text should be enclosed in the appropriate - comment syntax for the file format. We also recommend that a - file or class name and description of purpose be included on the - same "printed page" as the copyright notice for easier - identification within third-party archives. - - Copyright [yyyy] [name of copyright owner] - - Licensed under the Apache License, Version 2.0 (the "License"); - you may not use this file except in compliance with the License. - You may obtain a copy of the License at - - http://www.apache.org/licenses/LICENSE-2.0 - - Unless required by applicable law or agreed to in writing, software - distributed under the License is distributed on an "AS IS" BASIS, - WITHOUT WARRANTIES OR CONDITIONS OF ANY KIND, either express or implied. - See the License for the specific language governing permissions and - limitations under the License. + + Apache License + Version 2.0, January 2004 + http://www.apache.org/licenses/ + + TERMS AND CONDITIONS FOR USE, REPRODUCTION, AND DISTRIBUTION + + 1. Definitions. + + "License" shall mean the terms and conditions for use, reproduction, + and distribution as defined by Sections 1 through 9 of this document. + + "Licensor" shall mean the copyright owner or entity authorized by + the copyright owner that is granting the License. + + "Legal Entity" shall mean the union of the acting entity and all + other entities that control, are controlled by, or are under common + control with that entity. For the purposes of this definition, + "control" means (i) the power, direct or indirect, to cause the + direction or management of such entity, whether by contract or + otherwise, or (ii) ownership of fifty percent (50%) or more of the + outstanding shares, or (iii) beneficial ownership of such entity. + + "You" (or "Your") shall mean an individual or Legal Entity + exercising permissions granted by this License. + + "Source" form shall mean the preferred form for making modifications, + including but not limited to software source code, documentation + source, and configuration files. + + "Object" form shall mean any form resulting from mechanical + transformation or translation of a Source form, including but + not limited to compiled object code, generated documentation, + and conversions to other media types. + + "Work" shall mean the work of authorship, whether in Source or + Object form, made available under the License, as indicated by a + copyright notice that is included in or attached to the work + (an example is provided in the Appendix below). + + "Derivative Works" shall mean any work, whether in Source or Object + form, that is based on (or derived from) the Work and for which the + editorial revisions, annotations, elaborations, or other modifications + represent, as a whole, an original work of authorship. For the purposes + of this License, Derivative Works shall not include works that remain + separable from, or merely link (or bind by name) to the interfaces of, + the Work and Derivative Works thereof. + + "Contribution" shall mean any work of authorship, including + the original version of the Work and any modifications or additions + to that Work or Derivative Works thereof, that is intentionally + submitted to Licensor for inclusion in the Work by the copyright owner + or by an individual or Legal Entity authorized to submit on behalf of + the copyright owner. For the purposes of this definition, "submitted" + means any form of electronic, verbal, or written communication sent + to the Licensor or its representatives, including but not limited to + communication on electronic mailing lists, source code control systems, + and issue tracking systems that are managed by, or on behalf of, the + Licensor for the purpose of discussing and improving the Work, but + excluding communication that is conspicuously marked or otherwise + designated in writing by the copyright owner as "Not a Contribution." + + "Contributor" shall mean Licensor and any individual or Legal Entity + on behalf of whom a Contribution has been received by Licensor and + subsequently incorporated within the Work. + + 2. Grant of Copyright License. Subject to the terms and conditions of + this License, each Contributor hereby grants to You a perpetual, + worldwide, non-exclusive, no-charge, royalty-free, irrevocable + copyright license to reproduce, prepare Derivative Works of, + publicly display, publicly perform, sublicense, and distribute the + Work and such Derivative Works in Source or Object form. + + 3. Grant of Patent License. Subject to the terms and conditions of + this License, each Contributor hereby grants to You a perpetual, + worldwide, non-exclusive, no-charge, royalty-free, irrevocable + (except as stated in this section) patent license to make, have made, + use, offer to sell, sell, import, and otherwise transfer the Work, + where such license applies only to those patent claims licensable + by such Contributor that are necessarily infringed by their + Contribution(s) alone or by combination of their Contribution(s) + with the Work to which such Contribution(s) was submitted. If You + institute patent litigation against any entity (including a + cross-claim or counterclaim in a lawsuit) alleging that the Work + or a Contribution incorporated within the Work constitutes direct + or contributory patent infringement, then any patent licenses + granted to You under this License for that Work shall terminate + as of the date such litigation is filed. + + 4. Redistribution. You may reproduce and distribute copies of the + Work or Derivative Works thereof in any medium, with or without + modifications, and in Source or Object form, provided that You + meet the following conditions: + + (a) You must give any other recipients of the Work or + Derivative Works a copy of this License; and + + (b) You must cause any modified files to carry prominent notices + stating that You changed the files; and + + (c) You must retain, in the Source form of any Derivative Works + that You distribute, all copyright, patent, trademark, and + attribution notices from the Source form of the Work, + excluding those notices that do not pertain to any part of + the Derivative Works; and + + (d) If the Work includes a "NOTICE" text file as part of its + distribution, then any Derivative Works that You distribute must + include a readable copy of the attribution notices contained + within such NOTICE file, excluding those notices that do not + pertain to any part of the Derivative Works, in at least one + of the following places: within a NOTICE text file distributed + as part of the Derivative Works; within the Source form or + documentation, if provided along with the Derivative Works; or, + within a display generated by the Derivative Works, if and + wherever such third-party notices normally appear. The contents + of the NOTICE file are for informational purposes only and + do not modify the License. You may add Your own attribution + notices within Derivative Works that You distribute, alongside + or as an addendum to the NOTICE text from the Work, provided + that such additional attribution notices cannot be construed + as modifying the License. + + You may add Your own copyright statement to Your modifications and + may provide additional or different license terms and conditions + for use, reproduction, or distribution of Your modifications, or + for any such Derivative Works as a whole, provided Your use, + reproduction, and distribution of the Work otherwise complies with + the conditions stated in this License. + + 5. Submission of Contributions. Unless You explicitly state otherwise, + any Contribution intentionally submitted for inclusion in the Work + by You to the Licensor shall be under the terms and conditions of + this License, without any additional terms or conditions. + Notwithstanding the above, nothing herein shall supersede or modify + the terms of any separate license agreement you may have executed + with Licensor regarding such Contributions. + + 6. Trademarks. This License does not grant permission to use the trade + names, trademarks, service marks, or product names of the Licensor, + except as required for reasonable and customary use in describing the + origin of the Work and reproducing the content of the NOTICE file. + + 7. Disclaimer of Warranty. Unless required by applicable law or + agreed to in writing, Licensor provides the Work (and each + Contributor provides its Contributions) on an "AS IS" BASIS, + WITHOUT WARRANTIES OR CONDITIONS OF ANY KIND, either express or + implied, including, without limitation, any warranties or conditions + of TITLE, NON-INFRINGEMENT, MERCHANTABILITY, or FITNESS FOR A + PARTICULAR PURPOSE. You are solely responsible for determining the + appropriateness of using or redistributing the Work and assume any + risks associated with Your exercise of permissions under this License. + + 8. Limitation of Liability. In no event and under no legal theory, + whether in tort (including negligence), contract, or otherwise, + unless required by applicable law (such as deliberate and grossly + negligent acts) or agreed to in writing, shall any Contributor be + liable to You for damages, including any direct, indirect, special, + incidental, or consequential damages of any character arising as a + result of this License or out of the use or inability to use the + Work (including but not limited to damages for loss of goodwill, + work stoppage, computer failure or malfunction, or any and all + other commercial damages or losses), even if such Contributor + has been advised of the possibility of such damages. + + 9. Accepting Warranty or Additional Liability. While redistributing + the Work or Derivative Works thereof, You may choose to offer, + and charge a fee for, acceptance of support, warranty, indemnity, + or other liability obligations and/or rights consistent with this + License. However, in accepting such obligations, You may act only + on Your own behalf and on Your sole responsibility, not on behalf + of any other Contributor, and only if You agree to indemnify, + defend, and hold each Contributor harmless for any liability + incurred by, or claims asserted against, such Contributor by reason + of your accepting any such warranty or additional liability. + + END OF TERMS AND CONDITIONS + + APPENDIX: How to apply the Apache License to your work. + + To apply the Apache License to your work, attach the following + boilerplate notice, with the fields enclosed by brackets "[]" + replaced with your own identifying information. (Don't include + the brackets!) The text should be enclosed in the appropriate + comment syntax for the file format. We also recommend that a + file or class name and description of purpose be included on the + same "printed page" as the copyright notice for easier + identification within third-party archives. + + Copyright [yyyy] [name of copyright owner] + + Licensed under the Apache License, Version 2.0 (the "License"); + you may not use this file except in compliance with the License. + You may obtain a copy of the License at + + http://www.apache.org/licenses/LICENSE-2.0 + + Unless required by applicable law or agreed to in writing, software + distributed under the License is distributed on an "AS IS" BASIS, + WITHOUT WARRANTIES OR CONDITIONS OF ANY KIND, either express or implied. + See the License for the specific language governing permissions and + limitations under the License. diff --git a/README.md b/README.md index 20f0a04..3e657c4 100644 --- a/README.md +++ b/README.md @@ -1,9 +1,180 @@ -# Etc2Comp - Texture to ETC2 compressor - -Etc2Comp is a command line tool that converts textures (e.g. bitmaps) -into the ETC2 format. The tool is built with a focus on encoding performance -to reduce the amount of time required to compile asset heavy applications as -well as reduce overall application size. - -Important: This is not an official Google product. It is an experimental -library published as-is. +# Etc2Comp - Texture to ETC2 compressor + +Etc2Comp is a command line tool that converts textures (e.g. bitmaps) +into the [ETC2](https://en.wikipedia.org/wiki/Ericsson_Texture_Compression) +format. The tool is built with a focus on encoding performance +to reduce the amount of time required to compile asset heavy applications as +well as reduce overall application size. + +This repo provides source code that can be compiled into a binary. The +binary can then be used to convert textures to the ETC2 format. + +Important: This is not an official Google product. It is an experimental +library published as-is. Please see the CONTRIBUTORS.md file for information +about questions or issues. + +## Setup +This project uses [CMake](https://cmake.org/) to generate platform-specific +build files: + - Linux: make files + - OS X: Xcode workspace files + - Microsoft Windows: Visual Studio solution files + - Note: CMake supports other formats, but this doc only provides steps for + one of each platform for brevity. + +Refer to each platform's setup section to setup your environment and build +an Etc2Comp binary. Then skip to the usage section of this page for examples +of how to use the library. + +### Setup for OS X + build tested on this config: + OS X 10.9.5 i7 16GB RAM + Xcode 5.1.1 + cmake 3.2.3 + +Start by downloading and installing the following components if they are not +already installed on your development machine. + - *Xcode* version 5.1.1, or greater + - [CMake](https://cmake.org/download/) version 3.2.3, or greater + +To build the Etc2Comp binary: + 1. Open a *Terminal* window and navigate to the project directory. + 1. Run `mkdir build_xcode` + 1. Run `cd build_xcode` + 1. Run `cmake -G Xcode ../` + 1. Open *Xcode* and import the `build_xcode/EtcTest.xcodeproj` file. + 1. Open the Product menu and choose Build For -> Running. + 1. Once the build succeeds the binary located at `build_xcode/EtcTool/Debug/EtcTool` +can be executed. + +Optional +Xcode EtcTool ‘Run’ preferences +note: if the build_xcode/EtcTest.xcodeproj is manually deleted then some Xcode preferences +will need to be set by hand after cmake is run (these prefs are retained across +cmake updates if the .xcodeproj is not deleted/removed) + +1. Set the active scheme to ‘EtcTool’ +1. Edit the scheme +1. Select option ‘Run EtcTool’, then tab ‘Arguments’. +Add this launch argument: ‘-argfile ../../EtcTool/args.txt’ +1. Select tab ‘Options’ and set a custom working directory to: ‘$(SRCROOT)/Build_Xcode/EtcTool’ + +### SetUp for Windows + +1. Make a `build_vs2015` directory. +1. Invoke cmake to create the Visual Studio project. +1. open the 'EtcTest' solution +1. make the 'EtcTool' project the start up project +1. (optional) in the project properties, under 'Debugging ->command arguments' +add the argfile textfile thats included in the EtcTool directory. +example: -argfile C:\etc2\EtcTool\Args.txt + +### Setup For Linux +The Linux build was tested on this config: + Ubuntu desktop 14.04 + gcc/g++ 4.8 + cmake 2.8.12.2 + +1. Verify linux has cmake and C++-11 capable g++ installed +1. Open shell +1 Run `mkdir build_linux` +1 Run `cd build_linux` +1. Invoke cmake to create the make files. +1. Run ‘make’ +1. navigate to the newly created EtcTool directory ‘cd EtcTool’ +1. run the executable: ‘./EtcTool -argfile ../../EtcTool/args.txt’ + +Skip to the Usage section for more information about using the +tool. + +## Usage + +### Command Line Usage +EtcTool can be run from the command line with the following usage: + etctool.exe source_image [options ...] -output encoded_image + +The encoder will use an array of RGBA floats read from the source_image to create +an ETC1 or ETC2 encoded image in encoded_image. The RGBA floats should be in the +range [0:1]. + +Options: + + -analyze + -argfile additional command line arguments read from a file + -blockAtHV encodes a single block that contains the + pixel specified by the H V coordinates + -compare compares source_image to comparison_image + -effort number between 0 and 100 to specify the encoding quality + (100 is the highest quality) + -errormetric specify the error metric, the options are + rgba, rec709, numeric and normalxyz + -format ETC1, RGB8, SRGB8, RGBA8, SRGB8, RGB8A1, + SRGB8A1 or R11 + -help prints this message + -jobs or -j specifies the number of threads (default=1) + -normalizexyz normalize RGB to have a length of 1 + -verbose or -v shows status information during the encoding + process + +* -analyze will run an analysis of the encoding and place it in folder +"analysis_folder" (e.g. ../analysis/kodim05). within the analysis_folder, a folder +will be created with a name of the current date/time (e.g. 20151204_153306). this +date/time folder is used to compare encodings of the same texture over time. +within the date/time folder is a text file with several encoding stats and a 2x png +image showing the encoding mode for each 4x4 block. + +* -argfile allows additional command line arguments to be placed in a text file + +* -blockAtHV selects the 4x4 pixel subset of the source image at position (H,V). +This is mainly used for debugging + +* -compare compares the source image to the created encoded image. The encoding +will dictate what error analysis is used in the comparison. + +* -effort uses an "amount" between 0 and 100 to determine how much additional effort +to apply during the encoding. + +* -errormetric selects the fitting algorithm used by the encoder. "rgba" calculates +RMS error using RGB components that are weighted by A. "rec709" is similar to "rgba", +except the RGB components are also weighted according to Rec709. "numeric" +calculates RMS error using unweighted RGBA components. "normalize" calculates error +based on dot product and vector length for RGB and RMS error for A. + +* -help prints out the usage message + +* -jobs enables multi-threading to speed up image encoding + +* -normalizexyz normalizes the source RGB to have a length of 1. + +* -verbose shows information on the current encoding process. It will then display the +PSNR and time time it took to encode the image. + +Note: Path names can use slashes or backslashes. The tool will convert the +slashes to the appropriate polarity for the current platform. + + +## API + +The library supports two different APIs - a C-like API that is not heavily +class-based and a class-based API. + +main() in EtcTool.cpp contains an example of both APIs. + +The Encode() method now returns an EncodingStatus that contains bit flags for +reporting various warnings and flags encountered when encoding. + + +## Copyright +Copyright 2015 Etc2Comp Authors. + +Licensed under the Apache License, Version 2.0 (the "License"); +you may not use this file except in compliance with the License. +You may obtain a copy of the License at + + http://www.apache.org/licenses/LICENSE-2.0 + +Unless required by applicable law or agreed to in writing, software +distributed under the License is distributed on an "AS IS" BASIS, +WITHOUT WARRANTIES OR CONDITIONS OF ANY KIND, either express or implied. +See the License for the specific language governing permissions and +limitations under the License. diff --git a/third_party/lodepng/LICENSE b/third_party/lodepng/LICENSE new file mode 100644 index 0000000..9382c4d --- /dev/null +++ b/third_party/lodepng/LICENSE @@ -0,0 +1,25 @@ +LodePNG version 20160124 + +Copyright (c) 2005-2016 Lode Vandevenne + +This software is provided 'as-is', without any express or implied +warranty. In no event will the authors be held liable for any damages +arising from the use of this software. + +Permission is granted to anyone to use this software for any purpose, +including commercial applications, and to alter it and redistribute it +freely, subject to the following restrictions: + + 1. The origin of this software must not be misrepresented; you must not + claim that you wrote the original software. If you use this software + in a product, an acknowledgment in the product documentation would be + appreciated but is not required. + + 2. Altered source versions must be plainly marked as such, and must not be + misrepresented as being the original software. + + 3. This notice may not be removed or altered from any source + distribution. + +The manual and changelog are in the header file "lodepng.h" +Rename this file to lodepng.cpp to use it for C++, or to lodepng.c to use it for C. diff --git a/third_party/lodepng/lodepng.cpp b/third_party/lodepng/lodepng.cpp new file mode 100644 index 0000000..8be7f59 --- /dev/null +++ b/third_party/lodepng/lodepng.cpp @@ -0,0 +1,6168 @@ +/* +LodePNG version 20160124 + +Copyright (c) 2005-2016 Lode Vandevenne + +This software is provided 'as-is', without any express or implied +warranty. In no event will the authors be held liable for any damages +arising from the use of this software. + +Permission is granted to anyone to use this software for any purpose, +including commercial applications, and to alter it and redistribute it +freely, subject to the following restrictions: + + 1. The origin of this software must not be misrepresented; you must not + claim that you wrote the original software. If you use this software + in a product, an acknowledgment in the product documentation would be + appreciated but is not required. + + 2. Altered source versions must be plainly marked as such, and must not be + misrepresented as being the original software. + + 3. This notice may not be removed or altered from any source + distribution. +*/ + +/* +The manual and changelog are in the header file "lodepng.h" +Rename this file to lodepng.cpp to use it for C++, or to lodepng.c to use it for C. +*/ + +#include "lodepng.h" + +#include +#include + +#ifdef LODEPNG_COMPILE_CPP +#include +#endif /*LODEPNG_COMPILE_CPP*/ + +#if defined(_MSC_VER) && (_MSC_VER >= 1310) /*Visual Studio: A few warning types are not desired here.*/ +#pragma warning( disable : 4244 ) /*implicit conversions: not warned by gcc -Wall -Wextra and requires too much casts*/ +#pragma warning( disable : 4996 ) /*VS does not like fopen, but fopen_s is not standard C so unusable here*/ +#endif /*_MSC_VER */ + +const char* LODEPNG_VERSION_STRING = "20160124"; + +/* +This source file is built up in the following large parts. The code sections +with the "LODEPNG_COMPILE_" #defines divide this up further in an intermixed way. +-Tools for C and common code for PNG and Zlib +-C Code for Zlib (huffman, deflate, ...) +-C Code for PNG (file format chunks, adam7, PNG filters, color conversions, ...) +-The C++ wrapper around all of the above +*/ + +/*The malloc, realloc and free functions defined here with "lodepng_" in front +of the name, so that you can easily change them to others related to your +platform if needed. Everything else in the code calls these. Pass +-DLODEPNG_NO_COMPILE_ALLOCATORS to the compiler, or comment out +#define LODEPNG_COMPILE_ALLOCATORS in the header, to disable the ones here and +define them in your own project's source files without needing to change +lodepng source code. Don't forget to remove "static" if you copypaste them +from here.*/ + +#ifdef LODEPNG_COMPILE_ALLOCATORS +static void* lodepng_malloc(size_t size) +{ + return malloc(size); +} + +static void* lodepng_realloc(void* ptr, size_t new_size) +{ + return realloc(ptr, new_size); +} + +static void lodepng_free(void* ptr) +{ + free(ptr); +} +#else /*LODEPNG_COMPILE_ALLOCATORS*/ +void* lodepng_malloc(size_t size); +void* lodepng_realloc(void* ptr, size_t new_size); +void lodepng_free(void* ptr); +#endif /*LODEPNG_COMPILE_ALLOCATORS*/ + +/* ////////////////////////////////////////////////////////////////////////// */ +/* ////////////////////////////////////////////////////////////////////////// */ +/* // Tools for C, and common code for PNG and Zlib. // */ +/* ////////////////////////////////////////////////////////////////////////// */ +/* ////////////////////////////////////////////////////////////////////////// */ + +/* +Often in case of an error a value is assigned to a variable and then it breaks +out of a loop (to go to the cleanup phase of a function). This macro does that. +It makes the error handling code shorter and more readable. + +Example: if(!uivector_resizev(&frequencies_ll, 286, 0)) ERROR_BREAK(83); +*/ +#define CERROR_BREAK(errorvar, code)\ +{\ + errorvar = code;\ + break;\ +} + +/*version of CERROR_BREAK that assumes the common case where the error variable is named "error"*/ +#define ERROR_BREAK(code) CERROR_BREAK(error, code) + +/*Set error var to the error code, and return it.*/ +#define CERROR_RETURN_ERROR(errorvar, code)\ +{\ + errorvar = code;\ + return code;\ +} + +/*Try the code, if it returns error, also return the error.*/ +#define CERROR_TRY_RETURN(call)\ +{\ + unsigned error = call;\ + if(error) return error;\ +} + +/*Set error var to the error code, and return from the void function.*/ +#define CERROR_RETURN(errorvar, code)\ +{\ + errorvar = code;\ + return;\ +} + +/* +About uivector, ucvector and string: +-All of them wrap dynamic arrays or text strings in a similar way. +-LodePNG was originally written in C++. The vectors replace the std::vectors that were used in the C++ version. +-The string tools are made to avoid problems with compilers that declare things like strncat as deprecated. +-They're not used in the interface, only internally in this file as static functions. +-As with many other structs in this file, the init and cleanup functions serve as ctor and dtor. +*/ + +#ifdef LODEPNG_COMPILE_ZLIB +/*dynamic vector of unsigned ints*/ +typedef struct uivector +{ + unsigned* data; + size_t size; /*size in number of unsigned longs*/ + size_t allocsize; /*allocated size in bytes*/ +} uivector; + +static void uivector_cleanup(void* p) +{ + ((uivector*)p)->size = ((uivector*)p)->allocsize = 0; + lodepng_free(((uivector*)p)->data); + ((uivector*)p)->data = NULL; +} + +/*returns 1 if success, 0 if failure ==> nothing done*/ +static unsigned uivector_reserve(uivector* p, size_t allocsize) +{ + if(allocsize > p->allocsize) + { + size_t newsize = (allocsize > p->allocsize * 2) ? allocsize : (allocsize * 3 / 2); + void* data = lodepng_realloc(p->data, newsize); + if(data) + { + p->allocsize = newsize; + p->data = (unsigned*)data; + } + else return 0; /*error: not enough memory*/ + } + return 1; +} + +/*returns 1 if success, 0 if failure ==> nothing done*/ +static unsigned uivector_resize(uivector* p, size_t size) +{ + if(!uivector_reserve(p, size * sizeof(unsigned))) return 0; + p->size = size; + return 1; /*success*/ +} + +/*resize and give all new elements the value*/ +static unsigned uivector_resizev(uivector* p, size_t size, unsigned value) +{ + size_t oldsize = p->size, i; + if(!uivector_resize(p, size)) return 0; + for(i = oldsize; i < size; ++i) p->data[i] = value; + return 1; +} + +static void uivector_init(uivector* p) +{ + p->data = NULL; + p->size = p->allocsize = 0; +} + +#ifdef LODEPNG_COMPILE_ENCODER +/*returns 1 if success, 0 if failure ==> nothing done*/ +static unsigned uivector_push_back(uivector* p, unsigned c) +{ + if(!uivector_resize(p, p->size + 1)) return 0; + p->data[p->size - 1] = c; + return 1; +} +#endif /*LODEPNG_COMPILE_ENCODER*/ +#endif /*LODEPNG_COMPILE_ZLIB*/ + +/* /////////////////////////////////////////////////////////////////////////// */ + +/*dynamic vector of unsigned chars*/ +typedef struct ucvector +{ + unsigned char* data; + size_t size; /*used size*/ + size_t allocsize; /*allocated size*/ +} ucvector; + +/*returns 1 if success, 0 if failure ==> nothing done*/ +static unsigned ucvector_reserve(ucvector* p, size_t allocsize) +{ + if(allocsize > p->allocsize) + { + size_t newsize = (allocsize > p->allocsize * 2) ? allocsize : (allocsize * 3 / 2); + void* data = lodepng_realloc(p->data, newsize); + if(data) + { + p->allocsize = newsize; + p->data = (unsigned char*)data; + } + else return 0; /*error: not enough memory*/ + } + return 1; +} + +/*returns 1 if success, 0 if failure ==> nothing done*/ +static unsigned ucvector_resize(ucvector* p, size_t size) +{ + if(!ucvector_reserve(p, size * sizeof(unsigned char))) return 0; + p->size = size; + return 1; /*success*/ +} + +#ifdef LODEPNG_COMPILE_PNG + +static void ucvector_cleanup(void* p) +{ + ((ucvector*)p)->size = ((ucvector*)p)->allocsize = 0; + lodepng_free(((ucvector*)p)->data); + ((ucvector*)p)->data = NULL; +} + +static void ucvector_init(ucvector* p) +{ + p->data = NULL; + p->size = p->allocsize = 0; +} +#endif /*LODEPNG_COMPILE_PNG*/ + +#ifdef LODEPNG_COMPILE_ZLIB +/*you can both convert from vector to buffer&size and vica versa. If you use +init_buffer to take over a buffer and size, it is not needed to use cleanup*/ +static void ucvector_init_buffer(ucvector* p, unsigned char* buffer, size_t size) +{ + p->data = buffer; + p->allocsize = p->size = size; +} +#endif /*LODEPNG_COMPILE_ZLIB*/ + +#if (defined(LODEPNG_COMPILE_PNG) && defined(LODEPNG_COMPILE_ANCILLARY_CHUNKS)) || defined(LODEPNG_COMPILE_ENCODER) +/*returns 1 if success, 0 if failure ==> nothing done*/ +static unsigned ucvector_push_back(ucvector* p, unsigned char c) +{ + if(!ucvector_resize(p, p->size + 1)) return 0; + p->data[p->size - 1] = c; + return 1; +} +#endif /*defined(LODEPNG_COMPILE_PNG) || defined(LODEPNG_COMPILE_ENCODER)*/ + + +/* ////////////////////////////////////////////////////////////////////////// */ + +#ifdef LODEPNG_COMPILE_PNG +#ifdef LODEPNG_COMPILE_ANCILLARY_CHUNKS +/*returns 1 if success, 0 if failure ==> nothing done*/ +static unsigned string_resize(char** out, size_t size) +{ + char* data = (char*)lodepng_realloc(*out, size + 1); + if(data) + { + data[size] = 0; /*null termination char*/ + *out = data; + } + return data != 0; +} + +/*init a {char*, size_t} pair for use as string*/ +static void string_init(char** out) +{ + *out = NULL; + string_resize(out, 0); +} + +/*free the above pair again*/ +static void string_cleanup(char** out) +{ + lodepng_free(*out); + *out = NULL; +} + +static void string_set(char** out, const char* in) +{ + size_t insize = strlen(in), i; + if(string_resize(out, insize)) + { + for(i = 0; i != insize; ++i) + { + (*out)[i] = in[i]; + } + } +} +#endif /*LODEPNG_COMPILE_ANCILLARY_CHUNKS*/ +#endif /*LODEPNG_COMPILE_PNG*/ + +/* ////////////////////////////////////////////////////////////////////////// */ + +unsigned lodepng_read32bitInt(const unsigned char* buffer) +{ + return (unsigned)((buffer[0] << 24) | (buffer[1] << 16) | (buffer[2] << 8) | buffer[3]); +} + +#if defined(LODEPNG_COMPILE_PNG) || defined(LODEPNG_COMPILE_ENCODER) +/*buffer must have at least 4 allocated bytes available*/ +static void lodepng_set32bitInt(unsigned char* buffer, unsigned value) +{ + buffer[0] = (unsigned char)((value >> 24) & 0xff); + buffer[1] = (unsigned char)((value >> 16) & 0xff); + buffer[2] = (unsigned char)((value >> 8) & 0xff); + buffer[3] = (unsigned char)((value ) & 0xff); +} +#endif /*defined(LODEPNG_COMPILE_PNG) || defined(LODEPNG_COMPILE_ENCODER)*/ + +#ifdef LODEPNG_COMPILE_ENCODER +static void lodepng_add32bitInt(ucvector* buffer, unsigned value) +{ + ucvector_resize(buffer, buffer->size + 4); /*todo: give error if resize failed*/ + lodepng_set32bitInt(&buffer->data[buffer->size - 4], value); +} +#endif /*LODEPNG_COMPILE_ENCODER*/ + +/* ////////////////////////////////////////////////////////////////////////// */ +/* / File IO / */ +/* ////////////////////////////////////////////////////////////////////////// */ + +#ifdef LODEPNG_COMPILE_DISK + +unsigned lodepng_load_file(unsigned char** out, size_t* outsize, const char* filename) +{ + FILE* file; + long size; + + /*provide some proper output values if error will happen*/ + *out = 0; + *outsize = 0; + + file = fopen(filename, "rb"); + if(!file) return 78; + + /*get filesize:*/ + fseek(file , 0 , SEEK_END); + size = ftell(file); + rewind(file); + + /*read contents of the file into the vector*/ + *outsize = 0; + *out = (unsigned char*)lodepng_malloc((size_t)size); + if(size && (*out)) (*outsize) = fread(*out, 1, (size_t)size, file); + + fclose(file); + if(!(*out) && size) return 83; /*the above malloc failed*/ + return 0; +} + +/*write given buffer to the file, overwriting the file, it doesn't append to it.*/ +unsigned lodepng_save_file(const unsigned char* buffer, size_t buffersize, const char* filename) +{ + FILE* file; + file = fopen(filename, "wb" ); + if(!file) return 79; + fwrite((char*)buffer , 1 , buffersize, file); + fclose(file); + return 0; +} + +#endif /*LODEPNG_COMPILE_DISK*/ + +/* ////////////////////////////////////////////////////////////////////////// */ +/* ////////////////////////////////////////////////////////////////////////// */ +/* // End of common code and tools. Begin of Zlib related code. // */ +/* ////////////////////////////////////////////////////////////////////////// */ +/* ////////////////////////////////////////////////////////////////////////// */ + +#ifdef LODEPNG_COMPILE_ZLIB +#ifdef LODEPNG_COMPILE_ENCODER +/*TODO: this ignores potential out of memory errors*/ +#define addBitToStream(/*size_t**/ bitpointer, /*ucvector**/ bitstream, /*unsigned char*/ bit)\ +{\ + /*add a new byte at the end*/\ + if(((*bitpointer) & 7) == 0) ucvector_push_back(bitstream, (unsigned char)0);\ + /*earlier bit of huffman code is in a lesser significant bit of an earlier byte*/\ + (bitstream->data[bitstream->size - 1]) |= (bit << ((*bitpointer) & 0x7));\ + ++(*bitpointer);\ +} + +static void addBitsToStream(size_t* bitpointer, ucvector* bitstream, unsigned value, size_t nbits) +{ + size_t i; + for(i = 0; i != nbits; ++i) addBitToStream(bitpointer, bitstream, (unsigned char)((value >> i) & 1)); +} + +static void addBitsToStreamReversed(size_t* bitpointer, ucvector* bitstream, unsigned value, size_t nbits) +{ + size_t i; + for(i = 0; i != nbits; ++i) addBitToStream(bitpointer, bitstream, (unsigned char)((value >> (nbits - 1 - i)) & 1)); +} +#endif /*LODEPNG_COMPILE_ENCODER*/ + +#ifdef LODEPNG_COMPILE_DECODER + +#define READBIT(bitpointer, bitstream) ((bitstream[bitpointer >> 3] >> (bitpointer & 0x7)) & (unsigned char)1) + +static unsigned char readBitFromStream(size_t* bitpointer, const unsigned char* bitstream) +{ + unsigned char result = (unsigned char)(READBIT(*bitpointer, bitstream)); + ++(*bitpointer); + return result; +} + +static unsigned readBitsFromStream(size_t* bitpointer, const unsigned char* bitstream, size_t nbits) +{ + unsigned result = 0, i; + for(i = 0; i != nbits; ++i) + { + result += ((unsigned)READBIT(*bitpointer, bitstream)) << i; + ++(*bitpointer); + } + return result; +} +#endif /*LODEPNG_COMPILE_DECODER*/ + +/* ////////////////////////////////////////////////////////////////////////// */ +/* / Deflate - Huffman / */ +/* ////////////////////////////////////////////////////////////////////////// */ + +#define FIRST_LENGTH_CODE_INDEX 257 +#define LAST_LENGTH_CODE_INDEX 285 +/*256 literals, the end code, some length codes, and 2 unused codes*/ +#define NUM_DEFLATE_CODE_SYMBOLS 288 +/*the distance codes have their own symbols, 30 used, 2 unused*/ +#define NUM_DISTANCE_SYMBOLS 32 +/*the code length codes. 0-15: code lengths, 16: copy previous 3-6 times, 17: 3-10 zeros, 18: 11-138 zeros*/ +#define NUM_CODE_LENGTH_CODES 19 + +/*the base lengths represented by codes 257-285*/ +static const unsigned LENGTHBASE[29] + = {3, 4, 5, 6, 7, 8, 9, 10, 11, 13, 15, 17, 19, 23, 27, 31, 35, 43, 51, 59, + 67, 83, 99, 115, 131, 163, 195, 227, 258}; + +/*the extra bits used by codes 257-285 (added to base length)*/ +static const unsigned LENGTHEXTRA[29] + = {0, 0, 0, 0, 0, 0, 0, 0, 1, 1, 1, 1, 2, 2, 2, 2, 3, 3, 3, 3, + 4, 4, 4, 4, 5, 5, 5, 5, 0}; + +/*the base backwards distances (the bits of distance codes appear after length codes and use their own huffman tree)*/ +static const unsigned DISTANCEBASE[30] + = {1, 2, 3, 4, 5, 7, 9, 13, 17, 25, 33, 49, 65, 97, 129, 193, 257, 385, 513, + 769, 1025, 1537, 2049, 3073, 4097, 6145, 8193, 12289, 16385, 24577}; + +/*the extra bits of backwards distances (added to base)*/ +static const unsigned DISTANCEEXTRA[30] + = {0, 0, 0, 0, 1, 1, 2, 2, 3, 3, 4, 4, 5, 5, 6, 6, 7, 7, 8, + 8, 9, 9, 10, 10, 11, 11, 12, 12, 13, 13}; + +/*the order in which "code length alphabet code lengths" are stored, out of this +the huffman tree of the dynamic huffman tree lengths is generated*/ +static const unsigned CLCL_ORDER[NUM_CODE_LENGTH_CODES] + = {16, 17, 18, 0, 8, 7, 9, 6, 10, 5, 11, 4, 12, 3, 13, 2, 14, 1, 15}; + +/* ////////////////////////////////////////////////////////////////////////// */ + +/* +Huffman tree struct, containing multiple representations of the tree +*/ +typedef struct HuffmanTree +{ + unsigned* tree2d; + unsigned* tree1d; + unsigned* lengths; /*the lengths of the codes of the 1d-tree*/ + unsigned maxbitlen; /*maximum number of bits a single code can get*/ + unsigned numcodes; /*number of symbols in the alphabet = number of codes*/ +} HuffmanTree; + +/*function used for debug purposes to draw the tree in ascii art with C++*/ +/* +static void HuffmanTree_draw(HuffmanTree* tree) +{ + std::cout << "tree. length: " << tree->numcodes << " maxbitlen: " << tree->maxbitlen << std::endl; + for(size_t i = 0; i != tree->tree1d.size; ++i) + { + if(tree->lengths.data[i]) + std::cout << i << " " << tree->tree1d.data[i] << " " << tree->lengths.data[i] << std::endl; + } + std::cout << std::endl; +}*/ + +static void HuffmanTree_init(HuffmanTree* tree) +{ + tree->tree2d = 0; + tree->tree1d = 0; + tree->lengths = 0; +} + +static void HuffmanTree_cleanup(HuffmanTree* tree) +{ + lodepng_free(tree->tree2d); + lodepng_free(tree->tree1d); + lodepng_free(tree->lengths); +} + +/*the tree representation used by the decoder. return value is error*/ +static unsigned HuffmanTree_make2DTree(HuffmanTree* tree) +{ + unsigned nodefilled = 0; /*up to which node it is filled*/ + unsigned treepos = 0; /*position in the tree (1 of the numcodes columns)*/ + unsigned n, i; + + tree->tree2d = (unsigned*)lodepng_malloc(tree->numcodes * 2 * sizeof(unsigned)); + if(!tree->tree2d) return 83; /*alloc fail*/ + + /* + convert tree1d[] to tree2d[][]. In the 2D array, a value of 32767 means + uninited, a value >= numcodes is an address to another bit, a value < numcodes + is a code. The 2 rows are the 2 possible bit values (0 or 1), there are as + many columns as codes - 1. + A good huffman tree has N * 2 - 1 nodes, of which N - 1 are internal nodes. + Here, the internal nodes are stored (what their 0 and 1 option point to). + There is only memory for such good tree currently, if there are more nodes + (due to too long length codes), error 55 will happen + */ + for(n = 0; n < tree->numcodes * 2; ++n) + { + tree->tree2d[n] = 32767; /*32767 here means the tree2d isn't filled there yet*/ + } + + for(n = 0; n < tree->numcodes; ++n) /*the codes*/ + { + for(i = 0; i != tree->lengths[n]; ++i) /*the bits for this code*/ + { + unsigned char bit = (unsigned char)((tree->tree1d[n] >> (tree->lengths[n] - i - 1)) & 1); + /*oversubscribed, see comment in lodepng_error_text*/ + if(treepos > 2147483647 || treepos + 2 > tree->numcodes) return 55; + if(tree->tree2d[2 * treepos + bit] == 32767) /*not yet filled in*/ + { + if(i + 1 == tree->lengths[n]) /*last bit*/ + { + tree->tree2d[2 * treepos + bit] = n; /*put the current code in it*/ + treepos = 0; + } + else + { + /*put address of the next step in here, first that address has to be found of course + (it's just nodefilled + 1)...*/ + ++nodefilled; + /*addresses encoded with numcodes added to it*/ + tree->tree2d[2 * treepos + bit] = nodefilled + tree->numcodes; + treepos = nodefilled; + } + } + else treepos = tree->tree2d[2 * treepos + bit] - tree->numcodes; + } + } + + for(n = 0; n < tree->numcodes * 2; ++n) + { + if(tree->tree2d[n] == 32767) tree->tree2d[n] = 0; /*remove possible remaining 32767's*/ + } + + return 0; +} + +/* +Second step for the ...makeFromLengths and ...makeFromFrequencies functions. +numcodes, lengths and maxbitlen must already be filled in correctly. return +value is error. +*/ +static unsigned HuffmanTree_makeFromLengths2(HuffmanTree* tree) +{ + uivector blcount; + uivector nextcode; + unsigned error = 0; + unsigned bits, n; + + uivector_init(&blcount); + uivector_init(&nextcode); + + tree->tree1d = (unsigned*)lodepng_malloc(tree->numcodes * sizeof(unsigned)); + if(!tree->tree1d) error = 83; /*alloc fail*/ + + if(!uivector_resizev(&blcount, tree->maxbitlen + 1, 0) + || !uivector_resizev(&nextcode, tree->maxbitlen + 1, 0)) + error = 83; /*alloc fail*/ + + if(!error) + { + /*step 1: count number of instances of each code length*/ + for(bits = 0; bits != tree->numcodes; ++bits) ++blcount.data[tree->lengths[bits]]; + /*step 2: generate the nextcode values*/ + for(bits = 1; bits <= tree->maxbitlen; ++bits) + { + nextcode.data[bits] = (nextcode.data[bits - 1] + blcount.data[bits - 1]) << 1; + } + /*step 3: generate all the codes*/ + for(n = 0; n != tree->numcodes; ++n) + { + if(tree->lengths[n] != 0) tree->tree1d[n] = nextcode.data[tree->lengths[n]]++; + } + } + + uivector_cleanup(&blcount); + uivector_cleanup(&nextcode); + + if(!error) return HuffmanTree_make2DTree(tree); + else return error; +} + +/* +given the code lengths (as stored in the PNG file), generate the tree as defined +by Deflate. maxbitlen is the maximum bits that a code in the tree can have. +return value is error. +*/ +static unsigned HuffmanTree_makeFromLengths(HuffmanTree* tree, const unsigned* bitlen, + size_t numcodes, unsigned maxbitlen) +{ + unsigned i; + tree->lengths = (unsigned*)lodepng_malloc(numcodes * sizeof(unsigned)); + if(!tree->lengths) return 83; /*alloc fail*/ + for(i = 0; i != numcodes; ++i) tree->lengths[i] = bitlen[i]; + tree->numcodes = (unsigned)numcodes; /*number of symbols*/ + tree->maxbitlen = maxbitlen; + return HuffmanTree_makeFromLengths2(tree); +} + +#ifdef LODEPNG_COMPILE_ENCODER + +/*BPM: Boundary Package Merge, see "A Fast and Space-Economical Algorithm for Length-Limited Coding", +Jyrki Katajainen, Alistair Moffat, Andrew Turpin, 1995.*/ + +/*chain node for boundary package merge*/ +typedef struct BPMNode +{ + int weight; /*the sum of all weights in this chain*/ + unsigned index; /*index of this leaf node (called "count" in the paper)*/ + struct BPMNode* tail; /*the next nodes in this chain (null if last)*/ + int in_use; +} BPMNode; + +/*lists of chains*/ +typedef struct BPMLists +{ + /*memory pool*/ + unsigned memsize; + BPMNode* memory; + unsigned numfree; + unsigned nextfree; + BPMNode** freelist; + /*two heads of lookahead chains per list*/ + unsigned listsize; + BPMNode** chains0; + BPMNode** chains1; +} BPMLists; + +/*creates a new chain node with the given parameters, from the memory in the lists */ +static BPMNode* bpmnode_create(BPMLists* lists, int weight, unsigned index, BPMNode* tail) +{ + unsigned i; + BPMNode* result; + + /*memory full, so garbage collect*/ + if(lists->nextfree >= lists->numfree) + { + /*mark only those that are in use*/ + for(i = 0; i != lists->memsize; ++i) lists->memory[i].in_use = 0; + for(i = 0; i != lists->listsize; ++i) + { + BPMNode* node; + for(node = lists->chains0[i]; node != 0; node = node->tail) node->in_use = 1; + for(node = lists->chains1[i]; node != 0; node = node->tail) node->in_use = 1; + } + /*collect those that are free*/ + lists->numfree = 0; + for(i = 0; i != lists->memsize; ++i) + { + if(!lists->memory[i].in_use) lists->freelist[lists->numfree++] = &lists->memory[i]; + } + lists->nextfree = 0; + } + + result = lists->freelist[lists->nextfree++]; + result->weight = weight; + result->index = index; + result->tail = tail; + return result; +} + +static int bpmnode_compare(const void* a, const void* b) +{ + int wa = ((const BPMNode*)a)->weight; + int wb = ((const BPMNode*)b)->weight; + if(wa < wb) return -1; + if(wa > wb) return 1; + /*make the qsort a stable sort*/ + return ((const BPMNode*)a)->index < ((const BPMNode*)b)->index ? 1 : -1; +} + +/*Boundary Package Merge step, numpresent is the amount of leaves, and c is the current chain.*/ +static void boundaryPM(BPMLists* lists, BPMNode* leaves, size_t numpresent, int c, int num) +{ + unsigned lastindex = lists->chains1[c]->index; + + if(c == 0) + { + if(lastindex >= numpresent) return; + lists->chains0[c] = lists->chains1[c]; + lists->chains1[c] = bpmnode_create(lists, leaves[lastindex].weight, lastindex + 1, 0); + } + else + { + /*sum of the weights of the head nodes of the previous lookahead chains.*/ + int sum = lists->chains0[c - 1]->weight + lists->chains1[c - 1]->weight; + lists->chains0[c] = lists->chains1[c]; + if(lastindex < numpresent && sum > leaves[lastindex].weight) + { + lists->chains1[c] = bpmnode_create(lists, leaves[lastindex].weight, lastindex + 1, lists->chains1[c]->tail); + return; + } + lists->chains1[c] = bpmnode_create(lists, sum, lastindex, lists->chains1[c - 1]); + /*in the end we are only interested in the chain of the last list, so no + need to recurse if we're at the last one (this gives measurable speedup)*/ + if(num + 1 < (int)(2 * numpresent - 2)) + { + boundaryPM(lists, leaves, numpresent, c - 1, num); + boundaryPM(lists, leaves, numpresent, c - 1, num); + } + } +} + +unsigned lodepng_huffman_code_lengths(unsigned* lengths, const unsigned* frequencies, + size_t numcodes, unsigned maxbitlen) +{ + unsigned error = 0; + unsigned i; + size_t numpresent = 0; /*number of symbols with non-zero frequency*/ + BPMNode* leaves; /*the symbols, only those with > 0 frequency*/ + + if(numcodes == 0) return 80; /*error: a tree of 0 symbols is not supposed to be made*/ + if((1ull << maxbitlen) < numcodes) return 80; /*error: represent all symbols*/ + + leaves = (BPMNode*)lodepng_malloc(numcodes * sizeof(*leaves)); + if(!leaves) return 83; /*alloc fail*/ + + for(i = 0; i != numcodes; ++i) + { + if(frequencies[i] > 0) + { + leaves[numpresent].weight = (int)frequencies[i]; + leaves[numpresent].index = i; + ++numpresent; + } + } + + for(i = 0; i != numcodes; ++i) lengths[i] = 0; + + /*ensure at least two present symbols. There should be at least one symbol + according to RFC 1951 section 3.2.7. Some decoders incorrectly require two. To + make these work as well ensure there are at least two symbols. The + Package-Merge code below also doesn't work correctly if there's only one + symbol, it'd give it the theoritical 0 bits but in practice zlib wants 1 bit*/ + if(numpresent == 0) + { + lengths[0] = lengths[1] = 1; /*note that for RFC 1951 section 3.2.7, only lengths[0] = 1 is needed*/ + } + else if(numpresent == 1) + { + lengths[leaves[0].index] = 1; + lengths[leaves[0].index == 0 ? 1 : 0] = 1; + } + else + { + BPMLists lists; + BPMNode* node; + + qsort(leaves, numpresent, sizeof(BPMNode), bpmnode_compare); + + lists.listsize = maxbitlen; + lists.memsize = 2 * maxbitlen * (maxbitlen + 1); + lists.nextfree = 0; + lists.numfree = lists.memsize; + lists.memory = (BPMNode*)lodepng_malloc(lists.memsize * sizeof(*lists.memory)); + lists.freelist = (BPMNode**)lodepng_malloc(lists.memsize * sizeof(BPMNode*)); + lists.chains0 = (BPMNode**)lodepng_malloc(lists.listsize * sizeof(BPMNode*)); + lists.chains1 = (BPMNode**)lodepng_malloc(lists.listsize * sizeof(BPMNode*)); + if(!lists.memory || !lists.freelist || !lists.chains0 || !lists.chains1) error = 83; /*alloc fail*/ + + if(!error) + { + for(i = 0; i != lists.memsize; ++i) lists.freelist[i] = &lists.memory[i]; + + bpmnode_create(&lists, leaves[0].weight, 1, 0); + bpmnode_create(&lists, leaves[1].weight, 2, 0); + + for(i = 0; i != lists.listsize; ++i) + { + lists.chains0[i] = &lists.memory[0]; + lists.chains1[i] = &lists.memory[1]; + } + + /*each boundaryPM call adds one chain to the last list, and we need 2 * numpresent - 2 chains.*/ + for(i = 2; i != 2 * numpresent - 2; ++i) boundaryPM(&lists, leaves, numpresent, (int)maxbitlen - 1, (int)i); + + for(node = lists.chains1[maxbitlen - 1]; node; node = node->tail) + { + for(i = 0; i != node->index; ++i) ++lengths[leaves[i].index]; + } + } + + lodepng_free(lists.memory); + lodepng_free(lists.freelist); + lodepng_free(lists.chains0); + lodepng_free(lists.chains1); + } + + lodepng_free(leaves); + return error; +} + +/*Create the Huffman tree given the symbol frequencies*/ +static unsigned HuffmanTree_makeFromFrequencies(HuffmanTree* tree, const unsigned* frequencies, + size_t mincodes, size_t numcodes, unsigned maxbitlen) +{ + unsigned error = 0; + while(!frequencies[numcodes - 1] && numcodes > mincodes) --numcodes; /*trim zeroes*/ + tree->maxbitlen = maxbitlen; + tree->numcodes = (unsigned)numcodes; /*number of symbols*/ + tree->lengths = (unsigned*)lodepng_realloc(tree->lengths, numcodes * sizeof(unsigned)); + if(!tree->lengths) return 83; /*alloc fail*/ + /*initialize all lengths to 0*/ + memset(tree->lengths, 0, numcodes * sizeof(unsigned)); + + error = lodepng_huffman_code_lengths(tree->lengths, frequencies, numcodes, maxbitlen); + if(!error) error = HuffmanTree_makeFromLengths2(tree); + return error; +} + +static unsigned HuffmanTree_getCode(const HuffmanTree* tree, unsigned index) +{ + return tree->tree1d[index]; +} + +static unsigned HuffmanTree_getLength(const HuffmanTree* tree, unsigned index) +{ + return tree->lengths[index]; +} +#endif /*LODEPNG_COMPILE_ENCODER*/ + +/*get the literal and length code tree of a deflated block with fixed tree, as per the deflate specification*/ +static unsigned generateFixedLitLenTree(HuffmanTree* tree) +{ + unsigned i, error = 0; + unsigned* bitlen = (unsigned*)lodepng_malloc(NUM_DEFLATE_CODE_SYMBOLS * sizeof(unsigned)); + if(!bitlen) return 83; /*alloc fail*/ + + /*288 possible codes: 0-255=literals, 256=endcode, 257-285=lengthcodes, 286-287=unused*/ + for(i = 0; i <= 143; ++i) bitlen[i] = 8; + for(i = 144; i <= 255; ++i) bitlen[i] = 9; + for(i = 256; i <= 279; ++i) bitlen[i] = 7; + for(i = 280; i <= 287; ++i) bitlen[i] = 8; + + error = HuffmanTree_makeFromLengths(tree, bitlen, NUM_DEFLATE_CODE_SYMBOLS, 15); + + lodepng_free(bitlen); + return error; +} + +/*get the distance code tree of a deflated block with fixed tree, as specified in the deflate specification*/ +static unsigned generateFixedDistanceTree(HuffmanTree* tree) +{ + unsigned i, error = 0; + unsigned* bitlen = (unsigned*)lodepng_malloc(NUM_DISTANCE_SYMBOLS * sizeof(unsigned)); + if(!bitlen) return 83; /*alloc fail*/ + + /*there are 32 distance codes, but 30-31 are unused*/ + for(i = 0; i != NUM_DISTANCE_SYMBOLS; ++i) bitlen[i] = 5; + error = HuffmanTree_makeFromLengths(tree, bitlen, NUM_DISTANCE_SYMBOLS, 15); + + lodepng_free(bitlen); + return error; +} + +#ifdef LODEPNG_COMPILE_DECODER + +/* +returns the code, or (unsigned)(-1) if error happened +inbitlength is the length of the complete buffer, in bits (so its byte length times 8) +*/ +static unsigned huffmanDecodeSymbol(const unsigned char* in, size_t* bp, + const HuffmanTree* codetree, size_t inbitlength) +{ + unsigned treepos = 0, ct; + for(;;) + { + if(*bp >= inbitlength) return (unsigned)(-1); /*error: end of input memory reached without endcode*/ + /* + decode the symbol from the tree. The "readBitFromStream" code is inlined in + the expression below because this is the biggest bottleneck while decoding + */ + ct = codetree->tree2d[(treepos << 1) + READBIT(*bp, in)]; + ++(*bp); + if(ct < codetree->numcodes) return ct; /*the symbol is decoded, return it*/ + else treepos = ct - codetree->numcodes; /*symbol not yet decoded, instead move tree position*/ + + if(treepos >= codetree->numcodes) return (unsigned)(-1); /*error: it appeared outside the codetree*/ + } +} +#endif /*LODEPNG_COMPILE_DECODER*/ + +#ifdef LODEPNG_COMPILE_DECODER + +/* ////////////////////////////////////////////////////////////////////////// */ +/* / Inflator (Decompressor) / */ +/* ////////////////////////////////////////////////////////////////////////// */ + +/*get the tree of a deflated block with fixed tree, as specified in the deflate specification*/ +static void getTreeInflateFixed(HuffmanTree* tree_ll, HuffmanTree* tree_d) +{ + /*TODO: check for out of memory errors*/ + generateFixedLitLenTree(tree_ll); + generateFixedDistanceTree(tree_d); +} + +/*get the tree of a deflated block with dynamic tree, the tree itself is also Huffman compressed with a known tree*/ +static unsigned getTreeInflateDynamic(HuffmanTree* tree_ll, HuffmanTree* tree_d, + const unsigned char* in, size_t* bp, size_t inlength) +{ + /*make sure that length values that aren't filled in will be 0, or a wrong tree will be generated*/ + unsigned error = 0; + unsigned n, HLIT, HDIST, HCLEN, i; + size_t inbitlength = inlength * 8; + + /*see comments in deflateDynamic for explanation of the context and these variables, it is analogous*/ + unsigned* bitlen_ll = 0; /*lit,len code lengths*/ + unsigned* bitlen_d = 0; /*dist code lengths*/ + /*code length code lengths ("clcl"), the bit lengths of the huffman tree used to compress bitlen_ll and bitlen_d*/ + unsigned* bitlen_cl = 0; + HuffmanTree tree_cl; /*the code tree for code length codes (the huffman tree for compressed huffman trees)*/ + + if((*bp) + 14 > (inlength << 3)) return 49; /*error: the bit pointer is or will go past the memory*/ + + /*number of literal/length codes + 257. Unlike the spec, the value 257 is added to it here already*/ + HLIT = readBitsFromStream(bp, in, 5) + 257; + /*number of distance codes. Unlike the spec, the value 1 is added to it here already*/ + HDIST = readBitsFromStream(bp, in, 5) + 1; + /*number of code length codes. Unlike the spec, the value 4 is added to it here already*/ + HCLEN = readBitsFromStream(bp, in, 4) + 4; + + if((*bp) + HCLEN * 3 > (inlength << 3)) return 50; /*error: the bit pointer is or will go past the memory*/ + + HuffmanTree_init(&tree_cl); + + while(!error) + { + /*read the code length codes out of 3 * (amount of code length codes) bits*/ + + bitlen_cl = (unsigned*)lodepng_malloc(NUM_CODE_LENGTH_CODES * sizeof(unsigned)); + if(!bitlen_cl) ERROR_BREAK(83 /*alloc fail*/); + + for(i = 0; i != NUM_CODE_LENGTH_CODES; ++i) + { + if(i < HCLEN) bitlen_cl[CLCL_ORDER[i]] = readBitsFromStream(bp, in, 3); + else bitlen_cl[CLCL_ORDER[i]] = 0; /*if not, it must stay 0*/ + } + + error = HuffmanTree_makeFromLengths(&tree_cl, bitlen_cl, NUM_CODE_LENGTH_CODES, 7); + if(error) break; + + /*now we can use this tree to read the lengths for the tree that this function will return*/ + bitlen_ll = (unsigned*)lodepng_malloc(NUM_DEFLATE_CODE_SYMBOLS * sizeof(unsigned)); + bitlen_d = (unsigned*)lodepng_malloc(NUM_DISTANCE_SYMBOLS * sizeof(unsigned)); + if(!bitlen_ll || !bitlen_d) ERROR_BREAK(83 /*alloc fail*/); + for(i = 0; i != NUM_DEFLATE_CODE_SYMBOLS; ++i) bitlen_ll[i] = 0; + for(i = 0; i != NUM_DISTANCE_SYMBOLS; ++i) bitlen_d[i] = 0; + + /*i is the current symbol we're reading in the part that contains the code lengths of lit/len and dist codes*/ + i = 0; + while(i < HLIT + HDIST) + { + unsigned code = huffmanDecodeSymbol(in, bp, &tree_cl, inbitlength); + if(code <= 15) /*a length code*/ + { + if(i < HLIT) bitlen_ll[i] = code; + else bitlen_d[i - HLIT] = code; + ++i; + } + else if(code == 16) /*repeat previous*/ + { + unsigned replength = 3; /*read in the 2 bits that indicate repeat length (3-6)*/ + unsigned value; /*set value to the previous code*/ + + if(i == 0) ERROR_BREAK(54); /*can't repeat previous if i is 0*/ + + if((*bp + 2) > inbitlength) ERROR_BREAK(50); /*error, bit pointer jumps past memory*/ + replength += readBitsFromStream(bp, in, 2); + + if(i < HLIT + 1) value = bitlen_ll[i - 1]; + else value = bitlen_d[i - HLIT - 1]; + /*repeat this value in the next lengths*/ + for(n = 0; n < replength; ++n) + { + if(i >= HLIT + HDIST) ERROR_BREAK(13); /*error: i is larger than the amount of codes*/ + if(i < HLIT) bitlen_ll[i] = value; + else bitlen_d[i - HLIT] = value; + ++i; + } + } + else if(code == 17) /*repeat "0" 3-10 times*/ + { + unsigned replength = 3; /*read in the bits that indicate repeat length*/ + if((*bp + 3) > inbitlength) ERROR_BREAK(50); /*error, bit pointer jumps past memory*/ + replength += readBitsFromStream(bp, in, 3); + + /*repeat this value in the next lengths*/ + for(n = 0; n < replength; ++n) + { + if(i >= HLIT + HDIST) ERROR_BREAK(14); /*error: i is larger than the amount of codes*/ + + if(i < HLIT) bitlen_ll[i] = 0; + else bitlen_d[i - HLIT] = 0; + ++i; + } + } + else if(code == 18) /*repeat "0" 11-138 times*/ + { + unsigned replength = 11; /*read in the bits that indicate repeat length*/ + if((*bp + 7) > inbitlength) ERROR_BREAK(50); /*error, bit pointer jumps past memory*/ + replength += readBitsFromStream(bp, in, 7); + + /*repeat this value in the next lengths*/ + for(n = 0; n < replength; ++n) + { + if(i >= HLIT + HDIST) ERROR_BREAK(15); /*error: i is larger than the amount of codes*/ + + if(i < HLIT) bitlen_ll[i] = 0; + else bitlen_d[i - HLIT] = 0; + ++i; + } + } + else /*if(code == (unsigned)(-1))*/ /*huffmanDecodeSymbol returns (unsigned)(-1) in case of error*/ + { + if(code == (unsigned)(-1)) + { + /*return error code 10 or 11 depending on the situation that happened in huffmanDecodeSymbol + (10=no endcode, 11=wrong jump outside of tree)*/ + error = (*bp) > inbitlength ? 10 : 11; + } + else error = 16; /*unexisting code, this can never happen*/ + break; + } + } + if(error) break; + + if(bitlen_ll[256] == 0) ERROR_BREAK(64); /*the length of the end code 256 must be larger than 0*/ + + /*now we've finally got HLIT and HDIST, so generate the code trees, and the function is done*/ + error = HuffmanTree_makeFromLengths(tree_ll, bitlen_ll, NUM_DEFLATE_CODE_SYMBOLS, 15); + if(error) break; + error = HuffmanTree_makeFromLengths(tree_d, bitlen_d, NUM_DISTANCE_SYMBOLS, 15); + + break; /*end of error-while*/ + } + + lodepng_free(bitlen_cl); + lodepng_free(bitlen_ll); + lodepng_free(bitlen_d); + HuffmanTree_cleanup(&tree_cl); + + return error; +} + +/*inflate a block with dynamic of fixed Huffman tree*/ +static unsigned inflateHuffmanBlock(ucvector* out, const unsigned char* in, size_t* bp, + size_t* pos, size_t inlength, unsigned btype) +{ + unsigned error = 0; + HuffmanTree tree_ll; /*the huffman tree for literal and length codes*/ + HuffmanTree tree_d; /*the huffman tree for distance codes*/ + size_t inbitlength = inlength * 8; + + HuffmanTree_init(&tree_ll); + HuffmanTree_init(&tree_d); + + if(btype == 1) getTreeInflateFixed(&tree_ll, &tree_d); + else if(btype == 2) error = getTreeInflateDynamic(&tree_ll, &tree_d, in, bp, inlength); + + while(!error) /*decode all symbols until end reached, breaks at end code*/ + { + /*code_ll is literal, length or end code*/ + unsigned code_ll = huffmanDecodeSymbol(in, bp, &tree_ll, inbitlength); + if(code_ll <= 255) /*literal symbol*/ + { + /*ucvector_push_back would do the same, but for some reason the two lines below run 10% faster*/ + if(!ucvector_resize(out, (*pos) + 1)) ERROR_BREAK(83 /*alloc fail*/); + out->data[*pos] = (unsigned char)code_ll; + ++(*pos); + } + else if(code_ll >= FIRST_LENGTH_CODE_INDEX && code_ll <= LAST_LENGTH_CODE_INDEX) /*length code*/ + { + unsigned code_d, distance; + unsigned numextrabits_l, numextrabits_d; /*extra bits for length and distance*/ + size_t start, forward, backward, length; + + /*part 1: get length base*/ + length = LENGTHBASE[code_ll - FIRST_LENGTH_CODE_INDEX]; + + /*part 2: get extra bits and add the value of that to length*/ + numextrabits_l = LENGTHEXTRA[code_ll - FIRST_LENGTH_CODE_INDEX]; + if((*bp + numextrabits_l) > inbitlength) ERROR_BREAK(51); /*error, bit pointer will jump past memory*/ + length += readBitsFromStream(bp, in, numextrabits_l); + + /*part 3: get distance code*/ + code_d = huffmanDecodeSymbol(in, bp, &tree_d, inbitlength); + if(code_d > 29) + { + if(code_ll == (unsigned)(-1)) /*huffmanDecodeSymbol returns (unsigned)(-1) in case of error*/ + { + /*return error code 10 or 11 depending on the situation that happened in huffmanDecodeSymbol + (10=no endcode, 11=wrong jump outside of tree)*/ + error = (*bp) > inlength * 8 ? 10 : 11; + } + else error = 18; /*error: invalid distance code (30-31 are never used)*/ + break; + } + distance = DISTANCEBASE[code_d]; + + /*part 4: get extra bits from distance*/ + numextrabits_d = DISTANCEEXTRA[code_d]; + if((*bp + numextrabits_d) > inbitlength) ERROR_BREAK(51); /*error, bit pointer will jump past memory*/ + distance += readBitsFromStream(bp, in, numextrabits_d); + + /*part 5: fill in all the out[n] values based on the length and dist*/ + start = (*pos); + if(distance > start) ERROR_BREAK(52); /*too long backward distance*/ + backward = start - distance; + + if(!ucvector_resize(out, (*pos) + length)) ERROR_BREAK(83 /*alloc fail*/); + if (distance < length) { + for(forward = 0; forward < length; ++forward) + { + out->data[(*pos)++] = out->data[backward++]; + } + } else { + memcpy(out->data + *pos, out->data + backward, length); + *pos += length; + } + } + else if(code_ll == 256) + { + break; /*end code, break the loop*/ + } + else /*if(code == (unsigned)(-1))*/ /*huffmanDecodeSymbol returns (unsigned)(-1) in case of error*/ + { + /*return error code 10 or 11 depending on the situation that happened in huffmanDecodeSymbol + (10=no endcode, 11=wrong jump outside of tree)*/ + error = ((*bp) > inlength * 8) ? 10 : 11; + break; + } + } + + HuffmanTree_cleanup(&tree_ll); + HuffmanTree_cleanup(&tree_d); + + return error; +} + +static unsigned inflateNoCompression(ucvector* out, const unsigned char* in, size_t* bp, size_t* pos, size_t inlength) +{ + size_t p; + unsigned LEN, NLEN, n, error = 0; + + /*go to first boundary of byte*/ + while(((*bp) & 0x7) != 0) ++(*bp); + p = (*bp) / 8; /*byte position*/ + + /*read LEN (2 bytes) and NLEN (2 bytes)*/ + if(p + 4 >= inlength) return 52; /*error, bit pointer will jump past memory*/ + LEN = in[p] + 256u * in[p + 1]; p += 2; + NLEN = in[p] + 256u * in[p + 1]; p += 2; + + /*check if 16-bit NLEN is really the one's complement of LEN*/ + if(LEN + NLEN != 65535) return 21; /*error: NLEN is not one's complement of LEN*/ + + if(!ucvector_resize(out, (*pos) + LEN)) return 83; /*alloc fail*/ + + /*read the literal data: LEN bytes are now stored in the out buffer*/ + if(p + LEN > inlength) return 23; /*error: reading outside of in buffer*/ + for(n = 0; n < LEN; ++n) out->data[(*pos)++] = in[p++]; + + (*bp) = p * 8; + + return error; +} + +static unsigned lodepng_inflatev(ucvector* out, + const unsigned char* in, size_t insize, + const LodePNGDecompressSettings* settings) +{ + /*bit pointer in the "in" data, current byte is bp >> 3, current bit is bp & 0x7 (from lsb to msb of the byte)*/ + size_t bp = 0; + unsigned BFINAL = 0; + size_t pos = 0; /*byte position in the out buffer*/ + unsigned error = 0; + + (void)settings; + + while(!BFINAL) + { + unsigned BTYPE; + if(bp + 2 >= insize * 8) return 52; /*error, bit pointer will jump past memory*/ + BFINAL = readBitFromStream(&bp, in); + BTYPE = 1u * readBitFromStream(&bp, in); + BTYPE += 2u * readBitFromStream(&bp, in); + + if(BTYPE == 3) return 20; /*error: invalid BTYPE*/ + else if(BTYPE == 0) error = inflateNoCompression(out, in, &bp, &pos, insize); /*no compression*/ + else error = inflateHuffmanBlock(out, in, &bp, &pos, insize, BTYPE); /*compression, BTYPE 01 or 10*/ + + if(error) return error; + } + + return error; +} + +unsigned lodepng_inflate(unsigned char** out, size_t* outsize, + const unsigned char* in, size_t insize, + const LodePNGDecompressSettings* settings) +{ + unsigned error; + ucvector v; + ucvector_init_buffer(&v, *out, *outsize); + error = lodepng_inflatev(&v, in, insize, settings); + *out = v.data; + *outsize = v.size; + return error; +} + +static unsigned inflate(unsigned char** out, size_t* outsize, + const unsigned char* in, size_t insize, + const LodePNGDecompressSettings* settings) +{ + if(settings->custom_inflate) + { + return settings->custom_inflate(out, outsize, in, insize, settings); + } + else + { + return lodepng_inflate(out, outsize, in, insize, settings); + } +} + +#endif /*LODEPNG_COMPILE_DECODER*/ + +#ifdef LODEPNG_COMPILE_ENCODER + +/* ////////////////////////////////////////////////////////////////////////// */ +/* / Deflator (Compressor) / */ +/* ////////////////////////////////////////////////////////////////////////// */ + +static const size_t MAX_SUPPORTED_DEFLATE_LENGTH = 258; + +/*bitlen is the size in bits of the code*/ +static void addHuffmanSymbol(size_t* bp, ucvector* compressed, unsigned code, unsigned bitlen) +{ + addBitsToStreamReversed(bp, compressed, code, bitlen); +} + +/*search the index in the array, that has the largest value smaller than or equal to the given value, +given array must be sorted (if no value is smaller, it returns the size of the given array)*/ +static size_t searchCodeIndex(const unsigned* array, size_t array_size, size_t value) +{ + /*binary search (only small gain over linear). TODO: use CPU log2 instruction for getting symbols instead*/ + size_t left = 1; + size_t right = array_size - 1; + + while(left <= right) { + size_t mid = (left + right) >> 1; + if (array[mid] >= value) right = mid - 1; + else left = mid + 1; + } + if(left >= array_size || array[left] > value) left--; + return left; +} + +static void addLengthDistance(uivector* values, size_t length, size_t distance) +{ + /*values in encoded vector are those used by deflate: + 0-255: literal bytes + 256: end + 257-285: length/distance pair (length code, followed by extra length bits, distance code, extra distance bits) + 286-287: invalid*/ + + unsigned length_code = (unsigned)searchCodeIndex(LENGTHBASE, 29, length); + unsigned extra_length = (unsigned)(length - LENGTHBASE[length_code]); + unsigned dist_code = (unsigned)searchCodeIndex(DISTANCEBASE, 30, distance); + unsigned extra_distance = (unsigned)(distance - DISTANCEBASE[dist_code]); + + uivector_push_back(values, length_code + FIRST_LENGTH_CODE_INDEX); + uivector_push_back(values, extra_length); + uivector_push_back(values, dist_code); + uivector_push_back(values, extra_distance); +} + +/*3 bytes of data get encoded into two bytes. The hash cannot use more than 3 +bytes as input because 3 is the minimum match length for deflate*/ +static const unsigned HASH_NUM_VALUES = 65536; +static const unsigned HASH_BIT_MASK = 65535; /*HASH_NUM_VALUES - 1, but C90 does not like that as initializer*/ + +typedef struct Hash +{ + int* head; /*hash value to head circular pos - can be outdated if went around window*/ + /*circular pos to prev circular pos*/ + unsigned short* chain; + int* val; /*circular pos to hash value*/ + + /*TODO: do this not only for zeros but for any repeated byte. However for PNG + it's always going to be the zeros that dominate, so not important for PNG*/ + int* headz; /*similar to head, but for chainz*/ + unsigned short* chainz; /*those with same amount of zeros*/ + unsigned short* zeros; /*length of zeros streak, used as a second hash chain*/ +} Hash; + +static unsigned hash_init(Hash* hash, unsigned windowsize) +{ + unsigned i; + hash->head = (int*)lodepng_malloc(sizeof(int) * HASH_NUM_VALUES); + hash->val = (int*)lodepng_malloc(sizeof(int) * windowsize); + hash->chain = (unsigned short*)lodepng_malloc(sizeof(unsigned short) * windowsize); + + hash->zeros = (unsigned short*)lodepng_malloc(sizeof(unsigned short) * windowsize); + hash->headz = (int*)lodepng_malloc(sizeof(int) * (MAX_SUPPORTED_DEFLATE_LENGTH + 1)); + hash->chainz = (unsigned short*)lodepng_malloc(sizeof(unsigned short) * windowsize); + + if(!hash->head || !hash->chain || !hash->val || !hash->headz|| !hash->chainz || !hash->zeros) + { + return 83; /*alloc fail*/ + } + + /*initialize hash table*/ + for(i = 0; i != HASH_NUM_VALUES; ++i) hash->head[i] = -1; + for(i = 0; i != windowsize; ++i) hash->val[i] = -1; + for(i = 0; i != windowsize; ++i) hash->chain[i] = i; /*same value as index indicates uninitialized*/ + + for(i = 0; i <= MAX_SUPPORTED_DEFLATE_LENGTH; ++i) hash->headz[i] = -1; + for(i = 0; i != windowsize; ++i) hash->chainz[i] = i; /*same value as index indicates uninitialized*/ + + return 0; +} + +static void hash_cleanup(Hash* hash) +{ + lodepng_free(hash->head); + lodepng_free(hash->val); + lodepng_free(hash->chain); + + lodepng_free(hash->zeros); + lodepng_free(hash->headz); + lodepng_free(hash->chainz); +} + + + +static unsigned getHash(const unsigned char* data, size_t size, size_t pos) +{ + unsigned result = 0; + if(pos + 2 < size) + { + /*A simple shift and xor hash is used. Since the data of PNGs is dominated + by zeroes due to the filters, a better hash does not have a significant + effect on speed in traversing the chain, and causes more time spend on + calculating the hash.*/ + result ^= (unsigned)(data[pos + 0] << 0u); + result ^= (unsigned)(data[pos + 1] << 4u); + result ^= (unsigned)(data[pos + 2] << 8u); + } else { + size_t amount, i; + if(pos >= size) return 0; + amount = size - pos; + for(i = 0; i != amount; ++i) result ^= (unsigned)(data[pos + i] << (i * 8u)); + } + return result & HASH_BIT_MASK; +} + +static unsigned countZeros(const unsigned char* data, size_t size, size_t pos) +{ + const unsigned char* start = data + pos; + const unsigned char* end = start + MAX_SUPPORTED_DEFLATE_LENGTH; + if(end > data + size) end = data + size; + data = start; + while(data != end && *data == 0) ++data; + /*subtracting two addresses returned as 32-bit number (max value is MAX_SUPPORTED_DEFLATE_LENGTH)*/ + return (unsigned)(data - start); +} + +/*wpos = pos & (windowsize - 1)*/ +static void updateHashChain(Hash* hash, size_t wpos, unsigned hashval, unsigned short numzeros) +{ + hash->val[wpos] = (int)hashval; + if(hash->head[hashval] != -1) hash->chain[wpos] = hash->head[hashval]; + hash->head[hashval] = (int)wpos; + + hash->zeros[wpos] = numzeros; + if(hash->headz[numzeros] != -1) hash->chainz[wpos] = hash->headz[numzeros]; + hash->headz[numzeros] = (int)wpos; +} + +/* +LZ77-encode the data. Return value is error code. The input are raw bytes, the output +is in the form of unsigned integers with codes representing for example literal bytes, or +length/distance pairs. +It uses a hash table technique to let it encode faster. When doing LZ77 encoding, a +sliding window (of windowsize) is used, and all past bytes in that window can be used as +the "dictionary". A brute force search through all possible distances would be slow, and +this hash technique is one out of several ways to speed this up. +*/ +static unsigned encodeLZ77(uivector* out, Hash* hash, + const unsigned char* in, size_t inpos, size_t insize, unsigned windowsize, + unsigned minmatch, unsigned nicematch, unsigned lazymatching) +{ + size_t pos; + unsigned i, error = 0; + /*for large window lengths, assume the user wants no compression loss. Otherwise, max hash chain length speedup.*/ + unsigned maxchainlength = windowsize >= 8192 ? windowsize : windowsize / 8; + unsigned maxlazymatch = windowsize >= 8192 ? MAX_SUPPORTED_DEFLATE_LENGTH : 64; + + unsigned usezeros = 1; /*not sure if setting it to false for windowsize < 8192 is better or worse*/ + unsigned numzeros = 0; + + unsigned offset; /*the offset represents the distance in LZ77 terminology*/ + unsigned length; + unsigned lazy = 0; + unsigned lazylength = 0, lazyoffset = 0; + unsigned hashval; + unsigned current_offset, current_length; + unsigned prev_offset; + const unsigned char *lastptr, *foreptr, *backptr; + unsigned hashpos; + + if(windowsize == 0 || windowsize > 32768) return 60; /*error: windowsize smaller/larger than allowed*/ + if((windowsize & (windowsize - 1)) != 0) return 90; /*error: must be power of two*/ + + if(nicematch > MAX_SUPPORTED_DEFLATE_LENGTH) nicematch = MAX_SUPPORTED_DEFLATE_LENGTH; + + for(pos = inpos; pos < insize; ++pos) + { + size_t wpos = pos & (windowsize - 1); /*position for in 'circular' hash buffers*/ + unsigned chainlength = 0; + + hashval = getHash(in, insize, pos); + + if(usezeros && hashval == 0) + { + if(numzeros == 0) numzeros = countZeros(in, insize, pos); + else if(pos + numzeros > insize || in[pos + numzeros - 1] != 0) --numzeros; + } + else + { + numzeros = 0; + } + + updateHashChain(hash, wpos, hashval, numzeros); + + /*the length and offset found for the current position*/ + length = 0; + offset = 0; + + hashpos = hash->chain[wpos]; + + lastptr = &in[insize < pos + MAX_SUPPORTED_DEFLATE_LENGTH ? insize : pos + MAX_SUPPORTED_DEFLATE_LENGTH]; + + /*search for the longest string*/ + prev_offset = 0; + for(;;) + { + if(chainlength++ >= maxchainlength) break; + current_offset = hashpos <= wpos ? (unsigned int)(wpos - hashpos) : (unsigned int)(wpos - hashpos + windowsize); + + if(current_offset < prev_offset) break; /*stop when went completely around the circular buffer*/ + prev_offset = current_offset; + if(current_offset > 0) + { + /*test the next characters*/ + foreptr = &in[pos]; + backptr = &in[pos - current_offset]; + + /*common case in PNGs is lots of zeros. Quickly skip over them as a speedup*/ + if(numzeros >= 3) + { + unsigned skip = hash->zeros[hashpos]; + if(skip > numzeros) skip = numzeros; + backptr += skip; + foreptr += skip; + } + + while(foreptr != lastptr && *backptr == *foreptr) /*maximum supported length by deflate is max length*/ + { + ++backptr; + ++foreptr; + } + current_length = (unsigned)(foreptr - &in[pos]); + + if(current_length > length) + { + length = current_length; /*the longest length*/ + offset = current_offset; /*the offset that is related to this longest length*/ + /*jump out once a length of max length is found (speed gain). This also jumps + out if length is MAX_SUPPORTED_DEFLATE_LENGTH*/ + if(current_length >= nicematch) break; + } + } + + if(hashpos == hash->chain[hashpos]) break; + + if(numzeros >= 3 && length > numzeros) + { + hashpos = hash->chainz[hashpos]; + if(hash->zeros[hashpos] != numzeros) break; + } + else + { + hashpos = hash->chain[hashpos]; + /*outdated hash value, happens if particular value was not encountered in whole last window*/ + if(hash->val[hashpos] != (int)hashval) break; + } + } + + if(lazymatching) + { + if(!lazy && length >= 3 && length <= maxlazymatch && length < MAX_SUPPORTED_DEFLATE_LENGTH) + { + lazy = 1; + lazylength = length; + lazyoffset = offset; + continue; /*try the next byte*/ + } + if(lazy) + { + lazy = 0; + if(pos == 0) ERROR_BREAK(81); + if(length > lazylength + 1) + { + /*push the previous character as literal*/ + if(!uivector_push_back(out, in[pos - 1])) ERROR_BREAK(83 /*alloc fail*/); + } + else + { + length = lazylength; + offset = lazyoffset; + hash->head[hashval] = -1; /*the same hashchain update will be done, this ensures no wrong alteration*/ + hash->headz[numzeros] = -1; /*idem*/ + --pos; + } + } + } + if(length >= 3 && offset > windowsize) ERROR_BREAK(86 /*too big (or overflown negative) offset*/); + + /*encode it as length/distance pair or literal value*/ + if(length < 3) /*only lengths of 3 or higher are supported as length/distance pair*/ + { + if(!uivector_push_back(out, in[pos])) ERROR_BREAK(83 /*alloc fail*/); + } + else if(length < minmatch || (length == 3 && offset > 4096)) + { + /*compensate for the fact that longer offsets have more extra bits, a + length of only 3 may be not worth it then*/ + if(!uivector_push_back(out, in[pos])) ERROR_BREAK(83 /*alloc fail*/); + } + else + { + addLengthDistance(out, length, offset); + for(i = 1; i < length; ++i) + { + ++pos; + wpos = pos & (windowsize - 1); + hashval = getHash(in, insize, pos); + if(usezeros && hashval == 0) + { + if(numzeros == 0) numzeros = countZeros(in, insize, pos); + else if(pos + numzeros > insize || in[pos + numzeros - 1] != 0) --numzeros; + } + else + { + numzeros = 0; + } + updateHashChain(hash, wpos, hashval, numzeros); + } + } + } /*end of the loop through each character of input*/ + + return error; +} + +/* /////////////////////////////////////////////////////////////////////////// */ + +static unsigned deflateNoCompression(ucvector* out, const unsigned char* data, size_t datasize) +{ + /*non compressed deflate block data: 1 bit BFINAL,2 bits BTYPE,(5 bits): it jumps to start of next byte, + 2 bytes LEN, 2 bytes NLEN, LEN bytes literal DATA*/ + + size_t i, j, numdeflateblocks = (datasize + 65534) / 65535; + unsigned datapos = 0; + for(i = 0; i != numdeflateblocks; ++i) + { + unsigned BFINAL, BTYPE, LEN, NLEN; + unsigned char firstbyte; + + BFINAL = (i == numdeflateblocks - 1); + BTYPE = 0; + + firstbyte = (unsigned char)(BFINAL + ((BTYPE & 1) << 1) + ((BTYPE & 2) << 1)); + ucvector_push_back(out, firstbyte); + + LEN = 65535; + if(datasize - datapos < 65535) LEN = (unsigned)datasize - datapos; + NLEN = 65535 - LEN; + + ucvector_push_back(out, (unsigned char)(LEN & 255)); + ucvector_push_back(out, (unsigned char)(LEN >> 8)); + ucvector_push_back(out, (unsigned char)(NLEN & 255)); + ucvector_push_back(out, (unsigned char)(NLEN >> 8)); + + /*Decompressed data*/ + for(j = 0; j < 65535 && datapos < datasize; ++j) + { + ucvector_push_back(out, data[datapos++]); + } + } + + return 0; +} + +/* +write the lz77-encoded data, which has lit, len and dist codes, to compressed stream using huffman trees. +tree_ll: the tree for lit and len codes. +tree_d: the tree for distance codes. +*/ +static void writeLZ77data(size_t* bp, ucvector* out, const uivector* lz77_encoded, + const HuffmanTree* tree_ll, const HuffmanTree* tree_d) +{ + size_t i = 0; + for(i = 0; i != lz77_encoded->size; ++i) + { + unsigned val = lz77_encoded->data[i]; + addHuffmanSymbol(bp, out, HuffmanTree_getCode(tree_ll, val), HuffmanTree_getLength(tree_ll, val)); + if(val > 256) /*for a length code, 3 more things have to be added*/ + { + unsigned length_index = val - FIRST_LENGTH_CODE_INDEX; + unsigned n_length_extra_bits = LENGTHEXTRA[length_index]; + unsigned length_extra_bits = lz77_encoded->data[++i]; + + unsigned distance_code = lz77_encoded->data[++i]; + + unsigned distance_index = distance_code; + unsigned n_distance_extra_bits = DISTANCEEXTRA[distance_index]; + unsigned distance_extra_bits = lz77_encoded->data[++i]; + + addBitsToStream(bp, out, length_extra_bits, n_length_extra_bits); + addHuffmanSymbol(bp, out, HuffmanTree_getCode(tree_d, distance_code), + HuffmanTree_getLength(tree_d, distance_code)); + addBitsToStream(bp, out, distance_extra_bits, n_distance_extra_bits); + } + } +} + +/*Deflate for a block of type "dynamic", that is, with freely, optimally, created huffman trees*/ +static unsigned deflateDynamic(ucvector* out, size_t* bp, Hash* hash, + const unsigned char* data, size_t datapos, size_t dataend, + const LodePNGCompressSettings* settings, unsigned final) +{ + unsigned error = 0; + + /* + A block is compressed as follows: The PNG data is lz77 encoded, resulting in + literal bytes and length/distance pairs. This is then huffman compressed with + two huffman trees. One huffman tree is used for the lit and len values ("ll"), + another huffman tree is used for the dist values ("d"). These two trees are + stored using their code lengths, and to compress even more these code lengths + are also run-length encoded and huffman compressed. This gives a huffman tree + of code lengths "cl". The code lenghts used to describe this third tree are + the code length code lengths ("clcl"). + */ + + /*The lz77 encoded data, represented with integers since there will also be length and distance codes in it*/ + uivector lz77_encoded; + HuffmanTree tree_ll; /*tree for lit,len values*/ + HuffmanTree tree_d; /*tree for distance codes*/ + HuffmanTree tree_cl; /*tree for encoding the code lengths representing tree_ll and tree_d*/ + uivector frequencies_ll; /*frequency of lit,len codes*/ + uivector frequencies_d; /*frequency of dist codes*/ + uivector frequencies_cl; /*frequency of code length codes*/ + uivector bitlen_lld; /*lit,len,dist code lenghts (int bits), literally (without repeat codes).*/ + uivector bitlen_lld_e; /*bitlen_lld encoded with repeat codes (this is a rudemtary run length compression)*/ + /*bitlen_cl is the code length code lengths ("clcl"). The bit lengths of codes to represent tree_cl + (these are written as is in the file, it would be crazy to compress these using yet another huffman + tree that needs to be represented by yet another set of code lengths)*/ + uivector bitlen_cl; + size_t datasize = dataend - datapos; + + /* + Due to the huffman compression of huffman tree representations ("two levels"), there are some anologies: + bitlen_lld is to tree_cl what data is to tree_ll and tree_d. + bitlen_lld_e is to bitlen_lld what lz77_encoded is to data. + bitlen_cl is to bitlen_lld_e what bitlen_lld is to lz77_encoded. + */ + + unsigned BFINAL = final; + size_t numcodes_ll, numcodes_d, i; + unsigned HLIT, HDIST, HCLEN; + + uivector_init(&lz77_encoded); + HuffmanTree_init(&tree_ll); + HuffmanTree_init(&tree_d); + HuffmanTree_init(&tree_cl); + uivector_init(&frequencies_ll); + uivector_init(&frequencies_d); + uivector_init(&frequencies_cl); + uivector_init(&bitlen_lld); + uivector_init(&bitlen_lld_e); + uivector_init(&bitlen_cl); + + /*This while loop never loops due to a break at the end, it is here to + allow breaking out of it to the cleanup phase on error conditions.*/ + while(!error) + { + if(settings->use_lz77) + { + error = encodeLZ77(&lz77_encoded, hash, data, datapos, dataend, settings->windowsize, + settings->minmatch, settings->nicematch, settings->lazymatching); + if(error) break; + } + else + { + if(!uivector_resize(&lz77_encoded, datasize)) ERROR_BREAK(83 /*alloc fail*/); + for(i = datapos; i < dataend; ++i) lz77_encoded.data[i - datapos] = data[i]; /*no LZ77, but still will be Huffman compressed*/ + } + + if(!uivector_resizev(&frequencies_ll, 286, 0)) ERROR_BREAK(83 /*alloc fail*/); + if(!uivector_resizev(&frequencies_d, 30, 0)) ERROR_BREAK(83 /*alloc fail*/); + + /*Count the frequencies of lit, len and dist codes*/ + for(i = 0; i != lz77_encoded.size; ++i) + { + unsigned symbol = lz77_encoded.data[i]; + ++frequencies_ll.data[symbol]; + if(symbol > 256) + { + unsigned dist = lz77_encoded.data[i + 2]; + ++frequencies_d.data[dist]; + i += 3; + } + } + frequencies_ll.data[256] = 1; /*there will be exactly 1 end code, at the end of the block*/ + + /*Make both huffman trees, one for the lit and len codes, one for the dist codes*/ + error = HuffmanTree_makeFromFrequencies(&tree_ll, frequencies_ll.data, 257, frequencies_ll.size, 15); + if(error) break; + /*2, not 1, is chosen for mincodes: some buggy PNG decoders require at least 2 symbols in the dist tree*/ + error = HuffmanTree_makeFromFrequencies(&tree_d, frequencies_d.data, 2, frequencies_d.size, 15); + if(error) break; + + numcodes_ll = tree_ll.numcodes; if(numcodes_ll > 286) numcodes_ll = 286; + numcodes_d = tree_d.numcodes; if(numcodes_d > 30) numcodes_d = 30; + /*store the code lengths of both generated trees in bitlen_lld*/ + for(i = 0; i != numcodes_ll; ++i) uivector_push_back(&bitlen_lld, HuffmanTree_getLength(&tree_ll, (unsigned)i)); + for(i = 0; i != numcodes_d; ++i) uivector_push_back(&bitlen_lld, HuffmanTree_getLength(&tree_d, (unsigned)i)); + + /*run-length compress bitlen_ldd into bitlen_lld_e by using repeat codes 16 (copy length 3-6 times), + 17 (3-10 zeroes), 18 (11-138 zeroes)*/ + for(i = 0; i != (unsigned)bitlen_lld.size; ++i) + { + unsigned j = 0; /*amount of repititions*/ + while(i + j + 1 < (unsigned)bitlen_lld.size && bitlen_lld.data[i + j + 1] == bitlen_lld.data[i]) ++j; + + if(bitlen_lld.data[i] == 0 && j >= 2) /*repeat code for zeroes*/ + { + ++j; /*include the first zero*/ + if(j <= 10) /*repeat code 17 supports max 10 zeroes*/ + { + uivector_push_back(&bitlen_lld_e, 17); + uivector_push_back(&bitlen_lld_e, j - 3); + } + else /*repeat code 18 supports max 138 zeroes*/ + { + if(j > 138) j = 138; + uivector_push_back(&bitlen_lld_e, 18); + uivector_push_back(&bitlen_lld_e, j - 11); + } + i += (j - 1); + } + else if(j >= 3) /*repeat code for value other than zero*/ + { + size_t k; + unsigned num = j / 6, rest = j % 6; + uivector_push_back(&bitlen_lld_e, bitlen_lld.data[i]); + for(k = 0; k < num; ++k) + { + uivector_push_back(&bitlen_lld_e, 16); + uivector_push_back(&bitlen_lld_e, 6 - 3); + } + if(rest >= 3) + { + uivector_push_back(&bitlen_lld_e, 16); + uivector_push_back(&bitlen_lld_e, rest - 3); + } + else j -= rest; + i += j; + } + else /*too short to benefit from repeat code*/ + { + uivector_push_back(&bitlen_lld_e, bitlen_lld.data[i]); + } + } + + /*generate tree_cl, the huffmantree of huffmantrees*/ + + if(!uivector_resizev(&frequencies_cl, NUM_CODE_LENGTH_CODES, 0)) ERROR_BREAK(83 /*alloc fail*/); + for(i = 0; i != bitlen_lld_e.size; ++i) + { + ++frequencies_cl.data[bitlen_lld_e.data[i]]; + /*after a repeat code come the bits that specify the number of repetitions, + those don't need to be in the frequencies_cl calculation*/ + if(bitlen_lld_e.data[i] >= 16) ++i; + } + + error = HuffmanTree_makeFromFrequencies(&tree_cl, frequencies_cl.data, + frequencies_cl.size, frequencies_cl.size, 7); + if(error) break; + + if(!uivector_resize(&bitlen_cl, tree_cl.numcodes)) ERROR_BREAK(83 /*alloc fail*/); + for(i = 0; i != tree_cl.numcodes; ++i) + { + /*lenghts of code length tree is in the order as specified by deflate*/ + bitlen_cl.data[i] = HuffmanTree_getLength(&tree_cl, CLCL_ORDER[i]); + } + while(bitlen_cl.data[bitlen_cl.size - 1] == 0 && bitlen_cl.size > 4) + { + /*remove zeros at the end, but minimum size must be 4*/ + if(!uivector_resize(&bitlen_cl, bitlen_cl.size - 1)) ERROR_BREAK(83 /*alloc fail*/); + } + if(error) break; + + /* + Write everything into the output + + After the BFINAL and BTYPE, the dynamic block consists out of the following: + - 5 bits HLIT, 5 bits HDIST, 4 bits HCLEN + - (HCLEN+4)*3 bits code lengths of code length alphabet + - HLIT + 257 code lenghts of lit/length alphabet (encoded using the code length + alphabet, + possible repetition codes 16, 17, 18) + - HDIST + 1 code lengths of distance alphabet (encoded using the code length + alphabet, + possible repetition codes 16, 17, 18) + - compressed data + - 256 (end code) + */ + + /*Write block type*/ + addBitToStream(bp, out, BFINAL); + addBitToStream(bp, out, 0); /*first bit of BTYPE "dynamic"*/ + addBitToStream(bp, out, 1); /*second bit of BTYPE "dynamic"*/ + + /*write the HLIT, HDIST and HCLEN values*/ + HLIT = (unsigned)(numcodes_ll - 257); + HDIST = (unsigned)(numcodes_d - 1); + HCLEN = (unsigned)bitlen_cl.size - 4; + /*trim zeroes for HCLEN. HLIT and HDIST were already trimmed at tree creation*/ + while(!bitlen_cl.data[HCLEN + 4 - 1] && HCLEN > 0) --HCLEN; + addBitsToStream(bp, out, HLIT, 5); + addBitsToStream(bp, out, HDIST, 5); + addBitsToStream(bp, out, HCLEN, 4); + + /*write the code lenghts of the code length alphabet*/ + for(i = 0; i != HCLEN + 4; ++i) addBitsToStream(bp, out, bitlen_cl.data[i], 3); + + /*write the lenghts of the lit/len AND the dist alphabet*/ + for(i = 0; i != bitlen_lld_e.size; ++i) + { + addHuffmanSymbol(bp, out, HuffmanTree_getCode(&tree_cl, bitlen_lld_e.data[i]), + HuffmanTree_getLength(&tree_cl, bitlen_lld_e.data[i])); + /*extra bits of repeat codes*/ + if(bitlen_lld_e.data[i] == 16) addBitsToStream(bp, out, bitlen_lld_e.data[++i], 2); + else if(bitlen_lld_e.data[i] == 17) addBitsToStream(bp, out, bitlen_lld_e.data[++i], 3); + else if(bitlen_lld_e.data[i] == 18) addBitsToStream(bp, out, bitlen_lld_e.data[++i], 7); + } + + /*write the compressed data symbols*/ + writeLZ77data(bp, out, &lz77_encoded, &tree_ll, &tree_d); + /*error: the length of the end code 256 must be larger than 0*/ + if(HuffmanTree_getLength(&tree_ll, 256) == 0) ERROR_BREAK(64); + + /*write the end code*/ + addHuffmanSymbol(bp, out, HuffmanTree_getCode(&tree_ll, 256), HuffmanTree_getLength(&tree_ll, 256)); + + break; /*end of error-while*/ + } + + /*cleanup*/ + uivector_cleanup(&lz77_encoded); + HuffmanTree_cleanup(&tree_ll); + HuffmanTree_cleanup(&tree_d); + HuffmanTree_cleanup(&tree_cl); + uivector_cleanup(&frequencies_ll); + uivector_cleanup(&frequencies_d); + uivector_cleanup(&frequencies_cl); + uivector_cleanup(&bitlen_lld_e); + uivector_cleanup(&bitlen_lld); + uivector_cleanup(&bitlen_cl); + + return error; +} + +static unsigned deflateFixed(ucvector* out, size_t* bp, Hash* hash, + const unsigned char* data, + size_t datapos, size_t dataend, + const LodePNGCompressSettings* settings, unsigned final) +{ + HuffmanTree tree_ll; /*tree for literal values and length codes*/ + HuffmanTree tree_d; /*tree for distance codes*/ + + unsigned BFINAL = final; + unsigned error = 0; + size_t i; + + HuffmanTree_init(&tree_ll); + HuffmanTree_init(&tree_d); + + generateFixedLitLenTree(&tree_ll); + generateFixedDistanceTree(&tree_d); + + addBitToStream(bp, out, BFINAL); + addBitToStream(bp, out, 1); /*first bit of BTYPE*/ + addBitToStream(bp, out, 0); /*second bit of BTYPE*/ + + if(settings->use_lz77) /*LZ77 encoded*/ + { + uivector lz77_encoded; + uivector_init(&lz77_encoded); + error = encodeLZ77(&lz77_encoded, hash, data, datapos, dataend, settings->windowsize, + settings->minmatch, settings->nicematch, settings->lazymatching); + if(!error) writeLZ77data(bp, out, &lz77_encoded, &tree_ll, &tree_d); + uivector_cleanup(&lz77_encoded); + } + else /*no LZ77, but still will be Huffman compressed*/ + { + for(i = datapos; i < dataend; ++i) + { + addHuffmanSymbol(bp, out, HuffmanTree_getCode(&tree_ll, data[i]), HuffmanTree_getLength(&tree_ll, data[i])); + } + } + /*add END code*/ + if(!error) addHuffmanSymbol(bp, out, HuffmanTree_getCode(&tree_ll, 256), HuffmanTree_getLength(&tree_ll, 256)); + + /*cleanup*/ + HuffmanTree_cleanup(&tree_ll); + HuffmanTree_cleanup(&tree_d); + + return error; +} + +static unsigned lodepng_deflatev(ucvector* out, const unsigned char* in, size_t insize, + const LodePNGCompressSettings* settings) +{ + unsigned error = 0; + size_t i, blocksize, numdeflateblocks; + size_t bp = 0; /*the bit pointer*/ + Hash hash; + + if(settings->btype > 2) return 61; + else if(settings->btype == 0) return deflateNoCompression(out, in, insize); + else if(settings->btype == 1) blocksize = insize; + else /*if(settings->btype == 2)*/ + { + /*on PNGs, deflate blocks of 65-262k seem to give most dense encoding*/ + blocksize = insize / 8 + 8; + if(blocksize < 65536) blocksize = 65536; + if(blocksize > 262144) blocksize = 262144; + } + + numdeflateblocks = (insize + blocksize - 1) / blocksize; + if(numdeflateblocks == 0) numdeflateblocks = 1; + + error = hash_init(&hash, settings->windowsize); + if(error) return error; + + for(i = 0; i != numdeflateblocks && !error; ++i) + { + unsigned final = (i == numdeflateblocks - 1); + size_t start = i * blocksize; + size_t end = start + blocksize; + if(end > insize) end = insize; + + if(settings->btype == 1) error = deflateFixed(out, &bp, &hash, in, start, end, settings, final); + else if(settings->btype == 2) error = deflateDynamic(out, &bp, &hash, in, start, end, settings, final); + } + + hash_cleanup(&hash); + + return error; +} + +unsigned lodepng_deflate(unsigned char** out, size_t* outsize, + const unsigned char* in, size_t insize, + const LodePNGCompressSettings* settings) +{ + unsigned error; + ucvector v; + ucvector_init_buffer(&v, *out, *outsize); + error = lodepng_deflatev(&v, in, insize, settings); + *out = v.data; + *outsize = v.size; + return error; +} + +static unsigned deflate(unsigned char** out, size_t* outsize, + const unsigned char* in, size_t insize, + const LodePNGCompressSettings* settings) +{ + if(settings->custom_deflate) + { + return settings->custom_deflate(out, outsize, in, insize, settings); + } + else + { + return lodepng_deflate(out, outsize, in, insize, settings); + } +} + +#endif /*LODEPNG_COMPILE_DECODER*/ + +/* ////////////////////////////////////////////////////////////////////////// */ +/* / Adler32 */ +/* ////////////////////////////////////////////////////////////////////////// */ + +static unsigned update_adler32(unsigned adler, const unsigned char* data, unsigned len) +{ + unsigned s1 = adler & 0xffff; + unsigned s2 = (adler >> 16) & 0xffff; + + while(len > 0) + { + /*at least 5550 sums can be done before the sums overflow, saving a lot of module divisions*/ + unsigned amount = len > 5550 ? 5550 : len; + len -= amount; + while(amount > 0) + { + s1 += (*data++); + s2 += s1; + --amount; + } + s1 %= 65521; + s2 %= 65521; + } + + return (s2 << 16) | s1; +} + +/*Return the adler32 of the bytes data[0..len-1]*/ +static unsigned adler32(const unsigned char* data, unsigned len) +{ + return update_adler32(1L, data, len); +} + +/* ////////////////////////////////////////////////////////////////////////// */ +/* / Zlib / */ +/* ////////////////////////////////////////////////////////////////////////// */ + +#ifdef LODEPNG_COMPILE_DECODER + +unsigned lodepng_zlib_decompress(unsigned char** out, size_t* outsize, const unsigned char* in, + size_t insize, const LodePNGDecompressSettings* settings) +{ + unsigned error = 0; + unsigned CM, CINFO, FDICT; + + if(insize < 2) return 53; /*error, size of zlib data too small*/ + /*read information from zlib header*/ + if((in[0] * 256 + in[1]) % 31 != 0) + { + /*error: 256 * in[0] + in[1] must be a multiple of 31, the FCHECK value is supposed to be made that way*/ + return 24; + } + + CM = in[0] & 15; + CINFO = (in[0] >> 4) & 15; + /*FCHECK = in[1] & 31;*/ /*FCHECK is already tested above*/ + FDICT = (in[1] >> 5) & 1; + /*FLEVEL = (in[1] >> 6) & 3;*/ /*FLEVEL is not used here*/ + + if(CM != 8 || CINFO > 7) + { + /*error: only compression method 8: inflate with sliding window of 32k is supported by the PNG spec*/ + return 25; + } + if(FDICT != 0) + { + /*error: the specification of PNG says about the zlib stream: + "The additional flags shall not specify a preset dictionary."*/ + return 26; + } + + error = inflate(out, outsize, in + 2, insize - 2, settings); + if(error) return error; + + if(!settings->ignore_adler32) + { + unsigned ADLER32 = lodepng_read32bitInt(&in[insize - 4]); + unsigned checksum = adler32(*out, (unsigned)(*outsize)); + if(checksum != ADLER32) return 58; /*error, adler checksum not correct, data must be corrupted*/ + } + + return 0; /*no error*/ +} + +static unsigned zlib_decompress(unsigned char** out, size_t* outsize, const unsigned char* in, + size_t insize, const LodePNGDecompressSettings* settings) +{ + if(settings->custom_zlib) + { + return settings->custom_zlib(out, outsize, in, insize, settings); + } + else + { + return lodepng_zlib_decompress(out, outsize, in, insize, settings); + } +} + +#endif /*LODEPNG_COMPILE_DECODER*/ + +#ifdef LODEPNG_COMPILE_ENCODER + +unsigned lodepng_zlib_compress(unsigned char** out, size_t* outsize, const unsigned char* in, + size_t insize, const LodePNGCompressSettings* settings) +{ + /*initially, *out must be NULL and outsize 0, if you just give some random *out + that's pointing to a non allocated buffer, this'll crash*/ + ucvector outv; + size_t i; + unsigned error; + unsigned char* deflatedata = 0; + size_t deflatesize = 0; + + /*zlib data: 1 byte CMF (CM+CINFO), 1 byte FLG, deflate data, 4 byte ADLER32 checksum of the Decompressed data*/ + unsigned CMF = 120; /*0b01111000: CM 8, CINFO 7. With CINFO 7, any window size up to 32768 can be used.*/ + unsigned FLEVEL = 0; + unsigned FDICT = 0; + unsigned CMFFLG = 256 * CMF + FDICT * 32 + FLEVEL * 64; + unsigned FCHECK = 31 - CMFFLG % 31; + CMFFLG += FCHECK; + + /*ucvector-controlled version of the output buffer, for dynamic array*/ + ucvector_init_buffer(&outv, *out, *outsize); + + ucvector_push_back(&outv, (unsigned char)(CMFFLG >> 8)); + ucvector_push_back(&outv, (unsigned char)(CMFFLG & 255)); + + error = deflate(&deflatedata, &deflatesize, in, insize, settings); + + if(!error) + { + unsigned ADLER32 = adler32(in, (unsigned)insize); + for(i = 0; i != deflatesize; ++i) ucvector_push_back(&outv, deflatedata[i]); + lodepng_free(deflatedata); + lodepng_add32bitInt(&outv, ADLER32); + } + + *out = outv.data; + *outsize = outv.size; + + return error; +} + +/* compress using the default or custom zlib function */ +static unsigned zlib_compress(unsigned char** out, size_t* outsize, const unsigned char* in, + size_t insize, const LodePNGCompressSettings* settings) +{ + if(settings->custom_zlib) + { + return settings->custom_zlib(out, outsize, in, insize, settings); + } + else + { + return lodepng_zlib_compress(out, outsize, in, insize, settings); + } +} + +#endif /*LODEPNG_COMPILE_ENCODER*/ + +#else /*no LODEPNG_COMPILE_ZLIB*/ + +#ifdef LODEPNG_COMPILE_DECODER +static unsigned zlib_decompress(unsigned char** out, size_t* outsize, const unsigned char* in, + size_t insize, const LodePNGDecompressSettings* settings) +{ + if(!settings->custom_zlib) return 87; /*no custom zlib function provided */ + return settings->custom_zlib(out, outsize, in, insize, settings); +} +#endif /*LODEPNG_COMPILE_DECODER*/ +#ifdef LODEPNG_COMPILE_ENCODER +static unsigned zlib_compress(unsigned char** out, size_t* outsize, const unsigned char* in, + size_t insize, const LodePNGCompressSettings* settings) +{ + if(!settings->custom_zlib) return 87; /*no custom zlib function provided */ + return settings->custom_zlib(out, outsize, in, insize, settings); +} +#endif /*LODEPNG_COMPILE_ENCODER*/ + +#endif /*LODEPNG_COMPILE_ZLIB*/ + +/* ////////////////////////////////////////////////////////////////////////// */ + +#ifdef LODEPNG_COMPILE_ENCODER + +/*this is a good tradeoff between speed and compression ratio*/ +#define DEFAULT_WINDOWSIZE 2048 + +void lodepng_compress_settings_init(LodePNGCompressSettings* settings) +{ + /*compress with dynamic huffman tree (not in the mathematical sense, just not the predefined one)*/ + settings->btype = 2; + settings->use_lz77 = 1; + settings->windowsize = DEFAULT_WINDOWSIZE; + settings->minmatch = 3; + settings->nicematch = 128; + settings->lazymatching = 1; + + settings->custom_zlib = 0; + settings->custom_deflate = 0; + settings->custom_context = 0; +} + +const LodePNGCompressSettings lodepng_default_compress_settings = {2, 1, DEFAULT_WINDOWSIZE, 3, 128, 1, 0, 0, 0}; + + +#endif /*LODEPNG_COMPILE_ENCODER*/ + +#ifdef LODEPNG_COMPILE_DECODER + +void lodepng_decompress_settings_init(LodePNGDecompressSettings* settings) +{ + settings->ignore_adler32 = 0; + + settings->custom_zlib = 0; + settings->custom_inflate = 0; + settings->custom_context = 0; +} + +const LodePNGDecompressSettings lodepng_default_decompress_settings = {0, 0, 0, 0}; + +#endif /*LODEPNG_COMPILE_DECODER*/ + +/* ////////////////////////////////////////////////////////////////////////// */ +/* ////////////////////////////////////////////////////////////////////////// */ +/* // End of Zlib related code. Begin of PNG related code. // */ +/* ////////////////////////////////////////////////////////////////////////// */ +/* ////////////////////////////////////////////////////////////////////////// */ + +#ifdef LODEPNG_COMPILE_PNG + +/* ////////////////////////////////////////////////////////////////////////// */ +/* / CRC32 / */ +/* ////////////////////////////////////////////////////////////////////////// */ + + +#ifndef LODEPNG_NO_COMPILE_CRC +/* CRC polynomial: 0xedb88320 */ +static unsigned lodepng_crc32_table[256] = { + 0u, 1996959894u, 3993919788u, 2567524794u, 124634137u, 1886057615u, 3915621685u, 2657392035u, + 249268274u, 2044508324u, 3772115230u, 2547177864u, 162941995u, 2125561021u, 3887607047u, 2428444049u, + 498536548u, 1789927666u, 4089016648u, 2227061214u, 450548861u, 1843258603u, 4107580753u, 2211677639u, + 325883990u, 1684777152u, 4251122042u, 2321926636u, 335633487u, 1661365465u, 4195302755u, 2366115317u, + 997073096u, 1281953886u, 3579855332u, 2724688242u, 1006888145u, 1258607687u, 3524101629u, 2768942443u, + 901097722u, 1119000684u, 3686517206u, 2898065728u, 853044451u, 1172266101u, 3705015759u, 2882616665u, + 651767980u, 1373503546u, 3369554304u, 3218104598u, 565507253u, 1454621731u, 3485111705u, 3099436303u, + 671266974u, 1594198024u, 3322730930u, 2970347812u, 795835527u, 1483230225u, 3244367275u, 3060149565u, + 1994146192u, 31158534u, 2563907772u, 4023717930u, 1907459465u, 112637215u, 2680153253u, 3904427059u, + 2013776290u, 251722036u, 2517215374u, 3775830040u, 2137656763u, 141376813u, 2439277719u, 3865271297u, + 1802195444u, 476864866u, 2238001368u, 4066508878u, 1812370925u, 453092731u, 2181625025u, 4111451223u, + 1706088902u, 314042704u, 2344532202u, 4240017532u, 1658658271u, 366619977u, 2362670323u, 4224994405u, + 1303535960u, 984961486u, 2747007092u, 3569037538u, 1256170817u, 1037604311u, 2765210733u, 3554079995u, + 1131014506u, 879679996u, 2909243462u, 3663771856u, 1141124467u, 855842277u, 2852801631u, 3708648649u, + 1342533948u, 654459306u, 3188396048u, 3373015174u, 1466479909u, 544179635u, 3110523913u, 3462522015u, + 1591671054u, 702138776u, 2966460450u, 3352799412u, 1504918807u, 783551873u, 3082640443u, 3233442989u, + 3988292384u, 2596254646u, 62317068u, 1957810842u, 3939845945u, 2647816111u, 81470997u, 1943803523u, + 3814918930u, 2489596804u, 225274430u, 2053790376u, 3826175755u, 2466906013u, 167816743u, 2097651377u, + 4027552580u, 2265490386u, 503444072u, 1762050814u, 4150417245u, 2154129355u, 426522225u, 1852507879u, + 4275313526u, 2312317920u, 282753626u, 1742555852u, 4189708143u, 2394877945u, 397917763u, 1622183637u, + 3604390888u, 2714866558u, 953729732u, 1340076626u, 3518719985u, 2797360999u, 1068828381u, 1219638859u, + 3624741850u, 2936675148u, 906185462u, 1090812512u, 3747672003u, 2825379669u, 829329135u, 1181335161u, + 3412177804u, 3160834842u, 628085408u, 1382605366u, 3423369109u, 3138078467u, 570562233u, 1426400815u, + 3317316542u, 2998733608u, 733239954u, 1555261956u, 3268935591u, 3050360625u, 752459403u, 1541320221u, + 2607071920u, 3965973030u, 1969922972u, 40735498u, 2617837225u, 3943577151u, 1913087877u, 83908371u, + 2512341634u, 3803740692u, 2075208622u, 213261112u, 2463272603u, 3855990285u, 2094854071u, 198958881u, + 2262029012u, 4057260610u, 1759359992u, 534414190u, 2176718541u, 4139329115u, 1873836001u, 414664567u, + 2282248934u, 4279200368u, 1711684554u, 285281116u, 2405801727u, 4167216745u, 1634467795u, 376229701u, + 2685067896u, 3608007406u, 1308918612u, 956543938u, 2808555105u, 3495958263u, 1231636301u, 1047427035u, + 2932959818u, 3654703836u, 1088359270u, 936918000u, 2847714899u, 3736837829u, 1202900863u, 817233897u, + 3183342108u, 3401237130u, 1404277552u, 615818150u, 3134207493u, 3453421203u, 1423857449u, 601450431u, + 3009837614u, 3294710456u, 1567103746u, 711928724u, 3020668471u, 3272380065u, 1510334235u, 755167117u +}; + +/*Return the CRC of the bytes buf[0..len-1].*/ +unsigned lodepng_crc32(const unsigned char* data, size_t length) +{ + unsigned r = 0xffffffffu; + size_t i; + for(i = 0; i < length; ++i) + { + r = lodepng_crc32_table[(r ^ data[i]) & 0xff] ^ (r >> 8); + } + return r ^ 0xffffffffu; +} +#else /* !LODEPNG_NO_COMPILE_CRC */ +unsigned lodepng_crc32(const unsigned char* data, size_t length); +#endif /* !LODEPNG_NO_COMPILE_CRC */ + +/* ////////////////////////////////////////////////////////////////////////// */ +/* / Reading and writing single bits and bytes from/to stream for LodePNG / */ +/* ////////////////////////////////////////////////////////////////////////// */ + +static unsigned char readBitFromReversedStream(size_t* bitpointer, const unsigned char* bitstream) +{ + unsigned char result = (unsigned char)((bitstream[(*bitpointer) >> 3] >> (7 - ((*bitpointer) & 0x7))) & 1); + ++(*bitpointer); + return result; +} + +static unsigned readBitsFromReversedStream(size_t* bitpointer, const unsigned char* bitstream, size_t nbits) +{ + unsigned result = 0; + size_t i; + for(i = nbits - 1; i < nbits; --i) + { + result += (unsigned)readBitFromReversedStream(bitpointer, bitstream) << i; + } + return result; +} + +#ifdef LODEPNG_COMPILE_DECODER +static void setBitOfReversedStream0(size_t* bitpointer, unsigned char* bitstream, unsigned char bit) +{ + /*the current bit in bitstream must be 0 for this to work*/ + if(bit) + { + /*earlier bit of huffman code is in a lesser significant bit of an earlier byte*/ + bitstream[(*bitpointer) >> 3] |= (bit << (7 - ((*bitpointer) & 0x7))); + } + ++(*bitpointer); +} +#endif /*LODEPNG_COMPILE_DECODER*/ + +static void setBitOfReversedStream(size_t* bitpointer, unsigned char* bitstream, unsigned char bit) +{ + /*the current bit in bitstream may be 0 or 1 for this to work*/ + if(bit == 0) bitstream[(*bitpointer) >> 3] &= (unsigned char)(~(1 << (7 - ((*bitpointer) & 0x7)))); + else bitstream[(*bitpointer) >> 3] |= (1 << (7 - ((*bitpointer) & 0x7))); + ++(*bitpointer); +} + +/* ////////////////////////////////////////////////////////////////////////// */ +/* / PNG chunks / */ +/* ////////////////////////////////////////////////////////////////////////// */ + +unsigned lodepng_chunk_length(const unsigned char* chunk) +{ + return lodepng_read32bitInt(&chunk[0]); +} + +void lodepng_chunk_type(char type[5], const unsigned char* chunk) +{ + unsigned i; + for(i = 0; i != 4; ++i) type[i] = (char)chunk[4 + i]; + type[4] = 0; /*null termination char*/ +} + +unsigned char lodepng_chunk_type_equals(const unsigned char* chunk, const char* type) +{ + if(strlen(type) != 4) return 0; + return (chunk[4] == type[0] && chunk[5] == type[1] && chunk[6] == type[2] && chunk[7] == type[3]); +} + +unsigned char lodepng_chunk_ancillary(const unsigned char* chunk) +{ + return((chunk[4] & 32) != 0); +} + +unsigned char lodepng_chunk_private(const unsigned char* chunk) +{ + return((chunk[6] & 32) != 0); +} + +unsigned char lodepng_chunk_safetocopy(const unsigned char* chunk) +{ + return((chunk[7] & 32) != 0); +} + +unsigned char* lodepng_chunk_data(unsigned char* chunk) +{ + return &chunk[8]; +} + +const unsigned char* lodepng_chunk_data_const(const unsigned char* chunk) +{ + return &chunk[8]; +} + +unsigned lodepng_chunk_check_crc(const unsigned char* chunk) +{ + unsigned length = lodepng_chunk_length(chunk); + unsigned CRC = lodepng_read32bitInt(&chunk[length + 8]); + /*the CRC is taken of the data and the 4 chunk type letters, not the length*/ + unsigned checksum = lodepng_crc32(&chunk[4], length + 4); + if(CRC != checksum) return 1; + else return 0; +} + +void lodepng_chunk_generate_crc(unsigned char* chunk) +{ + unsigned length = lodepng_chunk_length(chunk); + unsigned CRC = lodepng_crc32(&chunk[4], length + 4); + lodepng_set32bitInt(chunk + 8 + length, CRC); +} + +unsigned char* lodepng_chunk_next(unsigned char* chunk) +{ + unsigned total_chunk_length = lodepng_chunk_length(chunk) + 12; + return &chunk[total_chunk_length]; +} + +const unsigned char* lodepng_chunk_next_const(const unsigned char* chunk) +{ + unsigned total_chunk_length = lodepng_chunk_length(chunk) + 12; + return &chunk[total_chunk_length]; +} + +unsigned lodepng_chunk_append(unsigned char** out, size_t* outlength, const unsigned char* chunk) +{ + unsigned i; + unsigned total_chunk_length = lodepng_chunk_length(chunk) + 12; + unsigned char *chunk_start, *new_buffer; + size_t new_length = (*outlength) + total_chunk_length; + if(new_length < total_chunk_length || new_length < (*outlength)) return 77; /*integer overflow happened*/ + + new_buffer = (unsigned char*)lodepng_realloc(*out, new_length); + if(!new_buffer) return 83; /*alloc fail*/ + (*out) = new_buffer; + (*outlength) = new_length; + chunk_start = &(*out)[new_length - total_chunk_length]; + + for(i = 0; i != total_chunk_length; ++i) chunk_start[i] = chunk[i]; + + return 0; +} + +unsigned lodepng_chunk_create(unsigned char** out, size_t* outlength, unsigned length, + const char* type, const unsigned char* data) +{ + unsigned i; + unsigned char *chunk, *new_buffer; + size_t new_length = (*outlength) + length + 12; + if(new_length < length + 12 || new_length < (*outlength)) return 77; /*integer overflow happened*/ + new_buffer = (unsigned char*)lodepng_realloc(*out, new_length); + if(!new_buffer) return 83; /*alloc fail*/ + (*out) = new_buffer; + (*outlength) = new_length; + chunk = &(*out)[(*outlength) - length - 12]; + + /*1: length*/ + lodepng_set32bitInt(chunk, (unsigned)length); + + /*2: chunk name (4 letters)*/ + chunk[4] = (unsigned char)type[0]; + chunk[5] = (unsigned char)type[1]; + chunk[6] = (unsigned char)type[2]; + chunk[7] = (unsigned char)type[3]; + + /*3: the data*/ + for(i = 0; i != length; ++i) chunk[8 + i] = data[i]; + + /*4: CRC (of the chunkname characters and the data)*/ + lodepng_chunk_generate_crc(chunk); + + return 0; +} + +/* ////////////////////////////////////////////////////////////////////////// */ +/* / Color types and such / */ +/* ////////////////////////////////////////////////////////////////////////// */ + +/*return type is a LodePNG error code*/ +static unsigned checkColorValidity(LodePNGColorType colortype, unsigned bd) /*bd = bitdepth*/ +{ + switch(colortype) + { + case 0: if(!(bd == 1 || bd == 2 || bd == 4 || bd == 8 || bd == 16)) return 37; break; /*grey*/ + case 2: if(!( bd == 8 || bd == 16)) return 37; break; /*RGB*/ + case 3: if(!(bd == 1 || bd == 2 || bd == 4 || bd == 8 )) return 37; break; /*palette*/ + case 4: if(!( bd == 8 || bd == 16)) return 37; break; /*grey + alpha*/ + case 6: if(!( bd == 8 || bd == 16)) return 37; break; /*RGBA*/ + default: return 31; + } + return 0; /*allowed color type / bits combination*/ +} + +static unsigned getNumColorChannels(LodePNGColorType colortype) +{ + switch(colortype) + { + case 0: return 1; /*grey*/ + case 2: return 3; /*RGB*/ + case 3: return 1; /*palette*/ + case 4: return 2; /*grey + alpha*/ + case 6: return 4; /*RGBA*/ + } + return 0; /*unexisting color type*/ +} + +static unsigned lodepng_get_bpp_lct(LodePNGColorType colortype, unsigned bitdepth) +{ + /*bits per pixel is amount of channels * bits per channel*/ + return getNumColorChannels(colortype) * bitdepth; +} + +/* ////////////////////////////////////////////////////////////////////////// */ + +void lodepng_color_mode_init(LodePNGColorMode* info) +{ + info->key_defined = 0; + info->key_r = info->key_g = info->key_b = 0; + info->colortype = LCT_RGBA; + info->bitdepth = 8; + info->palette = 0; + info->palettesize = 0; +} + +void lodepng_color_mode_cleanup(LodePNGColorMode* info) +{ + lodepng_palette_clear(info); +} + +unsigned lodepng_color_mode_copy(LodePNGColorMode* dest, const LodePNGColorMode* source) +{ + size_t i; + lodepng_color_mode_cleanup(dest); + *dest = *source; + if(source->palette) + { + dest->palette = (unsigned char*)lodepng_malloc(1024); + if(!dest->palette && source->palettesize) return 83; /*alloc fail*/ + for(i = 0; i != source->palettesize * 4; ++i) dest->palette[i] = source->palette[i]; + } + return 0; +} + +static int lodepng_color_mode_equal(const LodePNGColorMode* a, const LodePNGColorMode* b) +{ + size_t i; + if(a->colortype != b->colortype) return 0; + if(a->bitdepth != b->bitdepth) return 0; + if(a->key_defined != b->key_defined) return 0; + if(a->key_defined) + { + if(a->key_r != b->key_r) return 0; + if(a->key_g != b->key_g) return 0; + if(a->key_b != b->key_b) return 0; + } + /*if one of the palette sizes is 0, then we consider it to be the same as the + other: it means that e.g. the palette was not given by the user and should be + considered the same as the palette inside the PNG.*/ + if(1/*a->palettesize != 0 && b->palettesize != 0*/) { + if(a->palettesize != b->palettesize) return 0; + for(i = 0; i != a->palettesize * 4; ++i) + { + if(a->palette[i] != b->palette[i]) return 0; + } + } + return 1; +} + +void lodepng_palette_clear(LodePNGColorMode* info) +{ + if(info->palette) lodepng_free(info->palette); + info->palette = 0; + info->palettesize = 0; +} + +unsigned lodepng_palette_add(LodePNGColorMode* info, + unsigned char r, unsigned char g, unsigned char b, unsigned char a) +{ + unsigned char* data; + /*the same resize technique as C++ std::vectors is used, and here it's made so that for a palette with + the max of 256 colors, it'll have the exact alloc size*/ + if(!info->palette) /*allocate palette if empty*/ + { + /*room for 256 colors with 4 bytes each*/ + data = (unsigned char*)lodepng_realloc(info->palette, 1024); + if(!data) return 83; /*alloc fail*/ + else info->palette = data; + } + info->palette[4 * info->palettesize + 0] = r; + info->palette[4 * info->palettesize + 1] = g; + info->palette[4 * info->palettesize + 2] = b; + info->palette[4 * info->palettesize + 3] = a; + ++info->palettesize; + return 0; +} + +unsigned lodepng_get_bpp(const LodePNGColorMode* info) +{ + /*calculate bits per pixel out of colortype and bitdepth*/ + return lodepng_get_bpp_lct(info->colortype, info->bitdepth); +} + +unsigned lodepng_get_channels(const LodePNGColorMode* info) +{ + return getNumColorChannels(info->colortype); +} + +unsigned lodepng_is_greyscale_type(const LodePNGColorMode* info) +{ + return info->colortype == LCT_GREY || info->colortype == LCT_GREY_ALPHA; +} + +unsigned lodepng_is_alpha_type(const LodePNGColorMode* info) +{ + return (info->colortype & 4) != 0; /*4 or 6*/ +} + +unsigned lodepng_is_palette_type(const LodePNGColorMode* info) +{ + return info->colortype == LCT_PALETTE; +} + +unsigned lodepng_has_palette_alpha(const LodePNGColorMode* info) +{ + size_t i; + for(i = 0; i != info->palettesize; ++i) + { + if(info->palette[i * 4 + 3] < 255) return 1; + } + return 0; +} + +unsigned lodepng_can_have_alpha(const LodePNGColorMode* info) +{ + return info->key_defined + || lodepng_is_alpha_type(info) + || lodepng_has_palette_alpha(info); +} + +size_t lodepng_get_raw_size(unsigned w, unsigned h, const LodePNGColorMode* color) +{ + /*will not overflow for any color type if roughly w * h < 268435455*/ + int bpp = lodepng_get_bpp(color); + size_t n = w * h; + return ((n / 8) * bpp) + ((n & 7) * bpp + 7) / 8; +} + +size_t lodepng_get_raw_size_lct(unsigned w, unsigned h, LodePNGColorType colortype, unsigned bitdepth) +{ + /*will not overflow for any color type if roughly w * h < 268435455*/ + int bpp = lodepng_get_bpp_lct(colortype, bitdepth); + size_t n = w * h; + return ((n / 8) * bpp) + ((n & 7) * bpp + 7) / 8; +} + + +#ifdef LODEPNG_COMPILE_PNG +#ifdef LODEPNG_COMPILE_DECODER +/*in an idat chunk, each scanline is a multiple of 8 bits, unlike the lodepng output buffer*/ +static size_t lodepng_get_raw_size_idat(unsigned w, unsigned h, const LodePNGColorMode* color) +{ + /*will not overflow for any color type if roughly w * h < 268435455*/ + int bpp = lodepng_get_bpp(color); + size_t line = ((w / 8) * bpp) + ((w & 7) * bpp + 7) / 8; + return h * line; +} +#endif /*LODEPNG_COMPILE_DECODER*/ +#endif /*LODEPNG_COMPILE_PNG*/ + +#ifdef LODEPNG_COMPILE_ANCILLARY_CHUNKS + +static void LodePNGUnknownChunks_init(LodePNGInfo* info) +{ + unsigned i; + for(i = 0; i != 3; ++i) info->unknown_chunks_data[i] = 0; + for(i = 0; i != 3; ++i) info->unknown_chunks_size[i] = 0; +} + +static void LodePNGUnknownChunks_cleanup(LodePNGInfo* info) +{ + unsigned i; + for(i = 0; i != 3; ++i) lodepng_free(info->unknown_chunks_data[i]); +} + +static unsigned LodePNGUnknownChunks_copy(LodePNGInfo* dest, const LodePNGInfo* src) +{ + unsigned i; + + LodePNGUnknownChunks_cleanup(dest); + + for(i = 0; i != 3; ++i) + { + size_t j; + dest->unknown_chunks_size[i] = src->unknown_chunks_size[i]; + dest->unknown_chunks_data[i] = (unsigned char*)lodepng_malloc(src->unknown_chunks_size[i]); + if(!dest->unknown_chunks_data[i] && dest->unknown_chunks_size[i]) return 83; /*alloc fail*/ + for(j = 0; j < src->unknown_chunks_size[i]; ++j) + { + dest->unknown_chunks_data[i][j] = src->unknown_chunks_data[i][j]; + } + } + + return 0; +} + +/******************************************************************************/ + +static void LodePNGText_init(LodePNGInfo* info) +{ + info->text_num = 0; + info->text_keys = NULL; + info->text_strings = NULL; +} + +static void LodePNGText_cleanup(LodePNGInfo* info) +{ + size_t i; + for(i = 0; i != info->text_num; ++i) + { + string_cleanup(&info->text_keys[i]); + string_cleanup(&info->text_strings[i]); + } + lodepng_free(info->text_keys); + lodepng_free(info->text_strings); +} + +static unsigned LodePNGText_copy(LodePNGInfo* dest, const LodePNGInfo* source) +{ + size_t i = 0; + dest->text_keys = 0; + dest->text_strings = 0; + dest->text_num = 0; + for(i = 0; i != source->text_num; ++i) + { + CERROR_TRY_RETURN(lodepng_add_text(dest, source->text_keys[i], source->text_strings[i])); + } + return 0; +} + +void lodepng_clear_text(LodePNGInfo* info) +{ + LodePNGText_cleanup(info); +} + +unsigned lodepng_add_text(LodePNGInfo* info, const char* key, const char* str) +{ + char** new_keys = (char**)(lodepng_realloc(info->text_keys, sizeof(char*) * (info->text_num + 1))); + char** new_strings = (char**)(lodepng_realloc(info->text_strings, sizeof(char*) * (info->text_num + 1))); + if(!new_keys || !new_strings) + { + lodepng_free(new_keys); + lodepng_free(new_strings); + return 83; /*alloc fail*/ + } + + ++info->text_num; + info->text_keys = new_keys; + info->text_strings = new_strings; + + string_init(&info->text_keys[info->text_num - 1]); + string_set(&info->text_keys[info->text_num - 1], key); + + string_init(&info->text_strings[info->text_num - 1]); + string_set(&info->text_strings[info->text_num - 1], str); + + return 0; +} + +/******************************************************************************/ + +static void LodePNGIText_init(LodePNGInfo* info) +{ + info->itext_num = 0; + info->itext_keys = NULL; + info->itext_langtags = NULL; + info->itext_transkeys = NULL; + info->itext_strings = NULL; +} + +static void LodePNGIText_cleanup(LodePNGInfo* info) +{ + size_t i; + for(i = 0; i != info->itext_num; ++i) + { + string_cleanup(&info->itext_keys[i]); + string_cleanup(&info->itext_langtags[i]); + string_cleanup(&info->itext_transkeys[i]); + string_cleanup(&info->itext_strings[i]); + } + lodepng_free(info->itext_keys); + lodepng_free(info->itext_langtags); + lodepng_free(info->itext_transkeys); + lodepng_free(info->itext_strings); +} + +static unsigned LodePNGIText_copy(LodePNGInfo* dest, const LodePNGInfo* source) +{ + size_t i = 0; + dest->itext_keys = 0; + dest->itext_langtags = 0; + dest->itext_transkeys = 0; + dest->itext_strings = 0; + dest->itext_num = 0; + for(i = 0; i != source->itext_num; ++i) + { + CERROR_TRY_RETURN(lodepng_add_itext(dest, source->itext_keys[i], source->itext_langtags[i], + source->itext_transkeys[i], source->itext_strings[i])); + } + return 0; +} + +void lodepng_clear_itext(LodePNGInfo* info) +{ + LodePNGIText_cleanup(info); +} + +unsigned lodepng_add_itext(LodePNGInfo* info, const char* key, const char* langtag, + const char* transkey, const char* str) +{ + char** new_keys = (char**)(lodepng_realloc(info->itext_keys, sizeof(char*) * (info->itext_num + 1))); + char** new_langtags = (char**)(lodepng_realloc(info->itext_langtags, sizeof(char*) * (info->itext_num + 1))); + char** new_transkeys = (char**)(lodepng_realloc(info->itext_transkeys, sizeof(char*) * (info->itext_num + 1))); + char** new_strings = (char**)(lodepng_realloc(info->itext_strings, sizeof(char*) * (info->itext_num + 1))); + if(!new_keys || !new_langtags || !new_transkeys || !new_strings) + { + lodepng_free(new_keys); + lodepng_free(new_langtags); + lodepng_free(new_transkeys); + lodepng_free(new_strings); + return 83; /*alloc fail*/ + } + + ++info->itext_num; + info->itext_keys = new_keys; + info->itext_langtags = new_langtags; + info->itext_transkeys = new_transkeys; + info->itext_strings = new_strings; + + string_init(&info->itext_keys[info->itext_num - 1]); + string_set(&info->itext_keys[info->itext_num - 1], key); + + string_init(&info->itext_langtags[info->itext_num - 1]); + string_set(&info->itext_langtags[info->itext_num - 1], langtag); + + string_init(&info->itext_transkeys[info->itext_num - 1]); + string_set(&info->itext_transkeys[info->itext_num - 1], transkey); + + string_init(&info->itext_strings[info->itext_num - 1]); + string_set(&info->itext_strings[info->itext_num - 1], str); + + return 0; +} +#endif /*LODEPNG_COMPILE_ANCILLARY_CHUNKS*/ + +void lodepng_info_init(LodePNGInfo* info) +{ + lodepng_color_mode_init(&info->color); + info->interlace_method = 0; + info->compression_method = 0; + info->filter_method = 0; +#ifdef LODEPNG_COMPILE_ANCILLARY_CHUNKS + info->background_defined = 0; + info->background_r = info->background_g = info->background_b = 0; + + LodePNGText_init(info); + LodePNGIText_init(info); + + info->time_defined = 0; + info->phys_defined = 0; + + LodePNGUnknownChunks_init(info); +#endif /*LODEPNG_COMPILE_ANCILLARY_CHUNKS*/ +} + +void lodepng_info_cleanup(LodePNGInfo* info) +{ + lodepng_color_mode_cleanup(&info->color); +#ifdef LODEPNG_COMPILE_ANCILLARY_CHUNKS + LodePNGText_cleanup(info); + LodePNGIText_cleanup(info); + + LodePNGUnknownChunks_cleanup(info); +#endif /*LODEPNG_COMPILE_ANCILLARY_CHUNKS*/ +} + +unsigned lodepng_info_copy(LodePNGInfo* dest, const LodePNGInfo* source) +{ + lodepng_info_cleanup(dest); + *dest = *source; + lodepng_color_mode_init(&dest->color); + CERROR_TRY_RETURN(lodepng_color_mode_copy(&dest->color, &source->color)); + +#ifdef LODEPNG_COMPILE_ANCILLARY_CHUNKS + CERROR_TRY_RETURN(LodePNGText_copy(dest, source)); + CERROR_TRY_RETURN(LodePNGIText_copy(dest, source)); + + LodePNGUnknownChunks_init(dest); + CERROR_TRY_RETURN(LodePNGUnknownChunks_copy(dest, source)); +#endif /*LODEPNG_COMPILE_ANCILLARY_CHUNKS*/ + return 0; +} + +void lodepng_info_swap(LodePNGInfo* a, LodePNGInfo* b) +{ + LodePNGInfo temp = *a; + *a = *b; + *b = temp; +} + +/* ////////////////////////////////////////////////////////////////////////// */ + +/*index: bitgroup index, bits: bitgroup size(1, 2 or 4), in: bitgroup value, out: octet array to add bits to*/ +static void addColorBits(unsigned char* out, size_t index, unsigned bits, unsigned in) +{ + unsigned m = bits == 1 ? 7 : bits == 2 ? 3 : 1; /*8 / bits - 1*/ + /*p = the partial index in the byte, e.g. with 4 palettebits it is 0 for first half or 1 for second half*/ + unsigned p = index & m; + in &= (1u << bits) - 1u; /*filter out any other bits of the input value*/ + in = in << (bits * (m - p)); + if(p == 0) out[index * bits / 8] = in; + else out[index * bits / 8] |= in; +} + +typedef struct ColorTree ColorTree; + +/* +One node of a color tree +This is the data structure used to count the number of unique colors and to get a palette +index for a color. It's like an octree, but because the alpha channel is used too, each +node has 16 instead of 8 children. +*/ +struct ColorTree +{ + ColorTree* children[16]; /*up to 16 pointers to ColorTree of next level*/ + int index; /*the payload. Only has a meaningful value if this is in the last level*/ +}; + +static void color_tree_init(ColorTree* tree) +{ + int i; + for(i = 0; i != 16; ++i) tree->children[i] = 0; + tree->index = -1; +} + +static void color_tree_cleanup(ColorTree* tree) +{ + int i; + for(i = 0; i != 16; ++i) + { + if(tree->children[i]) + { + color_tree_cleanup(tree->children[i]); + lodepng_free(tree->children[i]); + } + } +} + +/*returns -1 if color not present, its index otherwise*/ +static int color_tree_get(ColorTree* tree, unsigned char r, unsigned char g, unsigned char b, unsigned char a) +{ + int bit = 0; + for(bit = 0; bit < 8; ++bit) + { + int i = 8 * ((r >> bit) & 1) + 4 * ((g >> bit) & 1) + 2 * ((b >> bit) & 1) + 1 * ((a >> bit) & 1); + if(!tree->children[i]) return -1; + else tree = tree->children[i]; + } + return tree ? tree->index : -1; +} + +#ifdef LODEPNG_COMPILE_ENCODER +static int color_tree_has(ColorTree* tree, unsigned char r, unsigned char g, unsigned char b, unsigned char a) +{ + return color_tree_get(tree, r, g, b, a) >= 0; +} +#endif /*LODEPNG_COMPILE_ENCODER*/ + +/*color is not allowed to already exist. +Index should be >= 0 (it's signed to be compatible with using -1 for "doesn't exist")*/ +static void color_tree_add(ColorTree* tree, + unsigned char r, unsigned char g, unsigned char b, unsigned char a, unsigned index) +{ + int bit; + for(bit = 0; bit < 8; ++bit) + { + int i = 8 * ((r >> bit) & 1) + 4 * ((g >> bit) & 1) + 2 * ((b >> bit) & 1) + 1 * ((a >> bit) & 1); + if(!tree->children[i]) + { + tree->children[i] = (ColorTree*)lodepng_malloc(sizeof(ColorTree)); + color_tree_init(tree->children[i]); + } + tree = tree->children[i]; + } + tree->index = (int)index; +} + +/*put a pixel, given its RGBA color, into image of any color type*/ +static unsigned rgba8ToPixel(unsigned char* out, size_t i, + const LodePNGColorMode* mode, ColorTree* tree /*for palette*/, + unsigned char r, unsigned char g, unsigned char b, unsigned char a) +{ + if(mode->colortype == LCT_GREY) + { + unsigned char grey = r; /*((unsigned short)r + g + b) / 3*/; + if(mode->bitdepth == 8) out[i] = grey; + else if(mode->bitdepth == 16) out[i * 2 + 0] = out[i * 2 + 1] = grey; + else + { + /*take the most significant bits of grey*/ + grey = (grey >> (8 - mode->bitdepth)) & ((1 << mode->bitdepth) - 1); + addColorBits(out, i, mode->bitdepth, grey); + } + } + else if(mode->colortype == LCT_RGB) + { + if(mode->bitdepth == 8) + { + out[i * 3 + 0] = r; + out[i * 3 + 1] = g; + out[i * 3 + 2] = b; + } + else + { + out[i * 6 + 0] = out[i * 6 + 1] = r; + out[i * 6 + 2] = out[i * 6 + 3] = g; + out[i * 6 + 4] = out[i * 6 + 5] = b; + } + } + else if(mode->colortype == LCT_PALETTE) + { + int index = color_tree_get(tree, r, g, b, a); + if(index < 0) return 82; /*color not in palette*/ + if(mode->bitdepth == 8) out[i] = index; + else addColorBits(out, i, mode->bitdepth, (unsigned)index); + } + else if(mode->colortype == LCT_GREY_ALPHA) + { + unsigned char grey = r; /*((unsigned short)r + g + b) / 3*/; + if(mode->bitdepth == 8) + { + out[i * 2 + 0] = grey; + out[i * 2 + 1] = a; + } + else if(mode->bitdepth == 16) + { + out[i * 4 + 0] = out[i * 4 + 1] = grey; + out[i * 4 + 2] = out[i * 4 + 3] = a; + } + } + else if(mode->colortype == LCT_RGBA) + { + if(mode->bitdepth == 8) + { + out[i * 4 + 0] = r; + out[i * 4 + 1] = g; + out[i * 4 + 2] = b; + out[i * 4 + 3] = a; + } + else + { + out[i * 8 + 0] = out[i * 8 + 1] = r; + out[i * 8 + 2] = out[i * 8 + 3] = g; + out[i * 8 + 4] = out[i * 8 + 5] = b; + out[i * 8 + 6] = out[i * 8 + 7] = a; + } + } + + return 0; /*no error*/ +} + +/*put a pixel, given its RGBA16 color, into image of any color 16-bitdepth type*/ +static void rgba16ToPixel(unsigned char* out, size_t i, + const LodePNGColorMode* mode, + unsigned short r, unsigned short g, unsigned short b, unsigned short a) +{ + if(mode->colortype == LCT_GREY) + { + unsigned short grey = r; /*((unsigned)r + g + b) / 3*/; + out[i * 2 + 0] = (grey >> 8) & 255; + out[i * 2 + 1] = grey & 255; + } + else if(mode->colortype == LCT_RGB) + { + out[i * 6 + 0] = (r >> 8) & 255; + out[i * 6 + 1] = r & 255; + out[i * 6 + 2] = (g >> 8) & 255; + out[i * 6 + 3] = g & 255; + out[i * 6 + 4] = (b >> 8) & 255; + out[i * 6 + 5] = b & 255; + } + else if(mode->colortype == LCT_GREY_ALPHA) + { + unsigned short grey = r; /*((unsigned)r + g + b) / 3*/; + out[i * 4 + 0] = (grey >> 8) & 255; + out[i * 4 + 1] = grey & 255; + out[i * 4 + 2] = (a >> 8) & 255; + out[i * 4 + 3] = a & 255; + } + else if(mode->colortype == LCT_RGBA) + { + out[i * 8 + 0] = (r >> 8) & 255; + out[i * 8 + 1] = r & 255; + out[i * 8 + 2] = (g >> 8) & 255; + out[i * 8 + 3] = g & 255; + out[i * 8 + 4] = (b >> 8) & 255; + out[i * 8 + 5] = b & 255; + out[i * 8 + 6] = (a >> 8) & 255; + out[i * 8 + 7] = a & 255; + } +} + +/*Get RGBA8 color of pixel with index i (y * width + x) from the raw image with given color type.*/ +static void getPixelColorRGBA8(unsigned char* r, unsigned char* g, + unsigned char* b, unsigned char* a, + const unsigned char* in, size_t i, + const LodePNGColorMode* mode) +{ + if(mode->colortype == LCT_GREY) + { + if(mode->bitdepth == 8) + { + *r = *g = *b = in[i]; + if(mode->key_defined && *r == mode->key_r) *a = 0; + else *a = 255; + } + else if(mode->bitdepth == 16) + { + *r = *g = *b = in[i * 2 + 0]; + if(mode->key_defined && 256U * in[i * 2 + 0] + in[i * 2 + 1] == mode->key_r) *a = 0; + else *a = 255; + } + else + { + unsigned highest = ((1U << mode->bitdepth) - 1U); /*highest possible value for this bit depth*/ + size_t j = i * mode->bitdepth; + unsigned value = readBitsFromReversedStream(&j, in, mode->bitdepth); + *r = *g = *b = (value * 255) / highest; + if(mode->key_defined && value == mode->key_r) *a = 0; + else *a = 255; + } + } + else if(mode->colortype == LCT_RGB) + { + if(mode->bitdepth == 8) + { + *r = in[i * 3 + 0]; *g = in[i * 3 + 1]; *b = in[i * 3 + 2]; + if(mode->key_defined && *r == mode->key_r && *g == mode->key_g && *b == mode->key_b) *a = 0; + else *a = 255; + } + else + { + *r = in[i * 6 + 0]; + *g = in[i * 6 + 2]; + *b = in[i * 6 + 4]; + if(mode->key_defined && 256U * in[i * 6 + 0] + in[i * 6 + 1] == mode->key_r + && 256U * in[i * 6 + 2] + in[i * 6 + 3] == mode->key_g + && 256U * in[i * 6 + 4] + in[i * 6 + 5] == mode->key_b) *a = 0; + else *a = 255; + } + } + else if(mode->colortype == LCT_PALETTE) + { + unsigned index; + if(mode->bitdepth == 8) index = in[i]; + else + { + size_t j = i * mode->bitdepth; + index = readBitsFromReversedStream(&j, in, mode->bitdepth); + } + + if(index >= mode->palettesize) + { + /*This is an error according to the PNG spec, but common PNG decoders make it black instead. + Done here too, slightly faster due to no error handling needed.*/ + *r = *g = *b = 0; + *a = 255; + } + else + { + *r = mode->palette[index * 4 + 0]; + *g = mode->palette[index * 4 + 1]; + *b = mode->palette[index * 4 + 2]; + *a = mode->palette[index * 4 + 3]; + } + } + else if(mode->colortype == LCT_GREY_ALPHA) + { + if(mode->bitdepth == 8) + { + *r = *g = *b = in[i * 2 + 0]; + *a = in[i * 2 + 1]; + } + else + { + *r = *g = *b = in[i * 4 + 0]; + *a = in[i * 4 + 2]; + } + } + else if(mode->colortype == LCT_RGBA) + { + if(mode->bitdepth == 8) + { + *r = in[i * 4 + 0]; + *g = in[i * 4 + 1]; + *b = in[i * 4 + 2]; + *a = in[i * 4 + 3]; + } + else + { + *r = in[i * 8 + 0]; + *g = in[i * 8 + 2]; + *b = in[i * 8 + 4]; + *a = in[i * 8 + 6]; + } + } +} + +/*Similar to getPixelColorRGBA8, but with all the for loops inside of the color +mode test cases, optimized to convert the colors much faster, when converting +to RGBA or RGB with 8 bit per cannel. buffer must be RGBA or RGB output with +enough memory, if has_alpha is true the output is RGBA. mode has the color mode +of the input buffer.*/ +static void getPixelColorsRGBA8(unsigned char* buffer, size_t numpixels, + unsigned has_alpha, const unsigned char* in, + const LodePNGColorMode* mode) +{ + unsigned num_channels = has_alpha ? 4 : 3; + size_t i; + if(mode->colortype == LCT_GREY) + { + if(mode->bitdepth == 8) + { + for(i = 0; i != numpixels; ++i, buffer += num_channels) + { + buffer[0] = buffer[1] = buffer[2] = in[i]; + if(has_alpha) buffer[3] = mode->key_defined && in[i] == mode->key_r ? 0 : 255; + } + } + else if(mode->bitdepth == 16) + { + for(i = 0; i != numpixels; ++i, buffer += num_channels) + { + buffer[0] = buffer[1] = buffer[2] = in[i * 2]; + if(has_alpha) buffer[3] = mode->key_defined && 256U * in[i * 2 + 0] + in[i * 2 + 1] == mode->key_r ? 0 : 255; + } + } + else + { + unsigned highest = ((1U << mode->bitdepth) - 1U); /*highest possible value for this bit depth*/ + size_t j = 0; + for(i = 0; i != numpixels; ++i, buffer += num_channels) + { + unsigned value = readBitsFromReversedStream(&j, in, mode->bitdepth); + buffer[0] = buffer[1] = buffer[2] = (value * 255) / highest; + if(has_alpha) buffer[3] = mode->key_defined && value == mode->key_r ? 0 : 255; + } + } + } + else if(mode->colortype == LCT_RGB) + { + if(mode->bitdepth == 8) + { + for(i = 0; i != numpixels; ++i, buffer += num_channels) + { + buffer[0] = in[i * 3 + 0]; + buffer[1] = in[i * 3 + 1]; + buffer[2] = in[i * 3 + 2]; + if(has_alpha) buffer[3] = mode->key_defined && buffer[0] == mode->key_r + && buffer[1]== mode->key_g && buffer[2] == mode->key_b ? 0 : 255; + } + } + else + { + for(i = 0; i != numpixels; ++i, buffer += num_channels) + { + buffer[0] = in[i * 6 + 0]; + buffer[1] = in[i * 6 + 2]; + buffer[2] = in[i * 6 + 4]; + if(has_alpha) buffer[3] = mode->key_defined + && 256U * in[i * 6 + 0] + in[i * 6 + 1] == mode->key_r + && 256U * in[i * 6 + 2] + in[i * 6 + 3] == mode->key_g + && 256U * in[i * 6 + 4] + in[i * 6 + 5] == mode->key_b ? 0 : 255; + } + } + } + else if(mode->colortype == LCT_PALETTE) + { + unsigned index; + size_t j = 0; + for(i = 0; i != numpixels; ++i, buffer += num_channels) + { + if(mode->bitdepth == 8) index = in[i]; + else index = readBitsFromReversedStream(&j, in, mode->bitdepth); + + if(index >= mode->palettesize) + { + /*This is an error according to the PNG spec, but most PNG decoders make it black instead. + Done here too, slightly faster due to no error handling needed.*/ + buffer[0] = buffer[1] = buffer[2] = 0; + if(has_alpha) buffer[3] = 255; + } + else + { + buffer[0] = mode->palette[index * 4 + 0]; + buffer[1] = mode->palette[index * 4 + 1]; + buffer[2] = mode->palette[index * 4 + 2]; + if(has_alpha) buffer[3] = mode->palette[index * 4 + 3]; + } + } + } + else if(mode->colortype == LCT_GREY_ALPHA) + { + if(mode->bitdepth == 8) + { + for(i = 0; i != numpixels; ++i, buffer += num_channels) + { + buffer[0] = buffer[1] = buffer[2] = in[i * 2 + 0]; + if(has_alpha) buffer[3] = in[i * 2 + 1]; + } + } + else + { + for(i = 0; i != numpixels; ++i, buffer += num_channels) + { + buffer[0] = buffer[1] = buffer[2] = in[i * 4 + 0]; + if(has_alpha) buffer[3] = in[i * 4 + 2]; + } + } + } + else if(mode->colortype == LCT_RGBA) + { + if(mode->bitdepth == 8) + { + for(i = 0; i != numpixels; ++i, buffer += num_channels) + { + buffer[0] = in[i * 4 + 0]; + buffer[1] = in[i * 4 + 1]; + buffer[2] = in[i * 4 + 2]; + if(has_alpha) buffer[3] = in[i * 4 + 3]; + } + } + else + { + for(i = 0; i != numpixels; ++i, buffer += num_channels) + { + buffer[0] = in[i * 8 + 0]; + buffer[1] = in[i * 8 + 2]; + buffer[2] = in[i * 8 + 4]; + if(has_alpha) buffer[3] = in[i * 8 + 6]; + } + } + } +} + +/*Get RGBA16 color of pixel with index i (y * width + x) from the raw image with +given color type, but the given color type must be 16-bit itself.*/ +static void getPixelColorRGBA16(unsigned short* r, unsigned short* g, unsigned short* b, unsigned short* a, + const unsigned char* in, size_t i, const LodePNGColorMode* mode) +{ + if(mode->colortype == LCT_GREY) + { + *r = *g = *b = 256 * in[i * 2 + 0] + in[i * 2 + 1]; + if(mode->key_defined && 256U * in[i * 2 + 0] + in[i * 2 + 1] == mode->key_r) *a = 0; + else *a = 65535; + } + else if(mode->colortype == LCT_RGB) + { + *r = 256u * in[i * 6 + 0] + in[i * 6 + 1]; + *g = 256u * in[i * 6 + 2] + in[i * 6 + 3]; + *b = 256u * in[i * 6 + 4] + in[i * 6 + 5]; + if(mode->key_defined + && 256u * in[i * 6 + 0] + in[i * 6 + 1] == mode->key_r + && 256u * in[i * 6 + 2] + in[i * 6 + 3] == mode->key_g + && 256u * in[i * 6 + 4] + in[i * 6 + 5] == mode->key_b) *a = 0; + else *a = 65535; + } + else if(mode->colortype == LCT_GREY_ALPHA) + { + *r = *g = *b = 256u * in[i * 4 + 0] + in[i * 4 + 1]; + *a = 256u * in[i * 4 + 2] + in[i * 4 + 3]; + } + else if(mode->colortype == LCT_RGBA) + { + *r = 256u * in[i * 8 + 0] + in[i * 8 + 1]; + *g = 256u * in[i * 8 + 2] + in[i * 8 + 3]; + *b = 256u * in[i * 8 + 4] + in[i * 8 + 5]; + *a = 256u * in[i * 8 + 6] + in[i * 8 + 7]; + } +} + +unsigned lodepng_convert(unsigned char* out, const unsigned char* in, + const LodePNGColorMode* mode_out, const LodePNGColorMode* mode_in, + unsigned w, unsigned h) +{ + int i; + ColorTree tree; + size_t numpixels = w * h; + + if(lodepng_color_mode_equal(mode_out, mode_in)) + { + size_t numbytes = lodepng_get_raw_size(w, h, mode_in); + for(i = 0; i != numbytes; ++i) out[i] = in[i]; + return 0; + } + + if(mode_out->colortype == LCT_PALETTE) + { + size_t palettesize = mode_out->palettesize; + const unsigned char* palette = mode_out->palette; + size_t palsize = 1u << mode_out->bitdepth; + /*if the user specified output palette but did not give the values, assume + they want the values of the input color type (assuming that one is palette). + Note that we never create a new palette ourselves.*/ + if(palettesize == 0) + { + palettesize = mode_in->palettesize; + palette = mode_in->palette; + } + if(palettesize < palsize) palsize = palettesize; + color_tree_init(&tree); + for(i = 0; i != palsize; ++i) + { + const unsigned char* p = &palette[i * 4]; + color_tree_add(&tree, p[0], p[1], p[2], p[3], i); + } + } + + if(mode_in->bitdepth == 16 && mode_out->bitdepth == 16) + { + for(i = 0; i != numpixels; ++i) + { + unsigned short r = 0, g = 0, b = 0, a = 0; + getPixelColorRGBA16(&r, &g, &b, &a, in, i, mode_in); + rgba16ToPixel(out, i, mode_out, r, g, b, a); + } + } + else if(mode_out->bitdepth == 8 && mode_out->colortype == LCT_RGBA) + { + getPixelColorsRGBA8(out, numpixels, 1, in, mode_in); + } + else if(mode_out->bitdepth == 8 && mode_out->colortype == LCT_RGB) + { + getPixelColorsRGBA8(out, numpixels, 0, in, mode_in); + } + else + { + unsigned char r = 0, g = 0, b = 0, a = 0; + for(i = 0; i != numpixels; ++i) + { + getPixelColorRGBA8(&r, &g, &b, &a, in, i, mode_in); + CERROR_TRY_RETURN(rgba8ToPixel(out, i, mode_out, &tree, r, g, b, a)); + } + } + + if(mode_out->colortype == LCT_PALETTE) + { + color_tree_cleanup(&tree); + } + + return 0; /*no error*/ +} + +#ifdef LODEPNG_COMPILE_ENCODER + +void lodepng_color_profile_init(LodePNGColorProfile* profile) +{ + profile->colored = 0; + profile->key = 0; + profile->alpha = 0; + profile->key_r = profile->key_g = profile->key_b = 0; + profile->numcolors = 0; + profile->bits = 1; +} + +/*function used for debug purposes with C++*/ +/*void printColorProfile(LodePNGColorProfile* p) +{ + std::cout << "colored: " << (int)p->colored << ", "; + std::cout << "key: " << (int)p->key << ", "; + std::cout << "key_r: " << (int)p->key_r << ", "; + std::cout << "key_g: " << (int)p->key_g << ", "; + std::cout << "key_b: " << (int)p->key_b << ", "; + std::cout << "alpha: " << (int)p->alpha << ", "; + std::cout << "numcolors: " << (int)p->numcolors << ", "; + std::cout << "bits: " << (int)p->bits << std::endl; +}*/ + +/*Returns how many bits needed to represent given value (max 8 bit)*/ +static unsigned getValueRequiredBits(unsigned char value) +{ + if(value == 0 || value == 255) return 1; + /*The scaling of 2-bit and 4-bit values uses multiples of 85 and 17*/ + if(value % 17 == 0) return value % 85 == 0 ? 2 : 4; + return 8; +} + +/*profile must already have been inited with mode. +It's ok to set some parameters of profile to done already.*/ +unsigned lodepng_get_color_profile(LodePNGColorProfile* profile, + const unsigned char* in, unsigned w, unsigned h, + const LodePNGColorMode* mode) +{ + unsigned error = 0; + size_t i; + ColorTree tree; + size_t numpixels = w * h; + + unsigned colored_done = lodepng_is_greyscale_type(mode) ? 1 : 0; + unsigned alpha_done = lodepng_can_have_alpha(mode) ? 0 : 1; + unsigned numcolors_done = 0; + unsigned bpp = lodepng_get_bpp(mode); + unsigned bits_done = bpp == 1 ? 1 : 0; + unsigned maxnumcolors = 257; + unsigned sixteen = 0; + if(bpp <= 8) maxnumcolors = bpp == 1 ? 2 : (bpp == 2 ? 4 : (bpp == 4 ? 16 : 256)); + + color_tree_init(&tree); + + /*Check if the 16-bit input is truly 16-bit*/ + if(mode->bitdepth == 16) + { + unsigned short r, g, b, a; + for(i = 0; i != numpixels; ++i) + { + getPixelColorRGBA16(&r, &g, &b, &a, in, i, mode); + if((r & 255) != ((r >> 8) & 255) || (g & 255) != ((g >> 8) & 255) || + (b & 255) != ((b >> 8) & 255) || (a & 255) != ((a >> 8) & 255)) /*first and second byte differ*/ + { + sixteen = 1; + break; + } + } + } + + if(sixteen) + { + unsigned short r = 0, g = 0, b = 0, a = 0; + profile->bits = 16; + bits_done = numcolors_done = 1; /*counting colors no longer useful, palette doesn't support 16-bit*/ + + for(i = 0; i != numpixels; ++i) + { + getPixelColorRGBA16(&r, &g, &b, &a, in, i, mode); + + if(!colored_done && (r != g || r != b)) + { + profile->colored = 1; + colored_done = 1; + } + + if(!alpha_done) + { + unsigned matchkey = (r == profile->key_r && g == profile->key_g && b == profile->key_b); + if(a != 65535 && (a != 0 || (profile->key && !matchkey))) + { + profile->alpha = 1; + alpha_done = 1; + if(profile->bits < 8) profile->bits = 8; /*PNG has no alphachannel modes with less than 8-bit per channel*/ + } + else if(a == 0 && !profile->alpha && !profile->key) + { + profile->key = 1; + profile->key_r = r; + profile->key_g = g; + profile->key_b = b; + } + else if(a == 65535 && profile->key && matchkey) + { + /* Color key cannot be used if an opaque pixel also has that RGB color. */ + profile->alpha = 1; + alpha_done = 1; + } + } + + if(alpha_done && numcolors_done && colored_done && bits_done) break; + } + } + else /* < 16-bit */ + { + for(i = 0; i != numpixels; ++i) + { + unsigned char r = 0, g = 0, b = 0, a = 0; + getPixelColorRGBA8(&r, &g, &b, &a, in, i, mode); + + if(!bits_done && profile->bits < 8) + { + /*only r is checked, < 8 bits is only relevant for greyscale*/ + unsigned bits = getValueRequiredBits(r); + if(bits > profile->bits) profile->bits = bits; + } + bits_done = (profile->bits >= bpp); + + if(!colored_done && (r != g || r != b)) + { + profile->colored = 1; + colored_done = 1; + if(profile->bits < 8) profile->bits = 8; /*PNG has no colored modes with less than 8-bit per channel*/ + } + + if(!alpha_done) + { + unsigned matchkey = (r == profile->key_r && g == profile->key_g && b == profile->key_b); + if(a != 255 && (a != 0 || (profile->key && !matchkey))) + { + profile->alpha = 1; + alpha_done = 1; + if(profile->bits < 8) profile->bits = 8; /*PNG has no alphachannel modes with less than 8-bit per channel*/ + } + else if(a == 0 && !profile->alpha && !profile->key) + { + profile->key = 1; + profile->key_r = r; + profile->key_g = g; + profile->key_b = b; + } + else if(a == 255 && profile->key && matchkey) + { + /* Color key cannot be used if an opaque pixel also has that RGB color. */ + profile->alpha = 1; + alpha_done = 1; + if(profile->bits < 8) profile->bits = 8; /*PNG has no alphachannel modes with less than 8-bit per channel*/ + } + } + + if(!numcolors_done) + { + if(!color_tree_has(&tree, r, g, b, a)) + { + color_tree_add(&tree, r, g, b, a, profile->numcolors); + if(profile->numcolors < 256) + { + unsigned char* p = profile->palette; + unsigned n = profile->numcolors; + p[n * 4 + 0] = r; + p[n * 4 + 1] = g; + p[n * 4 + 2] = b; + p[n * 4 + 3] = a; + } + ++profile->numcolors; + numcolors_done = profile->numcolors >= maxnumcolors; + } + } + + if(alpha_done && numcolors_done && colored_done && bits_done) break; + } + + /*make the profile's key always 16-bit for consistency - repeat each byte twice*/ + profile->key_r += (profile->key_r << 8); + profile->key_g += (profile->key_g << 8); + profile->key_b += (profile->key_b << 8); + } + + color_tree_cleanup(&tree); + return error; +} + +/*Automatically chooses color type that gives smallest amount of bits in the +output image, e.g. grey if there are only greyscale pixels, palette if there +are less than 256 colors, ... +Updates values of mode with a potentially smaller color model. mode_out should +contain the user chosen color model, but will be overwritten with the new chosen one.*/ +unsigned lodepng_auto_choose_color(LodePNGColorMode* mode_out, + const unsigned char* image, unsigned w, unsigned h, + const LodePNGColorMode* mode_in) +{ + LodePNGColorProfile prof; + unsigned error = 0; + unsigned i, n, palettebits, grey_ok, palette_ok; + + lodepng_color_profile_init(&prof); + error = lodepng_get_color_profile(&prof, image, w, h, mode_in); + if(error) return error; + mode_out->key_defined = 0; + + if(prof.key && w * h <= 16) + { + prof.alpha = 1; /*too few pixels to justify tRNS chunk overhead*/ + if(prof.bits < 8) prof.bits = 8; /*PNG has no alphachannel modes with less than 8-bit per channel*/ + } + grey_ok = !prof.colored && !prof.alpha; /*grey without alpha, with potentially low bits*/ + n = prof.numcolors; + palettebits = n <= 2 ? 1 : (n <= 4 ? 2 : (n <= 16 ? 4 : 8)); + palette_ok = n <= 256 && (n * 2 < w * h) && prof.bits <= 8; + if(w * h < n * 2) palette_ok = 0; /*don't add palette overhead if image has only a few pixels*/ + if(grey_ok && prof.bits <= palettebits) palette_ok = 0; /*grey is less overhead*/ + + if(palette_ok) + { + unsigned char* p = prof.palette; + lodepng_palette_clear(mode_out); /*remove potential earlier palette*/ + for(i = 0; i != prof.numcolors; ++i) + { + error = lodepng_palette_add(mode_out, p[i * 4 + 0], p[i * 4 + 1], p[i * 4 + 2], p[i * 4 + 3]); + if(error) break; + } + + mode_out->colortype = LCT_PALETTE; + mode_out->bitdepth = palettebits; + + if(mode_in->colortype == LCT_PALETTE && mode_in->palettesize >= mode_out->palettesize + && mode_in->bitdepth == mode_out->bitdepth) + { + /*If input should have same palette colors, keep original to preserve its order and prevent conversion*/ + lodepng_color_mode_cleanup(mode_out); + lodepng_color_mode_copy(mode_out, mode_in); + } + } + else /*8-bit or 16-bit per channel*/ + { + mode_out->bitdepth = prof.bits; + mode_out->colortype = prof.alpha ? (prof.colored ? LCT_RGBA : LCT_GREY_ALPHA) + : (prof.colored ? LCT_RGB : LCT_GREY); + + if(prof.key && !prof.alpha) + { + unsigned mask = (1u << mode_out->bitdepth) - 1u; /*profile always uses 16-bit, mask converts it*/ + mode_out->key_r = prof.key_r & mask; + mode_out->key_g = prof.key_g & mask; + mode_out->key_b = prof.key_b & mask; + mode_out->key_defined = 1; + } + } + + return error; +} + +#endif /* #ifdef LODEPNG_COMPILE_ENCODER */ + +/* +Paeth predicter, used by PNG filter type 4 +The parameters are of type short, but should come from unsigned chars, the shorts +are only needed to make the paeth calculation correct. +*/ +static unsigned char paethPredictor(short a, short b, short c) +{ + short pa = abs(b - c); + short pb = abs(a - c); + short pc = abs(a + b - c - c); + + if(pc < pa && pc < pb) return (unsigned char)c; + else if(pb < pa) return (unsigned char)b; + else return (unsigned char)a; +} + +/*shared values used by multiple Adam7 related functions*/ + +static const unsigned ADAM7_IX[7] = { 0, 4, 0, 2, 0, 1, 0 }; /*x start values*/ +static const unsigned ADAM7_IY[7] = { 0, 0, 4, 0, 2, 0, 1 }; /*y start values*/ +static const unsigned ADAM7_DX[7] = { 8, 8, 4, 4, 2, 2, 1 }; /*x delta values*/ +static const unsigned ADAM7_DY[7] = { 8, 8, 8, 4, 4, 2, 2 }; /*y delta values*/ + +/* +Outputs various dimensions and positions in the image related to the Adam7 reduced images. +passw: output containing the width of the 7 passes +passh: output containing the height of the 7 passes +filter_passstart: output containing the index of the start and end of each + reduced image with filter bytes +padded_passstart output containing the index of the start and end of each + reduced image when without filter bytes but with padded scanlines +passstart: output containing the index of the start and end of each reduced + image without padding between scanlines, but still padding between the images +w, h: width and height of non-interlaced image +bpp: bits per pixel +"padded" is only relevant if bpp is less than 8 and a scanline or image does not + end at a full byte +*/ +static void Adam7_getpassvalues(unsigned passw[7], unsigned passh[7], size_t filter_passstart[8], + size_t padded_passstart[8], size_t passstart[8], unsigned w, unsigned h, unsigned bpp) +{ + /*the passstart values have 8 values: the 8th one indicates the byte after the end of the 7th (= last) pass*/ + unsigned i; + + /*calculate width and height in pixels of each pass*/ + for(i = 0; i != 7; ++i) + { + passw[i] = (w + ADAM7_DX[i] - ADAM7_IX[i] - 1) / ADAM7_DX[i]; + passh[i] = (h + ADAM7_DY[i] - ADAM7_IY[i] - 1) / ADAM7_DY[i]; + if(passw[i] == 0) passh[i] = 0; + if(passh[i] == 0) passw[i] = 0; + } + + filter_passstart[0] = padded_passstart[0] = passstart[0] = 0; + for(i = 0; i != 7; ++i) + { + /*if passw[i] is 0, it's 0 bytes, not 1 (no filtertype-byte)*/ + filter_passstart[i + 1] = filter_passstart[i] + + ((passw[i] && passh[i]) ? passh[i] * (1 + (passw[i] * bpp + 7) / 8) : 0); + /*bits padded if needed to fill full byte at end of each scanline*/ + padded_passstart[i + 1] = padded_passstart[i] + passh[i] * ((passw[i] * bpp + 7) / 8); + /*only padded at end of reduced image*/ + passstart[i + 1] = passstart[i] + (passh[i] * passw[i] * bpp + 7) / 8; + } +} + +#ifdef LODEPNG_COMPILE_DECODER + +/* ////////////////////////////////////////////////////////////////////////// */ +/* / PNG Decoder / */ +/* ////////////////////////////////////////////////////////////////////////// */ + +/*read the information from the header and store it in the LodePNGInfo. return value is error*/ +unsigned lodepng_inspect(unsigned* w, unsigned* h, LodePNGState* state, + const unsigned char* in, size_t insize) +{ + LodePNGInfo* info = &state->info_png; + if(insize == 0 || in == 0) + { + CERROR_RETURN_ERROR(state->error, 48); /*error: the given data is empty*/ + } + if(insize < 33) + { + CERROR_RETURN_ERROR(state->error, 27); /*error: the data length is smaller than the length of a PNG header*/ + } + + /*when decoding a new PNG image, make sure all parameters created after previous decoding are reset*/ + lodepng_info_cleanup(info); + lodepng_info_init(info); + + if(in[0] != 137 || in[1] != 80 || in[2] != 78 || in[3] != 71 + || in[4] != 13 || in[5] != 10 || in[6] != 26 || in[7] != 10) + { + CERROR_RETURN_ERROR(state->error, 28); /*error: the first 8 bytes are not the correct PNG signature*/ + } + if(lodepng_chunk_length(in + 8) != 13) + { + CERROR_RETURN_ERROR(state->error, 94); /*error: header size must be 13 bytes*/ + } + if(!lodepng_chunk_type_equals(in + 8, "IHDR")) + { + CERROR_RETURN_ERROR(state->error, 29); /*error: it doesn't start with a IHDR chunk!*/ + } + + /*read the values given in the header*/ + *w = lodepng_read32bitInt(&in[16]); + *h = lodepng_read32bitInt(&in[20]); + info->color.bitdepth = in[24]; + info->color.colortype = (LodePNGColorType)in[25]; + info->compression_method = in[26]; + info->filter_method = in[27]; + info->interlace_method = in[28]; + + if(*w == 0 || *h == 0) + { + CERROR_RETURN_ERROR(state->error, 93); + } + + if(!state->decoder.ignore_crc) + { + unsigned CRC = lodepng_read32bitInt(&in[29]); + unsigned checksum = lodepng_crc32(&in[12], 17); + if(CRC != checksum) + { + CERROR_RETURN_ERROR(state->error, 57); /*invalid CRC*/ + } + } + + /*error: only compression method 0 is allowed in the specification*/ + if(info->compression_method != 0) CERROR_RETURN_ERROR(state->error, 32); + /*error: only filter method 0 is allowed in the specification*/ + if(info->filter_method != 0) CERROR_RETURN_ERROR(state->error, 33); + /*error: only interlace methods 0 and 1 exist in the specification*/ + if(info->interlace_method > 1) CERROR_RETURN_ERROR(state->error, 34); + + state->error = checkColorValidity(info->color.colortype, info->color.bitdepth); + return state->error; +} + +static unsigned unfilterScanline(unsigned char* recon, const unsigned char* scanline, const unsigned char* precon, + size_t bytewidth, unsigned char filterType, size_t length) +{ + /* + For PNG filter method 0 + unfilter a PNG image scanline by scanline. when the pixels are smaller than 1 byte, + the filter works byte per byte (bytewidth = 1) + precon is the previous unfiltered scanline, recon the result, scanline the current one + the incoming scanlines do NOT include the filtertype byte, that one is given in the parameter filterType instead + recon and scanline MAY be the same memory address! precon must be disjoint. + */ + + size_t i; + switch(filterType) + { + case 0: + for(i = 0; i != length; ++i) recon[i] = scanline[i]; + break; + case 1: + for(i = 0; i != bytewidth; ++i) recon[i] = scanline[i]; + for(i = bytewidth; i < length; ++i) recon[i] = scanline[i] + recon[i - bytewidth]; + break; + case 2: + if(precon) + { + for(i = 0; i != length; ++i) recon[i] = scanline[i] + precon[i]; + } + else + { + for(i = 0; i != length; ++i) recon[i] = scanline[i]; + } + break; + case 3: + if(precon) + { + for(i = 0; i != bytewidth; ++i) recon[i] = scanline[i] + (precon[i] >> 1); + for(i = bytewidth; i < length; ++i) recon[i] = scanline[i] + ((recon[i - bytewidth] + precon[i]) >> 1); + } + else + { + for(i = 0; i != bytewidth; ++i) recon[i] = scanline[i]; + for(i = bytewidth; i < length; ++i) recon[i] = scanline[i] + (recon[i - bytewidth] >> 1); + } + break; + case 4: + if(precon) + { + for(i = 0; i != bytewidth; ++i) + { + recon[i] = (scanline[i] + precon[i]); /*paethPredictor(0, precon[i], 0) is always precon[i]*/ + } + for(i = bytewidth; i < length; ++i) + { + recon[i] = (scanline[i] + paethPredictor(recon[i - bytewidth], precon[i], precon[i - bytewidth])); + } + } + else + { + for(i = 0; i != bytewidth; ++i) + { + recon[i] = scanline[i]; + } + for(i = bytewidth; i < length; ++i) + { + /*paethPredictor(recon[i - bytewidth], 0, 0) is always recon[i - bytewidth]*/ + recon[i] = (scanline[i] + recon[i - bytewidth]); + } + } + break; + default: return 36; /*error: unexisting filter type given*/ + } + return 0; +} + +static unsigned unfilter(unsigned char* out, const unsigned char* in, unsigned w, unsigned h, unsigned bpp) +{ + /* + For PNG filter method 0 + this function unfilters a single image (e.g. without interlacing this is called once, with Adam7 seven times) + out must have enough bytes allocated already, in must have the scanlines + 1 filtertype byte per scanline + w and h are image dimensions or dimensions of reduced image, bpp is bits per pixel + in and out are allowed to be the same memory address (but aren't the same size since in has the extra filter bytes) + */ + + unsigned y; + unsigned char* prevline = 0; + + /*bytewidth is used for filtering, is 1 when bpp < 8, number of bytes per pixel otherwise*/ + size_t bytewidth = (bpp + 7) / 8; + size_t linebytes = (w * bpp + 7) / 8; + + for(y = 0; y < h; ++y) + { + size_t outindex = linebytes * y; + size_t inindex = (1 + linebytes) * y; /*the extra filterbyte added to each row*/ + unsigned char filterType = in[inindex]; + + CERROR_TRY_RETURN(unfilterScanline(&out[outindex], &in[inindex + 1], prevline, bytewidth, filterType, linebytes)); + + prevline = &out[outindex]; + } + + return 0; +} + +/* +in: Adam7 interlaced image, with no padding bits between scanlines, but between + reduced images so that each reduced image starts at a byte. +out: the same pixels, but re-ordered so that they're now a non-interlaced image with size w*h +bpp: bits per pixel +out has the following size in bits: w * h * bpp. +in is possibly bigger due to padding bits between reduced images. +out must be big enough AND must be 0 everywhere if bpp < 8 in the current implementation +(because that's likely a little bit faster) +NOTE: comments about padding bits are only relevant if bpp < 8 +*/ +static void Adam7_deinterlace(unsigned char* out, const unsigned char* in, unsigned w, unsigned h, unsigned bpp) +{ + unsigned passw[7], passh[7]; + size_t filter_passstart[8], padded_passstart[8], passstart[8]; + unsigned i; + + Adam7_getpassvalues(passw, passh, filter_passstart, padded_passstart, passstart, w, h, bpp); + + if(bpp >= 8) + { + for(i = 0; i != 7; ++i) + { + unsigned x, y, b; + size_t bytewidth = bpp / 8; + for(y = 0; y < passh[i]; ++y) + for(x = 0; x < passw[i]; ++x) + { + size_t pixelinstart = passstart[i] + (y * passw[i] + x) * bytewidth; + size_t pixeloutstart = ((ADAM7_IY[i] + y * ADAM7_DY[i]) * w + ADAM7_IX[i] + x * ADAM7_DX[i]) * bytewidth; + for(b = 0; b < bytewidth; ++b) + { + out[pixeloutstart + b] = in[pixelinstart + b]; + } + } + } + } + else /*bpp < 8: Adam7 with pixels < 8 bit is a bit trickier: with bit pointers*/ + { + for(i = 0; i != 7; ++i) + { + unsigned x, y, b; + unsigned ilinebits = bpp * passw[i]; + unsigned olinebits = bpp * w; + size_t obp, ibp; /*bit pointers (for out and in buffer)*/ + for(y = 0; y < passh[i]; ++y) + for(x = 0; x < passw[i]; ++x) + { + ibp = (8 * passstart[i]) + (y * ilinebits + x * bpp); + obp = (ADAM7_IY[i] + y * ADAM7_DY[i]) * olinebits + (ADAM7_IX[i] + x * ADAM7_DX[i]) * bpp; + for(b = 0; b < bpp; ++b) + { + unsigned char bit = readBitFromReversedStream(&ibp, in); + /*note that this function assumes the out buffer is completely 0, use setBitOfReversedStream otherwise*/ + setBitOfReversedStream0(&obp, out, bit); + } + } + } + } +} + +static void removePaddingBits(unsigned char* out, const unsigned char* in, + size_t olinebits, size_t ilinebits, unsigned h) +{ + /* + After filtering there are still padding bits if scanlines have non multiple of 8 bit amounts. They need + to be removed (except at last scanline of (Adam7-reduced) image) before working with pure image buffers + for the Adam7 code, the color convert code and the output to the user. + in and out are allowed to be the same buffer, in may also be higher but still overlapping; in must + have >= ilinebits*h bits, out must have >= olinebits*h bits, olinebits must be <= ilinebits + also used to move bits after earlier such operations happened, e.g. in a sequence of reduced images from Adam7 + only useful if (ilinebits - olinebits) is a value in the range 1..7 + */ + unsigned y; + size_t diff = ilinebits - olinebits; + size_t ibp = 0, obp = 0; /*input and output bit pointers*/ + for(y = 0; y < h; ++y) + { + size_t x; + for(x = 0; x < olinebits; ++x) + { + unsigned char bit = readBitFromReversedStream(&ibp, in); + setBitOfReversedStream(&obp, out, bit); + } + ibp += diff; + } +} + +/*out must be buffer big enough to contain full image, and in must contain the full decompressed data from +the IDAT chunks (with filter index bytes and possible padding bits) +return value is error*/ +static unsigned postProcessScanlines(unsigned char* out, unsigned char* in, + unsigned w, unsigned h, const LodePNGInfo* info_png) +{ + /* + This function converts the filtered-padded-interlaced data into pure 2D image buffer with the PNG's colortype. + Steps: + *) if no Adam7: 1) unfilter 2) remove padding bits (= posible extra bits per scanline if bpp < 8) + *) if adam7: 1) 7x unfilter 2) 7x remove padding bits 3) Adam7_deinterlace + NOTE: the in buffer will be overwritten with intermediate data! + */ + unsigned bpp = lodepng_get_bpp(&info_png->color); + if(bpp == 0) return 31; /*error: invalid colortype*/ + + if(info_png->interlace_method == 0) + { + if(bpp < 8 && w * bpp != ((w * bpp + 7) / 8) * 8) + { + CERROR_TRY_RETURN(unfilter(in, in, w, h, bpp)); + removePaddingBits(out, in, w * bpp, ((w * bpp + 7) / 8) * 8, h); + } + /*we can immediately filter into the out buffer, no other steps needed*/ + else CERROR_TRY_RETURN(unfilter(out, in, w, h, bpp)); + } + else /*interlace_method is 1 (Adam7)*/ + { + unsigned passw[7], passh[7]; size_t filter_passstart[8], padded_passstart[8], passstart[8]; + unsigned i; + + Adam7_getpassvalues(passw, passh, filter_passstart, padded_passstart, passstart, w, h, bpp); + + for(i = 0; i != 7; ++i) + { + CERROR_TRY_RETURN(unfilter(&in[padded_passstart[i]], &in[filter_passstart[i]], passw[i], passh[i], bpp)); + /*TODO: possible efficiency improvement: if in this reduced image the bits fit nicely in 1 scanline, + move bytes instead of bits or move not at all*/ + if(bpp < 8) + { + /*remove padding bits in scanlines; after this there still may be padding + bits between the different reduced images: each reduced image still starts nicely at a byte*/ + removePaddingBits(&in[passstart[i]], &in[padded_passstart[i]], passw[i] * bpp, + ((passw[i] * bpp + 7) / 8) * 8, passh[i]); + } + } + + Adam7_deinterlace(out, in, w, h, bpp); + } + + return 0; +} + +static unsigned readChunk_PLTE(LodePNGColorMode* color, const unsigned char* data, size_t chunkLength) +{ + unsigned pos = 0, i; + if(color->palette) lodepng_free(color->palette); + color->palettesize = chunkLength / 3; + color->palette = (unsigned char*)lodepng_malloc(4 * color->palettesize); + if(!color->palette && color->palettesize) + { + color->palettesize = 0; + return 83; /*alloc fail*/ + } + if(color->palettesize > 256) return 38; /*error: palette too big*/ + + for(i = 0; i != color->palettesize; ++i) + { + color->palette[4 * i + 0] = data[pos++]; /*R*/ + color->palette[4 * i + 1] = data[pos++]; /*G*/ + color->palette[4 * i + 2] = data[pos++]; /*B*/ + color->palette[4 * i + 3] = 255; /*alpha*/ + } + + return 0; /* OK */ +} + +static unsigned readChunk_tRNS(LodePNGColorMode* color, const unsigned char* data, size_t chunkLength) +{ + unsigned i; + if(color->colortype == LCT_PALETTE) + { + /*error: more alpha values given than there are palette entries*/ + if(chunkLength > color->palettesize) return 38; + + for(i = 0; i != chunkLength; ++i) color->palette[4 * i + 3] = data[i]; + } + else if(color->colortype == LCT_GREY) + { + /*error: this chunk must be 2 bytes for greyscale image*/ + if(chunkLength != 2) return 30; + + color->key_defined = 1; + color->key_r = color->key_g = color->key_b = 256u * data[0] + data[1]; + } + else if(color->colortype == LCT_RGB) + { + /*error: this chunk must be 6 bytes for RGB image*/ + if(chunkLength != 6) return 41; + + color->key_defined = 1; + color->key_r = 256u * data[0] + data[1]; + color->key_g = 256u * data[2] + data[3]; + color->key_b = 256u * data[4] + data[5]; + } + else return 42; /*error: tRNS chunk not allowed for other color models*/ + + return 0; /* OK */ +} + + +#ifdef LODEPNG_COMPILE_ANCILLARY_CHUNKS +/*background color chunk (bKGD)*/ +static unsigned readChunk_bKGD(LodePNGInfo* info, const unsigned char* data, size_t chunkLength) +{ + if(info->color.colortype == LCT_PALETTE) + { + /*error: this chunk must be 1 byte for indexed color image*/ + if(chunkLength != 1) return 43; + + info->background_defined = 1; + info->background_r = info->background_g = info->background_b = data[0]; + } + else if(info->color.colortype == LCT_GREY || info->color.colortype == LCT_GREY_ALPHA) + { + /*error: this chunk must be 2 bytes for greyscale image*/ + if(chunkLength != 2) return 44; + + info->background_defined = 1; + info->background_r = info->background_g = info->background_b = 256u * data[0] + data[1]; + } + else if(info->color.colortype == LCT_RGB || info->color.colortype == LCT_RGBA) + { + /*error: this chunk must be 6 bytes for greyscale image*/ + if(chunkLength != 6) return 45; + + info->background_defined = 1; + info->background_r = 256u * data[0] + data[1]; + info->background_g = 256u * data[2] + data[3]; + info->background_b = 256u * data[4] + data[5]; + } + + return 0; /* OK */ +} + +/*text chunk (tEXt)*/ +static unsigned readChunk_tEXt(LodePNGInfo* info, const unsigned char* data, size_t chunkLength) +{ + unsigned error = 0; + char *key = 0, *str = 0; + unsigned i; + + while(!error) /*not really a while loop, only used to break on error*/ + { + unsigned length, string2_begin; + + length = 0; + while(length < chunkLength && data[length] != 0) ++length; + /*even though it's not allowed by the standard, no error is thrown if + there's no null termination char, if the text is empty*/ + if(length < 1 || length > 79) CERROR_BREAK(error, 89); /*keyword too short or long*/ + + key = (char*)lodepng_malloc(length + 1); + if(!key) CERROR_BREAK(error, 83); /*alloc fail*/ + + key[length] = 0; + for(i = 0; i != length; ++i) key[i] = (char)data[i]; + + string2_begin = length + 1; /*skip keyword null terminator*/ + + length = chunkLength < string2_begin ? 0 : (unsigned int)(chunkLength - string2_begin); + str = (char*)lodepng_malloc(length + 1); + if(!str) CERROR_BREAK(error, 83); /*alloc fail*/ + + str[length] = 0; + for(i = 0; i != length; ++i) str[i] = (char)data[string2_begin + i]; + + error = lodepng_add_text(info, key, str); + + break; + } + + lodepng_free(key); + lodepng_free(str); + + return error; +} + +/*compressed text chunk (zTXt)*/ +static unsigned readChunk_zTXt(LodePNGInfo* info, const LodePNGDecompressSettings* zlibsettings, + const unsigned char* data, size_t chunkLength) +{ + unsigned error = 0; + unsigned i; + + unsigned length, string2_begin; + char *key = 0; + ucvector decoded; + + ucvector_init(&decoded); + + while(!error) /*not really a while loop, only used to break on error*/ + { + for(length = 0; length < chunkLength && data[length] != 0; ++length) ; + if(length + 2 >= chunkLength) CERROR_BREAK(error, 75); /*no null termination, corrupt?*/ + if(length < 1 || length > 79) CERROR_BREAK(error, 89); /*keyword too short or long*/ + + key = (char*)lodepng_malloc(length + 1); + if(!key) CERROR_BREAK(error, 83); /*alloc fail*/ + + key[length] = 0; + for(i = 0; i != length; ++i) key[i] = (char)data[i]; + + if(data[length + 1] != 0) CERROR_BREAK(error, 72); /*the 0 byte indicating compression must be 0*/ + + string2_begin = length + 2; + if(string2_begin > chunkLength) CERROR_BREAK(error, 75); /*no null termination, corrupt?*/ + + length = (unsigned int)(chunkLength - string2_begin); + /*will fail if zlib error, e.g. if length is too small*/ + error = zlib_decompress(&decoded.data, &decoded.size, + (unsigned char*)(&data[string2_begin]), + length, zlibsettings); + if(error) break; + ucvector_push_back(&decoded, 0); + + error = lodepng_add_text(info, key, (char*)decoded.data); + + break; + } + + lodepng_free(key); + ucvector_cleanup(&decoded); + + return error; +} + +/*international text chunk (iTXt)*/ +static unsigned readChunk_iTXt(LodePNGInfo* info, const LodePNGDecompressSettings* zlibsettings, + const unsigned char* data, size_t chunkLength) +{ + unsigned error = 0; + unsigned i; + + unsigned length, begin, compressed; + char *key = 0, *langtag = 0, *transkey = 0; + ucvector decoded; + ucvector_init(&decoded); + + while(!error) /*not really a while loop, only used to break on error*/ + { + /*Quick check if the chunk length isn't too small. Even without check + it'd still fail with other error checks below if it's too short. This just gives a different error code.*/ + if(chunkLength < 5) CERROR_BREAK(error, 30); /*iTXt chunk too short*/ + + /*read the key*/ + for(length = 0; length < chunkLength && data[length] != 0; ++length) ; + if(length + 3 >= chunkLength) CERROR_BREAK(error, 75); /*no null termination char, corrupt?*/ + if(length < 1 || length > 79) CERROR_BREAK(error, 89); /*keyword too short or long*/ + + key = (char*)lodepng_malloc(length + 1); + if(!key) CERROR_BREAK(error, 83); /*alloc fail*/ + + key[length] = 0; + for(i = 0; i != length; ++i) key[i] = (char)data[i]; + + /*read the compression method*/ + compressed = data[length + 1]; + if(data[length + 2] != 0) CERROR_BREAK(error, 72); /*the 0 byte indicating compression must be 0*/ + + /*even though it's not allowed by the standard, no error is thrown if + there's no null termination char, if the text is empty for the next 3 texts*/ + + /*read the langtag*/ + begin = length + 3; + length = 0; + for(i = begin; i < chunkLength && data[i] != 0; ++i) ++length; + + langtag = (char*)lodepng_malloc(length + 1); + if(!langtag) CERROR_BREAK(error, 83); /*alloc fail*/ + + langtag[length] = 0; + for(i = 0; i != length; ++i) langtag[i] = (char)data[begin + i]; + + /*read the transkey*/ + begin += length + 1; + length = 0; + for(i = begin; i < chunkLength && data[i] != 0; ++i) ++length; + + transkey = (char*)lodepng_malloc(length + 1); + if(!transkey) CERROR_BREAK(error, 83); /*alloc fail*/ + + transkey[length] = 0; + for(i = 0; i != length; ++i) transkey[i] = (char)data[begin + i]; + + /*read the actual text*/ + begin += length + 1; + + length = chunkLength < begin ? 0 : (unsigned int)(chunkLength - begin); + + if(compressed) + { + /*will fail if zlib error, e.g. if length is too small*/ + error = zlib_decompress(&decoded.data, &decoded.size, + (unsigned char*)(&data[begin]), + length, zlibsettings); + if(error) break; + if(decoded.allocsize < decoded.size) decoded.allocsize = decoded.size; + ucvector_push_back(&decoded, 0); + } + else + { + if(!ucvector_resize(&decoded, length + 1)) CERROR_BREAK(error, 83 /*alloc fail*/); + + decoded.data[length] = 0; + for(i = 0; i != length; ++i) decoded.data[i] = data[begin + i]; + } + + error = lodepng_add_itext(info, key, langtag, transkey, (char*)decoded.data); + + break; + } + + lodepng_free(key); + lodepng_free(langtag); + lodepng_free(transkey); + ucvector_cleanup(&decoded); + + return error; +} + +static unsigned readChunk_tIME(LodePNGInfo* info, const unsigned char* data, size_t chunkLength) +{ + if(chunkLength != 7) return 73; /*invalid tIME chunk size*/ + + info->time_defined = 1; + info->time.year = 256u * data[0] + data[1]; + info->time.month = data[2]; + info->time.day = data[3]; + info->time.hour = data[4]; + info->time.minute = data[5]; + info->time.second = data[6]; + + return 0; /* OK */ +} + +static unsigned readChunk_pHYs(LodePNGInfo* info, const unsigned char* data, size_t chunkLength) +{ + if(chunkLength != 9) return 74; /*invalid pHYs chunk size*/ + + info->phys_defined = 1; + info->phys_x = 16777216u * data[0] + 65536u * data[1] + 256u * data[2] + data[3]; + info->phys_y = 16777216u * data[4] + 65536u * data[5] + 256u * data[6] + data[7]; + info->phys_unit = data[8]; + + return 0; /* OK */ +} +#endif /*LODEPNG_COMPILE_ANCILLARY_CHUNKS*/ + +/*read a PNG, the result will be in the same color type as the PNG (hence "generic")*/ +static void decodeGeneric(unsigned char** out, unsigned* w, unsigned* h, + LodePNGState* state, + const unsigned char* in, size_t insize) +{ + unsigned char IEND = 0; + const unsigned char* chunk; + size_t i; + ucvector idat; /*the data from idat chunks*/ + ucvector scanlines; + size_t predict; + size_t numpixels; + + /*for unknown chunk order*/ + unsigned unknown = 0; +#ifdef LODEPNG_COMPILE_ANCILLARY_CHUNKS + unsigned critical_pos = 1; /*1 = after IHDR, 2 = after PLTE, 3 = after IDAT*/ +#endif /*LODEPNG_COMPILE_ANCILLARY_CHUNKS*/ + + /*provide some proper output values if error will happen*/ + *out = 0; + + state->error = lodepng_inspect(w, h, state, in, insize); /*reads header and resets other parameters in state->info_png*/ + if(state->error) return; + + numpixels = *w * *h; + + /*multiplication overflow*/ + if(*h != 0 && numpixels / *h != *w) CERROR_RETURN(state->error, 92); + /*multiplication overflow possible further below. Allows up to 2^31-1 pixel + bytes with 16-bit RGBA, the rest is room for filter bytes.*/ + if(numpixels > 268435455) CERROR_RETURN(state->error, 92); + + ucvector_init(&idat); + chunk = &in[33]; /*first byte of the first chunk after the header*/ + + /*loop through the chunks, ignoring unknown chunks and stopping at IEND chunk. + IDAT data is put at the start of the in buffer*/ + while(!IEND && !state->error) + { + unsigned chunkLength; + const unsigned char* data; /*the data in the chunk*/ + + /*error: size of the in buffer too small to contain next chunk*/ + if((size_t)((chunk - in) + 12) > insize || chunk < in) CERROR_BREAK(state->error, 30); + + /*length of the data of the chunk, excluding the length bytes, chunk type and CRC bytes*/ + chunkLength = lodepng_chunk_length(chunk); + /*error: chunk length larger than the max PNG chunk size*/ + if(chunkLength > 2147483647) CERROR_BREAK(state->error, 63); + + if((size_t)((chunk - in) + chunkLength + 12) > insize || (chunk + chunkLength + 12) < in) + { + CERROR_BREAK(state->error, 64); /*error: size of the in buffer too small to contain next chunk*/ + } + + data = lodepng_chunk_data_const(chunk); + + /*IDAT chunk, containing compressed image data*/ + if(lodepng_chunk_type_equals(chunk, "IDAT")) + { + size_t oldsize = idat.size; + if(!ucvector_resize(&idat, oldsize + chunkLength)) CERROR_BREAK(state->error, 83 /*alloc fail*/); + for(i = 0; i != chunkLength; ++i) idat.data[oldsize + i] = data[i]; +#ifdef LODEPNG_COMPILE_ANCILLARY_CHUNKS + critical_pos = 3; +#endif /*LODEPNG_COMPILE_ANCILLARY_CHUNKS*/ + } + /*IEND chunk*/ + else if(lodepng_chunk_type_equals(chunk, "IEND")) + { + IEND = 1; + } + /*palette chunk (PLTE)*/ + else if(lodepng_chunk_type_equals(chunk, "PLTE")) + { + state->error = readChunk_PLTE(&state->info_png.color, data, chunkLength); + if(state->error) break; +#ifdef LODEPNG_COMPILE_ANCILLARY_CHUNKS + critical_pos = 2; +#endif /*LODEPNG_COMPILE_ANCILLARY_CHUNKS*/ + } + /*palette transparency chunk (tRNS)*/ + else if(lodepng_chunk_type_equals(chunk, "tRNS")) + { + state->error = readChunk_tRNS(&state->info_png.color, data, chunkLength); + if(state->error) break; + } +#ifdef LODEPNG_COMPILE_ANCILLARY_CHUNKS + /*background color chunk (bKGD)*/ + else if(lodepng_chunk_type_equals(chunk, "bKGD")) + { + state->error = readChunk_bKGD(&state->info_png, data, chunkLength); + if(state->error) break; + } + /*text chunk (tEXt)*/ + else if(lodepng_chunk_type_equals(chunk, "tEXt")) + { + if(state->decoder.read_text_chunks) + { + state->error = readChunk_tEXt(&state->info_png, data, chunkLength); + if(state->error) break; + } + } + /*compressed text chunk (zTXt)*/ + else if(lodepng_chunk_type_equals(chunk, "zTXt")) + { + if(state->decoder.read_text_chunks) + { + state->error = readChunk_zTXt(&state->info_png, &state->decoder.zlibsettings, data, chunkLength); + if(state->error) break; + } + } + /*international text chunk (iTXt)*/ + else if(lodepng_chunk_type_equals(chunk, "iTXt")) + { + if(state->decoder.read_text_chunks) + { + state->error = readChunk_iTXt(&state->info_png, &state->decoder.zlibsettings, data, chunkLength); + if(state->error) break; + } + } + else if(lodepng_chunk_type_equals(chunk, "tIME")) + { + state->error = readChunk_tIME(&state->info_png, data, chunkLength); + if(state->error) break; + } + else if(lodepng_chunk_type_equals(chunk, "pHYs")) + { + state->error = readChunk_pHYs(&state->info_png, data, chunkLength); + if(state->error) break; + } +#endif /*LODEPNG_COMPILE_ANCILLARY_CHUNKS*/ + else /*it's not an implemented chunk type, so ignore it: skip over the data*/ + { + /*error: unknown critical chunk (5th bit of first byte of chunk type is 0)*/ + if(!lodepng_chunk_ancillary(chunk)) CERROR_BREAK(state->error, 69); + + unknown = 1; +#ifdef LODEPNG_COMPILE_ANCILLARY_CHUNKS + if(state->decoder.remember_unknown_chunks) + { + state->error = lodepng_chunk_append(&state->info_png.unknown_chunks_data[critical_pos - 1], + &state->info_png.unknown_chunks_size[critical_pos - 1], chunk); + if(state->error) break; + } +#endif /*LODEPNG_COMPILE_ANCILLARY_CHUNKS*/ + } + + if(!state->decoder.ignore_crc && !unknown) /*check CRC if wanted, only on known chunk types*/ + { + if(lodepng_chunk_check_crc(chunk)) CERROR_BREAK(state->error, 57); /*invalid CRC*/ + } + + if(!IEND) chunk = lodepng_chunk_next_const(chunk); + } + + ucvector_init(&scanlines); + /*predict output size, to allocate exact size for output buffer to avoid more dynamic allocation. + If the decompressed size does not match the prediction, the image must be corrupt.*/ + if(state->info_png.interlace_method == 0) + { + /*The extra *h is added because this are the filter bytes every scanline starts with*/ + predict = lodepng_get_raw_size_idat(*w, *h, &state->info_png.color) + *h; + } + else + { + /*Adam-7 interlaced: predicted size is the sum of the 7 sub-images sizes*/ + const LodePNGColorMode* color = &state->info_png.color; + predict = 0; + predict += lodepng_get_raw_size_idat((*w + 7) >> 3, (*h + 7) >> 3, color) + ((*h + 7) >> 3); + if(*w > 4) predict += lodepng_get_raw_size_idat((*w + 3) >> 3, (*h + 7) >> 3, color) + ((*h + 7) >> 3); + predict += lodepng_get_raw_size_idat((*w + 3) >> 2, (*h + 3) >> 3, color) + ((*h + 3) >> 3); + if(*w > 2) predict += lodepng_get_raw_size_idat((*w + 1) >> 2, (*h + 3) >> 2, color) + ((*h + 3) >> 2); + predict += lodepng_get_raw_size_idat((*w + 1) >> 1, (*h + 1) >> 2, color) + ((*h + 1) >> 2); + if(*w > 1) predict += lodepng_get_raw_size_idat((*w + 0) >> 1, (*h + 1) >> 1, color) + ((*h + 1) >> 1); + predict += lodepng_get_raw_size_idat((*w + 0), (*h + 0) >> 1, color) + ((*h + 0) >> 1); + } + if(!state->error && !ucvector_reserve(&scanlines, predict)) state->error = 83; /*alloc fail*/ + if(!state->error) + { + state->error = zlib_decompress(&scanlines.data, &scanlines.size, idat.data, + idat.size, &state->decoder.zlibsettings); + if(!state->error && scanlines.size != predict) state->error = 91; /*decompressed size doesn't match prediction*/ + } + ucvector_cleanup(&idat); + + if(!state->error) + { + size_t outsize = lodepng_get_raw_size(*w, *h, &state->info_png.color); + *out = (unsigned char*)lodepng_malloc(outsize); + if(!*out) state->error = 83; /*alloc fail*/ + for(i = 0; i < outsize; i++) (*out)[i] = 0; + if(!state->error) state->error = postProcessScanlines(*out, scanlines.data, *w, *h, &state->info_png); + } + ucvector_cleanup(&scanlines); +} + +unsigned lodepng_decode(unsigned char** out, unsigned* w, unsigned* h, + LodePNGState* state, + const unsigned char* in, size_t insize) +{ + *out = 0; + decodeGeneric(out, w, h, state, in, insize); + if(state->error) return state->error; + if(!state->decoder.color_convert || lodepng_color_mode_equal(&state->info_raw, &state->info_png.color)) + { + /*same color type, no copying or converting of data needed*/ + /*store the info_png color settings on the info_raw so that the info_raw still reflects what colortype + the raw image has to the end user*/ + if(!state->decoder.color_convert) + { + state->error = lodepng_color_mode_copy(&state->info_raw, &state->info_png.color); + if(state->error) return state->error; + } + } + else + { + /*color conversion needed; sort of copy of the data*/ + unsigned char* data = *out; + size_t outsize; + + /*TODO: check if this works according to the statement in the documentation: "The converter can convert + from greyscale input color type, to 8-bit greyscale or greyscale with alpha"*/ + if(!(state->info_raw.colortype == LCT_RGB || state->info_raw.colortype == LCT_RGBA) + && !(state->info_raw.bitdepth == 8)) + { + return 56; /*unsupported color mode conversion*/ + } + + outsize = lodepng_get_raw_size(*w, *h, &state->info_raw); + *out = (unsigned char*)lodepng_malloc(outsize); + if(!(*out)) + { + state->error = 83; /*alloc fail*/ + } + else state->error = lodepng_convert(*out, data, &state->info_raw, + &state->info_png.color, *w, *h); + lodepng_free(data); + } + return state->error; +} + +unsigned lodepng_decode_memory(unsigned char** out, unsigned* w, unsigned* h, const unsigned char* in, + size_t insize, LodePNGColorType colortype, unsigned bitdepth) +{ + unsigned error; + LodePNGState state; + lodepng_state_init(&state); + state.info_raw.colortype = colortype; + state.info_raw.bitdepth = bitdepth; + error = lodepng_decode(out, w, h, &state, in, insize); + lodepng_state_cleanup(&state); + return error; +} + +unsigned lodepng_decode32(unsigned char** out, unsigned* w, unsigned* h, const unsigned char* in, size_t insize) +{ + return lodepng_decode_memory(out, w, h, in, insize, LCT_RGBA, 8); +} + +unsigned lodepng_decode24(unsigned char** out, unsigned* w, unsigned* h, const unsigned char* in, size_t insize) +{ + return lodepng_decode_memory(out, w, h, in, insize, LCT_RGB, 8); +} + +#ifdef LODEPNG_COMPILE_DISK +unsigned lodepng_decode_file(unsigned char** out, unsigned* w, unsigned* h, const char* filename, + LodePNGColorType colortype, unsigned bitdepth) +{ + unsigned char* buffer; + size_t buffersize; + unsigned error; + error = lodepng_load_file(&buffer, &buffersize, filename); + if(!error) error = lodepng_decode_memory(out, w, h, buffer, buffersize, colortype, bitdepth); + lodepng_free(buffer); + return error; +} + +unsigned lodepng_decode32_file(unsigned char** out, unsigned* w, unsigned* h, const char* filename) +{ + return lodepng_decode_file(out, w, h, filename, LCT_RGBA, 8); +} + +unsigned lodepng_decode24_file(unsigned char** out, unsigned* w, unsigned* h, const char* filename) +{ + return lodepng_decode_file(out, w, h, filename, LCT_RGB, 8); +} +#endif /*LODEPNG_COMPILE_DISK*/ + +void lodepng_decoder_settings_init(LodePNGDecoderSettings* settings) +{ + settings->color_convert = 1; +#ifdef LODEPNG_COMPILE_ANCILLARY_CHUNKS + settings->read_text_chunks = 1; + settings->remember_unknown_chunks = 0; +#endif /*LODEPNG_COMPILE_ANCILLARY_CHUNKS*/ + settings->ignore_crc = 0; + lodepng_decompress_settings_init(&settings->zlibsettings); +} + +#endif /*LODEPNG_COMPILE_DECODER*/ + +#if defined(LODEPNG_COMPILE_DECODER) || defined(LODEPNG_COMPILE_ENCODER) + +void lodepng_state_init(LodePNGState* state) +{ +#ifdef LODEPNG_COMPILE_DECODER + lodepng_decoder_settings_init(&state->decoder); +#endif /*LODEPNG_COMPILE_DECODER*/ +#ifdef LODEPNG_COMPILE_ENCODER + lodepng_encoder_settings_init(&state->encoder); +#endif /*LODEPNG_COMPILE_ENCODER*/ + lodepng_color_mode_init(&state->info_raw); + lodepng_info_init(&state->info_png); + state->error = 1; +} + +void lodepng_state_cleanup(LodePNGState* state) +{ + lodepng_color_mode_cleanup(&state->info_raw); + lodepng_info_cleanup(&state->info_png); +} + +void lodepng_state_copy(LodePNGState* dest, const LodePNGState* source) +{ + lodepng_state_cleanup(dest); + *dest = *source; + lodepng_color_mode_init(&dest->info_raw); + lodepng_info_init(&dest->info_png); + dest->error = lodepng_color_mode_copy(&dest->info_raw, &source->info_raw); if(dest->error) return; + dest->error = lodepng_info_copy(&dest->info_png, &source->info_png); if(dest->error) return; +} + +#endif /* defined(LODEPNG_COMPILE_DECODER) || defined(LODEPNG_COMPILE_ENCODER) */ + +#ifdef LODEPNG_COMPILE_ENCODER + +/* ////////////////////////////////////////////////////////////////////////// */ +/* / PNG Encoder / */ +/* ////////////////////////////////////////////////////////////////////////// */ + +/*chunkName must be string of 4 characters*/ +static unsigned addChunk(ucvector* out, const char* chunkName, const unsigned char* data, size_t length) +{ + CERROR_TRY_RETURN(lodepng_chunk_create(&out->data, &out->size, (unsigned)length, chunkName, data)); + out->allocsize = out->size; /*fix the allocsize again*/ + return 0; +} + +static void writeSignature(ucvector* out) +{ + /*8 bytes PNG signature, aka the magic bytes*/ + ucvector_push_back(out, 137); + ucvector_push_back(out, 80); + ucvector_push_back(out, 78); + ucvector_push_back(out, 71); + ucvector_push_back(out, 13); + ucvector_push_back(out, 10); + ucvector_push_back(out, 26); + ucvector_push_back(out, 10); +} + +static unsigned addChunk_IHDR(ucvector* out, unsigned w, unsigned h, + LodePNGColorType colortype, unsigned bitdepth, unsigned interlace_method) +{ + unsigned error = 0; + ucvector header; + ucvector_init(&header); + + lodepng_add32bitInt(&header, w); /*width*/ + lodepng_add32bitInt(&header, h); /*height*/ + ucvector_push_back(&header, (unsigned char)bitdepth); /*bit depth*/ + ucvector_push_back(&header, (unsigned char)colortype); /*color type*/ + ucvector_push_back(&header, 0); /*compression method*/ + ucvector_push_back(&header, 0); /*filter method*/ + ucvector_push_back(&header, interlace_method); /*interlace method*/ + + error = addChunk(out, "IHDR", header.data, header.size); + ucvector_cleanup(&header); + + return error; +} + +static unsigned addChunk_PLTE(ucvector* out, const LodePNGColorMode* info) +{ + unsigned error = 0; + size_t i; + ucvector PLTE; + ucvector_init(&PLTE); + for(i = 0; i != info->palettesize * 4; ++i) + { + /*add all channels except alpha channel*/ + if(i % 4 != 3) ucvector_push_back(&PLTE, info->palette[i]); + } + error = addChunk(out, "PLTE", PLTE.data, PLTE.size); + ucvector_cleanup(&PLTE); + + return error; +} + +static unsigned addChunk_tRNS(ucvector* out, const LodePNGColorMode* info) +{ + unsigned error = 0; + size_t i; + ucvector tRNS; + ucvector_init(&tRNS); + if(info->colortype == LCT_PALETTE) + { + size_t amount = info->palettesize; + /*the tail of palette values that all have 255 as alpha, does not have to be encoded*/ + for(i = info->palettesize; i != 0; --i) + { + if(info->palette[4 * (i - 1) + 3] == 255) --amount; + else break; + } + /*add only alpha channel*/ + for(i = 0; i != amount; ++i) ucvector_push_back(&tRNS, info->palette[4 * i + 3]); + } + else if(info->colortype == LCT_GREY) + { + if(info->key_defined) + { + ucvector_push_back(&tRNS, (unsigned char)(info->key_r >> 8)); + ucvector_push_back(&tRNS, (unsigned char)(info->key_r & 255)); + } + } + else if(info->colortype == LCT_RGB) + { + if(info->key_defined) + { + ucvector_push_back(&tRNS, (unsigned char)(info->key_r >> 8)); + ucvector_push_back(&tRNS, (unsigned char)(info->key_r & 255)); + ucvector_push_back(&tRNS, (unsigned char)(info->key_g >> 8)); + ucvector_push_back(&tRNS, (unsigned char)(info->key_g & 255)); + ucvector_push_back(&tRNS, (unsigned char)(info->key_b >> 8)); + ucvector_push_back(&tRNS, (unsigned char)(info->key_b & 255)); + } + } + + error = addChunk(out, "tRNS", tRNS.data, tRNS.size); + ucvector_cleanup(&tRNS); + + return error; +} + +static unsigned addChunk_IDAT(ucvector* out, const unsigned char* data, size_t datasize, + LodePNGCompressSettings* zlibsettings) +{ + ucvector zlibdata; + unsigned error = 0; + + /*compress with the Zlib compressor*/ + ucvector_init(&zlibdata); + error = zlib_compress(&zlibdata.data, &zlibdata.size, data, datasize, zlibsettings); + if(!error) error = addChunk(out, "IDAT", zlibdata.data, zlibdata.size); + ucvector_cleanup(&zlibdata); + + return error; +} + +static unsigned addChunk_IEND(ucvector* out) +{ + unsigned error = 0; + error = addChunk(out, "IEND", 0, 0); + return error; +} + +#ifdef LODEPNG_COMPILE_ANCILLARY_CHUNKS + +static unsigned addChunk_tEXt(ucvector* out, const char* keyword, const char* textstring) +{ + unsigned error = 0; + size_t i; + ucvector text; + ucvector_init(&text); + for(i = 0; keyword[i] != 0; ++i) ucvector_push_back(&text, (unsigned char)keyword[i]); + if(i < 1 || i > 79) return 89; /*error: invalid keyword size*/ + ucvector_push_back(&text, 0); /*0 termination char*/ + for(i = 0; textstring[i] != 0; ++i) ucvector_push_back(&text, (unsigned char)textstring[i]); + error = addChunk(out, "tEXt", text.data, text.size); + ucvector_cleanup(&text); + + return error; +} + +static unsigned addChunk_zTXt(ucvector* out, const char* keyword, const char* textstring, + LodePNGCompressSettings* zlibsettings) +{ + unsigned error = 0; + ucvector data, compressed; + size_t i, textsize = strlen(textstring); + + ucvector_init(&data); + ucvector_init(&compressed); + for(i = 0; keyword[i] != 0; ++i) ucvector_push_back(&data, (unsigned char)keyword[i]); + if(i < 1 || i > 79) return 89; /*error: invalid keyword size*/ + ucvector_push_back(&data, 0); /*0 termination char*/ + ucvector_push_back(&data, 0); /*compression method: 0*/ + + error = zlib_compress(&compressed.data, &compressed.size, + (unsigned char*)textstring, textsize, zlibsettings); + if(!error) + { + for(i = 0; i != compressed.size; ++i) ucvector_push_back(&data, compressed.data[i]); + error = addChunk(out, "zTXt", data.data, data.size); + } + + ucvector_cleanup(&compressed); + ucvector_cleanup(&data); + return error; +} + +static unsigned addChunk_iTXt(ucvector* out, unsigned compressed, const char* keyword, const char* langtag, + const char* transkey, const char* textstring, LodePNGCompressSettings* zlibsettings) +{ + unsigned error = 0; + ucvector data; + size_t i, textsize = strlen(textstring); + + ucvector_init(&data); + + for(i = 0; keyword[i] != 0; ++i) ucvector_push_back(&data, (unsigned char)keyword[i]); + if(i < 1 || i > 79) return 89; /*error: invalid keyword size*/ + ucvector_push_back(&data, 0); /*null termination char*/ + ucvector_push_back(&data, compressed ? 1 : 0); /*compression flag*/ + ucvector_push_back(&data, 0); /*compression method*/ + for(i = 0; langtag[i] != 0; ++i) ucvector_push_back(&data, (unsigned char)langtag[i]); + ucvector_push_back(&data, 0); /*null termination char*/ + for(i = 0; transkey[i] != 0; ++i) ucvector_push_back(&data, (unsigned char)transkey[i]); + ucvector_push_back(&data, 0); /*null termination char*/ + + if(compressed) + { + ucvector compressed_data; + ucvector_init(&compressed_data); + error = zlib_compress(&compressed_data.data, &compressed_data.size, + (unsigned char*)textstring, textsize, zlibsettings); + if(!error) + { + for(i = 0; i != compressed_data.size; ++i) ucvector_push_back(&data, compressed_data.data[i]); + } + ucvector_cleanup(&compressed_data); + } + else /*not compressed*/ + { + for(i = 0; textstring[i] != 0; ++i) ucvector_push_back(&data, (unsigned char)textstring[i]); + } + + if(!error) error = addChunk(out, "iTXt", data.data, data.size); + ucvector_cleanup(&data); + return error; +} + +static unsigned addChunk_bKGD(ucvector* out, const LodePNGInfo* info) +{ + unsigned error = 0; + ucvector bKGD; + ucvector_init(&bKGD); + if(info->color.colortype == LCT_GREY || info->color.colortype == LCT_GREY_ALPHA) + { + ucvector_push_back(&bKGD, (unsigned char)(info->background_r >> 8)); + ucvector_push_back(&bKGD, (unsigned char)(info->background_r & 255)); + } + else if(info->color.colortype == LCT_RGB || info->color.colortype == LCT_RGBA) + { + ucvector_push_back(&bKGD, (unsigned char)(info->background_r >> 8)); + ucvector_push_back(&bKGD, (unsigned char)(info->background_r & 255)); + ucvector_push_back(&bKGD, (unsigned char)(info->background_g >> 8)); + ucvector_push_back(&bKGD, (unsigned char)(info->background_g & 255)); + ucvector_push_back(&bKGD, (unsigned char)(info->background_b >> 8)); + ucvector_push_back(&bKGD, (unsigned char)(info->background_b & 255)); + } + else if(info->color.colortype == LCT_PALETTE) + { + ucvector_push_back(&bKGD, (unsigned char)(info->background_r & 255)); /*palette index*/ + } + + error = addChunk(out, "bKGD", bKGD.data, bKGD.size); + ucvector_cleanup(&bKGD); + + return error; +} + +static unsigned addChunk_tIME(ucvector* out, const LodePNGTime* time) +{ + unsigned error = 0; + unsigned char* data = (unsigned char*)lodepng_malloc(7); + if(!data) return 83; /*alloc fail*/ + data[0] = (unsigned char)(time->year >> 8); + data[1] = (unsigned char)(time->year & 255); + data[2] = (unsigned char)time->month; + data[3] = (unsigned char)time->day; + data[4] = (unsigned char)time->hour; + data[5] = (unsigned char)time->minute; + data[6] = (unsigned char)time->second; + error = addChunk(out, "tIME", data, 7); + lodepng_free(data); + return error; +} + +static unsigned addChunk_pHYs(ucvector* out, const LodePNGInfo* info) +{ + unsigned error = 0; + ucvector data; + ucvector_init(&data); + + lodepng_add32bitInt(&data, info->phys_x); + lodepng_add32bitInt(&data, info->phys_y); + ucvector_push_back(&data, info->phys_unit); + + error = addChunk(out, "pHYs", data.data, data.size); + ucvector_cleanup(&data); + + return error; +} + +#endif /*LODEPNG_COMPILE_ANCILLARY_CHUNKS*/ + +static void filterScanline(unsigned char* out, const unsigned char* scanline, const unsigned char* prevline, + size_t length, size_t bytewidth, unsigned char filterType) +{ + size_t i; + switch(filterType) + { + case 0: /*None*/ + for(i = 0; i != length; ++i) out[i] = scanline[i]; + break; + case 1: /*Sub*/ + for(i = 0; i != bytewidth; ++i) out[i] = scanline[i]; + for(i = bytewidth; i < length; ++i) out[i] = scanline[i] - scanline[i - bytewidth]; + break; + case 2: /*Up*/ + if(prevline) + { + for(i = 0; i != length; ++i) out[i] = scanline[i] - prevline[i]; + } + else + { + for(i = 0; i != length; ++i) out[i] = scanline[i]; + } + break; + case 3: /*Average*/ + if(prevline) + { + for(i = 0; i != bytewidth; ++i) out[i] = scanline[i] - (prevline[i] >> 1); + for(i = bytewidth; i < length; ++i) out[i] = scanline[i] - ((scanline[i - bytewidth] + prevline[i]) >> 1); + } + else + { + for(i = 0; i != bytewidth; ++i) out[i] = scanline[i]; + for(i = bytewidth; i < length; ++i) out[i] = scanline[i] - (scanline[i - bytewidth] >> 1); + } + break; + case 4: /*Paeth*/ + if(prevline) + { + /*paethPredictor(0, prevline[i], 0) is always prevline[i]*/ + for(i = 0; i != bytewidth; ++i) out[i] = (scanline[i] - prevline[i]); + for(i = bytewidth; i < length; ++i) + { + out[i] = (scanline[i] - paethPredictor(scanline[i - bytewidth], prevline[i], prevline[i - bytewidth])); + } + } + else + { + for(i = 0; i != bytewidth; ++i) out[i] = scanline[i]; + /*paethPredictor(scanline[i - bytewidth], 0, 0) is always scanline[i - bytewidth]*/ + for(i = bytewidth; i < length; ++i) out[i] = (scanline[i] - scanline[i - bytewidth]); + } + break; + default: return; /*unexisting filter type given*/ + } +} + +/* log2 approximation. A slight bit faster than std::log. */ +static float flog2(float f) +{ + float result = 0; + while(f > 32) { result += 4; f /= 16; } + while(f > 2) { ++result; f /= 2; } + return result + 1.442695f * (f * f * f / 3 - 3 * f * f / 2 + 3 * f - 1.83333f); +} + +static unsigned filter(unsigned char* out, const unsigned char* in, unsigned w, unsigned h, + const LodePNGColorMode* info, const LodePNGEncoderSettings* settings) +{ + /* + For PNG filter method 0 + out must be a buffer with as size: h + (w * h * bpp + 7) / 8, because there are + the scanlines with 1 extra byte per scanline + */ + + unsigned bpp = lodepng_get_bpp(info); + /*the width of a scanline in bytes, not including the filter type*/ + size_t linebytes = (w * bpp + 7) / 8; + /*bytewidth is used for filtering, is 1 when bpp < 8, number of bytes per pixel otherwise*/ + size_t bytewidth = (bpp + 7) / 8; + const unsigned char* prevline = 0; + unsigned x, y; + unsigned error = 0; + LodePNGFilterStrategy strategy = settings->filter_strategy; + + /* + There is a heuristic called the minimum sum of absolute differences heuristic, suggested by the PNG standard: + * If the image type is Palette, or the bit depth is smaller than 8, then do not filter the image (i.e. + use fixed filtering, with the filter None). + * (The other case) If the image type is Grayscale or RGB (with or without Alpha), and the bit depth is + not smaller than 8, then use adaptive filtering heuristic as follows: independently for each row, apply + all five filters and select the filter that produces the smallest sum of absolute values per row. + This heuristic is used if filter strategy is LFS_MINSUM and filter_palette_zero is true. + + If filter_palette_zero is true and filter_strategy is not LFS_MINSUM, the above heuristic is followed, + but for "the other case", whatever strategy filter_strategy is set to instead of the minimum sum + heuristic is used. + */ + if(settings->filter_palette_zero && + (info->colortype == LCT_PALETTE || info->bitdepth < 8)) strategy = LFS_ZERO; + + if(bpp == 0) return 31; /*error: invalid color type*/ + + if(strategy == LFS_ZERO) + { + for(y = 0; y != h; ++y) + { + size_t outindex = (1 + linebytes) * y; /*the extra filterbyte added to each row*/ + size_t inindex = linebytes * y; + out[outindex] = 0; /*filter type byte*/ + filterScanline(&out[outindex + 1], &in[inindex], prevline, linebytes, bytewidth, 0); + prevline = &in[inindex]; + } + } + else if(strategy == LFS_MINSUM) + { + /*adaptive filtering*/ + size_t sum[5]; + unsigned char* attempt[5]; /*five filtering attempts, one for each filter type*/ + size_t smallest = 0; + unsigned char type, bestType = 0; + + for(type = 0; type != 5; ++type) + { + attempt[type] = (unsigned char*)lodepng_malloc(linebytes); + if(!attempt[type]) return 83; /*alloc fail*/ + } + + if(!error) + { + for(y = 0; y != h; ++y) + { + /*try the 5 filter types*/ + for(type = 0; type != 5; ++type) + { + filterScanline(attempt[type], &in[y * linebytes], prevline, linebytes, bytewidth, type); + + /*calculate the sum of the result*/ + sum[type] = 0; + if(type == 0) + { + for(x = 0; x != linebytes; ++x) sum[type] += (unsigned char)(attempt[type][x]); + } + else + { + for(x = 0; x != linebytes; ++x) + { + /*For differences, each byte should be treated as signed, values above 127 are negative + (converted to signed char). Filtertype 0 isn't a difference though, so use unsigned there. + This means filtertype 0 is almost never chosen, but that is justified.*/ + unsigned char s = attempt[type][x]; + sum[type] += s < 128 ? s : (255U - s); + } + } + + /*check if this is smallest sum (or if type == 0 it's the first case so always store the values)*/ + if(type == 0 || sum[type] < smallest) + { + bestType = type; + smallest = sum[type]; + } + } + + prevline = &in[y * linebytes]; + + /*now fill the out values*/ + out[y * (linebytes + 1)] = bestType; /*the first byte of a scanline will be the filter type*/ + for(x = 0; x != linebytes; ++x) out[y * (linebytes + 1) + 1 + x] = attempt[bestType][x]; + } + } + + for(type = 0; type != 5; ++type) lodepng_free(attempt[type]); + } + else if(strategy == LFS_ENTROPY) + { + float sum[5]; + unsigned char* attempt[5]; /*five filtering attempts, one for each filter type*/ + float smallest = 0; + unsigned type, bestType = 0; + unsigned count[256]; + + for(type = 0; type != 5; ++type) + { + attempt[type] = (unsigned char*)lodepng_malloc(linebytes); + if(!attempt[type]) return 83; /*alloc fail*/ + } + + for(y = 0; y != h; ++y) + { + /*try the 5 filter types*/ + for(type = 0; type != 5; ++type) + { + filterScanline(attempt[type], &in[y * linebytes], prevline, linebytes, bytewidth, type); + for(x = 0; x != 256; ++x) count[x] = 0; + for(x = 0; x != linebytes; ++x) ++count[attempt[type][x]]; + ++count[type]; /*the filter type itself is part of the scanline*/ + sum[type] = 0; + for(x = 0; x != 256; ++x) + { + float p = count[x] / (float)(linebytes + 1); + sum[type] += count[x] == 0 ? 0 : flog2(1 / p) * p; + } + /*check if this is smallest sum (or if type == 0 it's the first case so always store the values)*/ + if(type == 0 || sum[type] < smallest) + { + bestType = type; + smallest = sum[type]; + } + } + + prevline = &in[y * linebytes]; + + /*now fill the out values*/ + out[y * (linebytes + 1)] = bestType; /*the first byte of a scanline will be the filter type*/ + for(x = 0; x != linebytes; ++x) out[y * (linebytes + 1) + 1 + x] = attempt[bestType][x]; + } + + for(type = 0; type != 5; ++type) lodepng_free(attempt[type]); + } + else if(strategy == LFS_PREDEFINED) + { + for(y = 0; y != h; ++y) + { + size_t outindex = (1 + linebytes) * y; /*the extra filterbyte added to each row*/ + size_t inindex = linebytes * y; + unsigned char type = settings->predefined_filters[y]; + out[outindex] = type; /*filter type byte*/ + filterScanline(&out[outindex + 1], &in[inindex], prevline, linebytes, bytewidth, type); + prevline = &in[inindex]; + } + } + else if(strategy == LFS_BRUTE_FORCE) + { + /*brute force filter chooser. + deflate the scanline after every filter attempt to see which one deflates best. + This is very slow and gives only slightly smaller, sometimes even larger, result*/ + size_t size[5]; + unsigned char* attempt[5]; /*five filtering attempts, one for each filter type*/ + size_t smallest = 0; + unsigned type = 0, bestType = 0; + unsigned char* dummy; + LodePNGCompressSettings zlibsettings = settings->zlibsettings; + /*use fixed tree on the attempts so that the tree is not adapted to the filtertype on purpose, + to simulate the true case where the tree is the same for the whole image. Sometimes it gives + better result with dynamic tree anyway. Using the fixed tree sometimes gives worse, but in rare + cases better compression. It does make this a bit less slow, so it's worth doing this.*/ + zlibsettings.btype = 1; + /*a custom encoder likely doesn't read the btype setting and is optimized for complete PNG + images only, so disable it*/ + zlibsettings.custom_zlib = 0; + zlibsettings.custom_deflate = 0; + for(type = 0; type != 5; ++type) + { + attempt[type] = (unsigned char*)lodepng_malloc(linebytes); + if(!attempt[type]) return 83; /*alloc fail*/ + } + for(y = 0; y != h; ++y) /*try the 5 filter types*/ + { + for(type = 0; type != 5; ++type) + { + size_t testsize = linebytes; + /*if(testsize > 8) testsize /= 8;*/ /*it already works good enough by testing a part of the row*/ + + filterScanline(attempt[type], &in[y * linebytes], prevline, linebytes, bytewidth, type); + size[type] = 0; + dummy = 0; + zlib_compress(&dummy, &size[type], attempt[type], testsize, &zlibsettings); + lodepng_free(dummy); + /*check if this is smallest size (or if type == 0 it's the first case so always store the values)*/ + if(type == 0 || size[type] < smallest) + { + bestType = type; + smallest = size[type]; + } + } + prevline = &in[y * linebytes]; + out[y * (linebytes + 1)] = bestType; /*the first byte of a scanline will be the filter type*/ + for(x = 0; x != linebytes; ++x) out[y * (linebytes + 1) + 1 + x] = attempt[bestType][x]; + } + for(type = 0; type != 5; ++type) free(attempt[type]); + } + else return 88; /* unknown filter strategy */ + + return error; +} + +static void addPaddingBits(unsigned char* out, const unsigned char* in, + size_t olinebits, size_t ilinebits, unsigned h) +{ + /*The opposite of the removePaddingBits function + olinebits must be >= ilinebits*/ + unsigned y; + size_t diff = olinebits - ilinebits; + size_t obp = 0, ibp = 0; /*bit pointers*/ + for(y = 0; y != h; ++y) + { + size_t x; + for(x = 0; x < ilinebits; ++x) + { + unsigned char bit = readBitFromReversedStream(&ibp, in); + setBitOfReversedStream(&obp, out, bit); + } + /*obp += diff; --> no, fill in some value in the padding bits too, to avoid + "Use of uninitialised value of size ###" warning from valgrind*/ + for(x = 0; x != diff; ++x) setBitOfReversedStream(&obp, out, 0); + } +} + +/* +in: non-interlaced image with size w*h +out: the same pixels, but re-ordered according to PNG's Adam7 interlacing, with + no padding bits between scanlines, but between reduced images so that each + reduced image starts at a byte. +bpp: bits per pixel +there are no padding bits, not between scanlines, not between reduced images +in has the following size in bits: w * h * bpp. +out is possibly bigger due to padding bits between reduced images +NOTE: comments about padding bits are only relevant if bpp < 8 +*/ +static void Adam7_interlace(unsigned char* out, const unsigned char* in, unsigned w, unsigned h, unsigned bpp) +{ + unsigned passw[7], passh[7]; + size_t filter_passstart[8], padded_passstart[8], passstart[8]; + unsigned i; + + Adam7_getpassvalues(passw, passh, filter_passstart, padded_passstart, passstart, w, h, bpp); + + if(bpp >= 8) + { + for(i = 0; i != 7; ++i) + { + unsigned x, y, b; + size_t bytewidth = bpp / 8; + for(y = 0; y < passh[i]; ++y) + for(x = 0; x < passw[i]; ++x) + { + size_t pixelinstart = ((ADAM7_IY[i] + y * ADAM7_DY[i]) * w + ADAM7_IX[i] + x * ADAM7_DX[i]) * bytewidth; + size_t pixeloutstart = passstart[i] + (y * passw[i] + x) * bytewidth; + for(b = 0; b < bytewidth; ++b) + { + out[pixeloutstart + b] = in[pixelinstart + b]; + } + } + } + } + else /*bpp < 8: Adam7 with pixels < 8 bit is a bit trickier: with bit pointers*/ + { + for(i = 0; i != 7; ++i) + { + unsigned x, y, b; + unsigned ilinebits = bpp * passw[i]; + unsigned olinebits = bpp * w; + size_t obp, ibp; /*bit pointers (for out and in buffer)*/ + for(y = 0; y < passh[i]; ++y) + for(x = 0; x < passw[i]; ++x) + { + ibp = (ADAM7_IY[i] + y * ADAM7_DY[i]) * olinebits + (ADAM7_IX[i] + x * ADAM7_DX[i]) * bpp; + obp = (8 * passstart[i]) + (y * ilinebits + x * bpp); + for(b = 0; b < bpp; ++b) + { + unsigned char bit = readBitFromReversedStream(&ibp, in); + setBitOfReversedStream(&obp, out, bit); + } + } + } + } +} + +/*out must be buffer big enough to contain uncompressed IDAT chunk data, and in must contain the full image. +return value is error**/ +static unsigned preProcessScanlines(unsigned char** out, size_t* outsize, const unsigned char* in, + unsigned w, unsigned h, + const LodePNGInfo* info_png, const LodePNGEncoderSettings* settings) +{ + /* + This function converts the pure 2D image with the PNG's colortype, into filtered-padded-interlaced data. Steps: + *) if no Adam7: 1) add padding bits (= posible extra bits per scanline if bpp < 8) 2) filter + *) if adam7: 1) Adam7_interlace 2) 7x add padding bits 3) 7x filter + */ + unsigned bpp = lodepng_get_bpp(&info_png->color); + unsigned error = 0; + + if(info_png->interlace_method == 0) + { + *outsize = h + (h * ((w * bpp + 7) / 8)); /*image size plus an extra byte per scanline + possible padding bits*/ + *out = (unsigned char*)lodepng_malloc(*outsize); + if(!(*out) && (*outsize)) error = 83; /*alloc fail*/ + + if(!error) + { + /*non multiple of 8 bits per scanline, padding bits needed per scanline*/ + if(bpp < 8 && w * bpp != ((w * bpp + 7) / 8) * 8) + { + unsigned char* padded = (unsigned char*)lodepng_malloc(h * ((w * bpp + 7) / 8)); + if(!padded) error = 83; /*alloc fail*/ + if(!error) + { + addPaddingBits(padded, in, ((w * bpp + 7) / 8) * 8, w * bpp, h); + error = filter(*out, padded, w, h, &info_png->color, settings); + } + lodepng_free(padded); + } + else + { + /*we can immediately filter into the out buffer, no other steps needed*/ + error = filter(*out, in, w, h, &info_png->color, settings); + } + } + } + else /*interlace_method is 1 (Adam7)*/ + { + unsigned passw[7], passh[7]; + size_t filter_passstart[8], padded_passstart[8], passstart[8]; + unsigned char* adam7; + + Adam7_getpassvalues(passw, passh, filter_passstart, padded_passstart, passstart, w, h, bpp); + + *outsize = filter_passstart[7]; /*image size plus an extra byte per scanline + possible padding bits*/ + *out = (unsigned char*)lodepng_malloc(*outsize); + if(!(*out)) error = 83; /*alloc fail*/ + + adam7 = (unsigned char*)lodepng_malloc(passstart[7]); + if(!adam7 && passstart[7]) error = 83; /*alloc fail*/ + + if(!error) + { + unsigned i; + + Adam7_interlace(adam7, in, w, h, bpp); + for(i = 0; i != 7; ++i) + { + if(bpp < 8) + { + unsigned char* padded = (unsigned char*)lodepng_malloc(padded_passstart[i + 1] - padded_passstart[i]); + if(!padded) ERROR_BREAK(83); /*alloc fail*/ + addPaddingBits(padded, &adam7[passstart[i]], + ((passw[i] * bpp + 7) / 8) * 8, passw[i] * bpp, passh[i]); + error = filter(&(*out)[filter_passstart[i]], padded, + passw[i], passh[i], &info_png->color, settings); + lodepng_free(padded); + } + else + { + error = filter(&(*out)[filter_passstart[i]], &adam7[padded_passstart[i]], + passw[i], passh[i], &info_png->color, settings); + } + + if(error) break; + } + } + + lodepng_free(adam7); + } + + return error; +} + +/* +palette must have 4 * palettesize bytes allocated, and given in format RGBARGBARGBARGBA... +returns 0 if the palette is opaque, +returns 1 if the palette has a single color with alpha 0 ==> color key +returns 2 if the palette is semi-translucent. +*/ +static unsigned getPaletteTranslucency(const unsigned char* palette, size_t palettesize) +{ + size_t i; + unsigned key = 0; + unsigned r = 0, g = 0, b = 0; /*the value of the color with alpha 0, so long as color keying is possible*/ + for(i = 0; i != palettesize; ++i) + { + if(!key && palette[4 * i + 3] == 0) + { + r = palette[4 * i + 0]; g = palette[4 * i + 1]; b = palette[4 * i + 2]; + key = 1; + i = (size_t)(-1); /*restart from beginning, to detect earlier opaque colors with key's value*/ + } + else if(palette[4 * i + 3] != 255) return 2; + /*when key, no opaque RGB may have key's RGB*/ + else if(key && r == palette[i * 4 + 0] && g == palette[i * 4 + 1] && b == palette[i * 4 + 2]) return 2; + } + return key; +} + +#ifdef LODEPNG_COMPILE_ANCILLARY_CHUNKS +static unsigned addUnknownChunks(ucvector* out, unsigned char* data, size_t datasize) +{ + unsigned char* inchunk = data; + while((size_t)(inchunk - data) < datasize) + { + CERROR_TRY_RETURN(lodepng_chunk_append(&out->data, &out->size, inchunk)); + out->allocsize = out->size; /*fix the allocsize again*/ + inchunk = lodepng_chunk_next(inchunk); + } + return 0; +} +#endif /*LODEPNG_COMPILE_ANCILLARY_CHUNKS*/ + +unsigned lodepng_encode(unsigned char** out, size_t* outsize, + const unsigned char* image, unsigned w, unsigned h, + LodePNGState* state) +{ + LodePNGInfo info; + ucvector outv; + unsigned char* data = 0; /*uncompressed version of the IDAT chunk data*/ + size_t datasize = 0; + + /*provide some proper output values if error will happen*/ + *out = 0; + *outsize = 0; + state->error = 0; + + lodepng_info_init(&info); + lodepng_info_copy(&info, &state->info_png); + + if((info.color.colortype == LCT_PALETTE || state->encoder.force_palette) + && (info.color.palettesize == 0 || info.color.palettesize > 256)) + { + state->error = 68; /*invalid palette size, it is only allowed to be 1-256*/ + return state->error; + } + + if(state->encoder.auto_convert) + { + state->error = lodepng_auto_choose_color(&info.color, image, w, h, &state->info_raw); + } + if(state->error) return state->error; + + if(state->encoder.zlibsettings.btype > 2) + { + CERROR_RETURN_ERROR(state->error, 61); /*error: unexisting btype*/ + } + if(state->info_png.interlace_method > 1) + { + CERROR_RETURN_ERROR(state->error, 71); /*error: unexisting interlace mode*/ + } + + state->error = checkColorValidity(info.color.colortype, info.color.bitdepth); + if(state->error) return state->error; /*error: unexisting color type given*/ + state->error = checkColorValidity(state->info_raw.colortype, state->info_raw.bitdepth); + if(state->error) return state->error; /*error: unexisting color type given*/ + + if(!lodepng_color_mode_equal(&state->info_raw, &info.color)) + { + unsigned char* converted; + size_t size = (w * h * lodepng_get_bpp(&info.color) + 7) / 8; + + converted = (unsigned char*)lodepng_malloc(size); + if(!converted && size) state->error = 83; /*alloc fail*/ + if(!state->error) + { + state->error = lodepng_convert(converted, image, &info.color, &state->info_raw, w, h); + } + if(!state->error) preProcessScanlines(&data, &datasize, converted, w, h, &info, &state->encoder); + lodepng_free(converted); + } + else preProcessScanlines(&data, &datasize, image, w, h, &info, &state->encoder); + + ucvector_init(&outv); + while(!state->error) /*while only executed once, to break on error*/ + { +#ifdef LODEPNG_COMPILE_ANCILLARY_CHUNKS + size_t i; +#endif /*LODEPNG_COMPILE_ANCILLARY_CHUNKS*/ + /*write signature and chunks*/ + writeSignature(&outv); + /*IHDR*/ + addChunk_IHDR(&outv, w, h, info.color.colortype, info.color.bitdepth, info.interlace_method); +#ifdef LODEPNG_COMPILE_ANCILLARY_CHUNKS + /*unknown chunks between IHDR and PLTE*/ + if(info.unknown_chunks_data[0]) + { + state->error = addUnknownChunks(&outv, info.unknown_chunks_data[0], info.unknown_chunks_size[0]); + if(state->error) break; + } +#endif /*LODEPNG_COMPILE_ANCILLARY_CHUNKS*/ + /*PLTE*/ + if(info.color.colortype == LCT_PALETTE) + { + addChunk_PLTE(&outv, &info.color); + } + if(state->encoder.force_palette && (info.color.colortype == LCT_RGB || info.color.colortype == LCT_RGBA)) + { + addChunk_PLTE(&outv, &info.color); + } + /*tRNS*/ + if(info.color.colortype == LCT_PALETTE && getPaletteTranslucency(info.color.palette, info.color.palettesize) != 0) + { + addChunk_tRNS(&outv, &info.color); + } + if((info.color.colortype == LCT_GREY || info.color.colortype == LCT_RGB) && info.color.key_defined) + { + addChunk_tRNS(&outv, &info.color); + } +#ifdef LODEPNG_COMPILE_ANCILLARY_CHUNKS + /*bKGD (must come between PLTE and the IDAt chunks*/ + if(info.background_defined) addChunk_bKGD(&outv, &info); + /*pHYs (must come before the IDAT chunks)*/ + if(info.phys_defined) addChunk_pHYs(&outv, &info); + + /*unknown chunks between PLTE and IDAT*/ + if(info.unknown_chunks_data[1]) + { + state->error = addUnknownChunks(&outv, info.unknown_chunks_data[1], info.unknown_chunks_size[1]); + if(state->error) break; + } +#endif /*LODEPNG_COMPILE_ANCILLARY_CHUNKS*/ + /*IDAT (multiple IDAT chunks must be consecutive)*/ + state->error = addChunk_IDAT(&outv, data, datasize, &state->encoder.zlibsettings); + if(state->error) break; +#ifdef LODEPNG_COMPILE_ANCILLARY_CHUNKS + /*tIME*/ + if(info.time_defined) addChunk_tIME(&outv, &info.time); + /*tEXt and/or zTXt*/ + for(i = 0; i != info.text_num; ++i) + { + if(strlen(info.text_keys[i]) > 79) + { + state->error = 66; /*text chunk too large*/ + break; + } + if(strlen(info.text_keys[i]) < 1) + { + state->error = 67; /*text chunk too small*/ + break; + } + if(state->encoder.text_compression) + { + addChunk_zTXt(&outv, info.text_keys[i], info.text_strings[i], &state->encoder.zlibsettings); + } + else + { + addChunk_tEXt(&outv, info.text_keys[i], info.text_strings[i]); + } + } + /*LodePNG version id in text chunk*/ + if(state->encoder.add_id) + { + unsigned alread_added_id_text = 0; + for(i = 0; i != info.text_num; ++i) + { + if(!strcmp(info.text_keys[i], "LodePNG")) + { + alread_added_id_text = 1; + break; + } + } + if(alread_added_id_text == 0) + { + addChunk_tEXt(&outv, "LodePNG", LODEPNG_VERSION_STRING); /*it's shorter as tEXt than as zTXt chunk*/ + } + } + /*iTXt*/ + for(i = 0; i != info.itext_num; ++i) + { + if(strlen(info.itext_keys[i]) > 79) + { + state->error = 66; /*text chunk too large*/ + break; + } + if(strlen(info.itext_keys[i]) < 1) + { + state->error = 67; /*text chunk too small*/ + break; + } + addChunk_iTXt(&outv, state->encoder.text_compression, + info.itext_keys[i], info.itext_langtags[i], info.itext_transkeys[i], info.itext_strings[i], + &state->encoder.zlibsettings); + } + + /*unknown chunks between IDAT and IEND*/ + if(info.unknown_chunks_data[2]) + { + state->error = addUnknownChunks(&outv, info.unknown_chunks_data[2], info.unknown_chunks_size[2]); + if(state->error) break; + } +#endif /*LODEPNG_COMPILE_ANCILLARY_CHUNKS*/ + addChunk_IEND(&outv); + + break; /*this isn't really a while loop; no error happened so break out now!*/ + } + + lodepng_info_cleanup(&info); + lodepng_free(data); + /*instead of cleaning the vector up, give it to the output*/ + *out = outv.data; + *outsize = outv.size; + + return state->error; +} + +unsigned lodepng_encode_memory(unsigned char** out, size_t* outsize, const unsigned char* image, + unsigned w, unsigned h, LodePNGColorType colortype, unsigned bitdepth) +{ + unsigned error; + LodePNGState state; + lodepng_state_init(&state); + state.info_raw.colortype = colortype; + state.info_raw.bitdepth = bitdepth; + state.info_png.color.colortype = colortype; + state.info_png.color.bitdepth = bitdepth; + lodepng_encode(out, outsize, image, w, h, &state); + error = state.error; + lodepng_state_cleanup(&state); + return error; +} + +unsigned lodepng_encode32(unsigned char** out, size_t* outsize, const unsigned char* image, unsigned w, unsigned h) +{ + return lodepng_encode_memory(out, outsize, image, w, h, LCT_RGBA, 8); +} + +unsigned lodepng_encode24(unsigned char** out, size_t* outsize, const unsigned char* image, unsigned w, unsigned h) +{ + return lodepng_encode_memory(out, outsize, image, w, h, LCT_RGB, 8); +} + +#ifdef LODEPNG_COMPILE_DISK +unsigned lodepng_encode_file(const char* filename, const unsigned char* image, unsigned w, unsigned h, + LodePNGColorType colortype, unsigned bitdepth) +{ + unsigned char* buffer; + size_t buffersize; + unsigned error = lodepng_encode_memory(&buffer, &buffersize, image, w, h, colortype, bitdepth); + if(!error) error = lodepng_save_file(buffer, buffersize, filename); + lodepng_free(buffer); + return error; +} + +unsigned lodepng_encode32_file(const char* filename, const unsigned char* image, unsigned w, unsigned h) +{ + return lodepng_encode_file(filename, image, w, h, LCT_RGBA, 8); +} + +unsigned lodepng_encode24_file(const char* filename, const unsigned char* image, unsigned w, unsigned h) +{ + return lodepng_encode_file(filename, image, w, h, LCT_RGB, 8); +} +#endif /*LODEPNG_COMPILE_DISK*/ + +void lodepng_encoder_settings_init(LodePNGEncoderSettings* settings) +{ + lodepng_compress_settings_init(&settings->zlibsettings); + settings->filter_palette_zero = 1; + settings->filter_strategy = LFS_MINSUM; + settings->auto_convert = 1; + settings->force_palette = 0; + settings->predefined_filters = 0; +#ifdef LODEPNG_COMPILE_ANCILLARY_CHUNKS + settings->add_id = 0; + settings->text_compression = 1; +#endif /*LODEPNG_COMPILE_ANCILLARY_CHUNKS*/ +} + +#endif /*LODEPNG_COMPILE_ENCODER*/ +#endif /*LODEPNG_COMPILE_PNG*/ + +#ifdef LODEPNG_COMPILE_ERROR_TEXT +/* +This returns the description of a numerical error code in English. This is also +the documentation of all the error codes. +*/ +const char* lodepng_error_text(unsigned code) +{ + switch(code) + { + case 0: return "no error, everything went ok"; + case 1: return "nothing done yet"; /*the Encoder/Decoder has done nothing yet, error checking makes no sense yet*/ + case 10: return "end of input memory reached without huffman end code"; /*while huffman decoding*/ + case 11: return "error in code tree made it jump outside of huffman tree"; /*while huffman decoding*/ + case 13: return "problem while processing dynamic deflate block"; + case 14: return "problem while processing dynamic deflate block"; + case 15: return "problem while processing dynamic deflate block"; + case 16: return "unexisting code while processing dynamic deflate block"; + case 17: return "end of out buffer memory reached while inflating"; + case 18: return "invalid distance code while inflating"; + case 19: return "end of out buffer memory reached while inflating"; + case 20: return "invalid deflate block BTYPE encountered while decoding"; + case 21: return "NLEN is not ones complement of LEN in a deflate block"; + /*end of out buffer memory reached while inflating: + This can happen if the inflated deflate data is longer than the amount of bytes required to fill up + all the pixels of the image, given the color depth and image dimensions. Something that doesn't + happen in a normal, well encoded, PNG image.*/ + case 22: return "end of out buffer memory reached while inflating"; + case 23: return "end of in buffer memory reached while inflating"; + case 24: return "invalid FCHECK in zlib header"; + case 25: return "invalid compression method in zlib header"; + case 26: return "FDICT encountered in zlib header while it's not used for PNG"; + case 27: return "PNG file is smaller than a PNG header"; + /*Checks the magic file header, the first 8 bytes of the PNG file*/ + case 28: return "incorrect PNG signature, it's no PNG or corrupted"; + case 29: return "first chunk is not the header chunk"; + case 30: return "chunk length too large, chunk broken off at end of file"; + case 31: return "illegal PNG color type or bpp"; + case 32: return "illegal PNG compression method"; + case 33: return "illegal PNG filter method"; + case 34: return "illegal PNG interlace method"; + case 35: return "chunk length of a chunk is too large or the chunk too small"; + case 36: return "illegal PNG filter type encountered"; + case 37: return "illegal bit depth for this color type given"; + case 38: return "the palette is too big"; /*more than 256 colors*/ + case 39: return "more palette alpha values given in tRNS chunk than there are colors in the palette"; + case 40: return "tRNS chunk has wrong size for greyscale image"; + case 41: return "tRNS chunk has wrong size for RGB image"; + case 42: return "tRNS chunk appeared while it was not allowed for this color type"; + case 43: return "bKGD chunk has wrong size for palette image"; + case 44: return "bKGD chunk has wrong size for greyscale image"; + case 45: return "bKGD chunk has wrong size for RGB image"; + case 48: return "empty input buffer given to decoder. Maybe caused by non-existing file?"; + case 49: return "jumped past memory while generating dynamic huffman tree"; + case 50: return "jumped past memory while generating dynamic huffman tree"; + case 51: return "jumped past memory while inflating huffman block"; + case 52: return "jumped past memory while inflating"; + case 53: return "size of zlib data too small"; + case 54: return "repeat symbol in tree while there was no value symbol yet"; + /*jumped past tree while generating huffman tree, this could be when the + tree will have more leaves than symbols after generating it out of the + given lenghts. They call this an oversubscribed dynamic bit lengths tree in zlib.*/ + case 55: return "jumped past tree while generating huffman tree"; + case 56: return "given output image colortype or bitdepth not supported for color conversion"; + case 57: return "invalid CRC encountered (checking CRC can be disabled)"; + case 58: return "invalid ADLER32 encountered (checking ADLER32 can be disabled)"; + case 59: return "requested color conversion not supported"; + case 60: return "invalid window size given in the settings of the encoder (must be 0-32768)"; + case 61: return "invalid BTYPE given in the settings of the encoder (only 0, 1 and 2 are allowed)"; + /*LodePNG leaves the choice of RGB to greyscale conversion formula to the user.*/ + case 62: return "conversion from color to greyscale not supported"; + case 63: return "length of a chunk too long, max allowed for PNG is 2147483647 bytes per chunk"; /*(2^31-1)*/ + /*this would result in the inability of a deflated block to ever contain an end code. It must be at least 1.*/ + case 64: return "the length of the END symbol 256 in the Huffman tree is 0"; + case 66: return "the length of a text chunk keyword given to the encoder is longer than the maximum of 79 bytes"; + case 67: return "the length of a text chunk keyword given to the encoder is smaller than the minimum of 1 byte"; + case 68: return "tried to encode a PLTE chunk with a palette that has less than 1 or more than 256 colors"; + case 69: return "unknown chunk type with 'critical' flag encountered by the decoder"; + case 71: return "unexisting interlace mode given to encoder (must be 0 or 1)"; + case 72: return "while decoding, unexisting compression method encountering in zTXt or iTXt chunk (it must be 0)"; + case 73: return "invalid tIME chunk size"; + case 74: return "invalid pHYs chunk size"; + /*length could be wrong, or data chopped off*/ + case 75: return "no null termination char found while decoding text chunk"; + case 76: return "iTXt chunk too short to contain required bytes"; + case 77: return "integer overflow in buffer size"; + case 78: return "failed to open file for reading"; /*file doesn't exist or couldn't be opened for reading*/ + case 79: return "failed to open file for writing"; + case 80: return "tried creating a tree of 0 symbols"; + case 81: return "lazy matching at pos 0 is impossible"; + case 82: return "color conversion to palette requested while a color isn't in palette"; + case 83: return "memory allocation failed"; + case 84: return "given image too small to contain all pixels to be encoded"; + case 86: return "impossible offset in lz77 encoding (internal bug)"; + case 87: return "must provide custom zlib function pointer if LODEPNG_COMPILE_ZLIB is not defined"; + case 88: return "invalid filter strategy given for LodePNGEncoderSettings.filter_strategy"; + case 89: return "text chunk keyword too short or long: must have size 1-79"; + /*the windowsize in the LodePNGCompressSettings. Requiring POT(==> & instead of %) makes encoding 12% faster.*/ + case 90: return "windowsize must be a power of two"; + case 91: return "invalid decompressed idat size"; + case 92: return "too many pixels, not supported"; + case 93: return "zero width or height is invalid"; + case 94: return "header chunk must have a size of 13 bytes"; + } + return "unknown error code"; +} +#endif /*LODEPNG_COMPILE_ERROR_TEXT*/ + +/* ////////////////////////////////////////////////////////////////////////// */ +/* ////////////////////////////////////////////////////////////////////////// */ +/* // C++ Wrapper // */ +/* ////////////////////////////////////////////////////////////////////////// */ +/* ////////////////////////////////////////////////////////////////////////// */ + +#ifdef LODEPNG_COMPILE_CPP +namespace lodepng +{ + +#ifdef LODEPNG_COMPILE_DISK +unsigned load_file(std::vector& buffer, const std::string& filename) +{ + std::ifstream file(filename.c_str(), std::ios::in|std::ios::binary|std::ios::ate); + if(!file) return 78; + + /*get filesize*/ + std::streamsize size = 0; + if(file.seekg(0, std::ios::end).good()) size = file.tellg(); + if(file.seekg(0, std::ios::beg).good()) size -= file.tellg(); + + /*read contents of the file into the vector*/ + buffer.resize(size_t(size)); + if(size > 0) file.read((char*)(&buffer[0]), size); + + return 0; /* OK */ +} + +/*write given buffer to the file, overwriting the file, it doesn't append to it.*/ +unsigned save_file(const std::vector& buffer, const std::string& filename) +{ + std::ofstream file(filename.c_str(), std::ios::out|std::ios::binary); + if(!file) return 79; + file.write(buffer.empty() ? 0 : (char*)&buffer[0], std::streamsize(buffer.size())); + return 0; +} +#endif /* LODEPNG_COMPILE_DISK */ + +#ifdef LODEPNG_COMPILE_ZLIB +#ifdef LODEPNG_COMPILE_DECODER +unsigned decompress(std::vector& out, const unsigned char* in, size_t insize, + const LodePNGDecompressSettings& settings) +{ + unsigned char* buffer = 0; + size_t buffersize = 0; + unsigned error = zlib_decompress(&buffer, &buffersize, in, insize, &settings); + if(buffer) + { + out.insert(out.end(), &buffer[0], &buffer[buffersize]); + lodepng_free(buffer); + } + return error; +} + +unsigned decompress(std::vector& out, const std::vector& in, + const LodePNGDecompressSettings& settings) +{ + return decompress(out, in.empty() ? 0 : &in[0], in.size(), settings); +} +#endif /* LODEPNG_COMPILE_DECODER */ + +#ifdef LODEPNG_COMPILE_ENCODER +unsigned compress(std::vector& out, const unsigned char* in, size_t insize, + const LodePNGCompressSettings& settings) +{ + unsigned char* buffer = 0; + size_t buffersize = 0; + unsigned error = zlib_compress(&buffer, &buffersize, in, insize, &settings); + if(buffer) + { + out.insert(out.end(), &buffer[0], &buffer[buffersize]); + lodepng_free(buffer); + } + return error; +} + +unsigned compress(std::vector& out, const std::vector& in, + const LodePNGCompressSettings& settings) +{ + return compress(out, in.empty() ? 0 : &in[0], in.size(), settings); +} +#endif /* LODEPNG_COMPILE_ENCODER */ +#endif /* LODEPNG_COMPILE_ZLIB */ + + +#ifdef LODEPNG_COMPILE_PNG + +State::State() +{ + lodepng_state_init(this); +} + +State::State(const State& other) +{ + lodepng_state_init(this); + lodepng_state_copy(this, &other); +} + +State::~State() +{ + lodepng_state_cleanup(this); +} + +State& State::operator=(const State& other) +{ + lodepng_state_copy(this, &other); + return *this; +} + +#ifdef LODEPNG_COMPILE_DECODER + +unsigned decode(std::vector& out, unsigned& w, unsigned& h, const unsigned char* in, + size_t insize, LodePNGColorType colortype, unsigned bitdepth) +{ + unsigned char* buffer; + unsigned error = lodepng_decode_memory(&buffer, &w, &h, in, insize, colortype, bitdepth); + if(buffer && !error) + { + State state; + state.info_raw.colortype = colortype; + state.info_raw.bitdepth = bitdepth; + size_t buffersize = lodepng_get_raw_size(w, h, &state.info_raw); + out.insert(out.end(), &buffer[0], &buffer[buffersize]); + lodepng_free(buffer); + } + return error; +} + +unsigned decode(std::vector& out, unsigned& w, unsigned& h, + const std::vector& in, LodePNGColorType colortype, unsigned bitdepth) +{ + return decode(out, w, h, in.empty() ? 0 : &in[0], (unsigned)in.size(), colortype, bitdepth); +} + +unsigned decode(std::vector& out, unsigned& w, unsigned& h, + State& state, + const unsigned char* in, size_t insize) +{ + unsigned char* buffer = NULL; + unsigned error = lodepng_decode(&buffer, &w, &h, &state, in, insize); + if(buffer && !error) + { + size_t buffersize = lodepng_get_raw_size(w, h, &state.info_raw); + out.insert(out.end(), &buffer[0], &buffer[buffersize]); + } + lodepng_free(buffer); + return error; +} + +unsigned decode(std::vector& out, unsigned& w, unsigned& h, + State& state, + const std::vector& in) +{ + return decode(out, w, h, state, in.empty() ? 0 : &in[0], in.size()); +} + +#ifdef LODEPNG_COMPILE_DISK +unsigned decode(std::vector& out, unsigned& w, unsigned& h, const std::string& filename, + LodePNGColorType colortype, unsigned bitdepth) +{ + std::vector buffer; + unsigned error = load_file(buffer, filename); + if(error) return error; + return decode(out, w, h, buffer, colortype, bitdepth); +} +#endif /* LODEPNG_COMPILE_DECODER */ +#endif /* LODEPNG_COMPILE_DISK */ + +#ifdef LODEPNG_COMPILE_ENCODER +unsigned encode(std::vector& out, const unsigned char* in, unsigned w, unsigned h, + LodePNGColorType colortype, unsigned bitdepth) +{ + unsigned char* buffer; + size_t buffersize; + unsigned error = lodepng_encode_memory(&buffer, &buffersize, in, w, h, colortype, bitdepth); + if(buffer) + { + out.insert(out.end(), &buffer[0], &buffer[buffersize]); + lodepng_free(buffer); + } + return error; +} + +unsigned encode(std::vector& out, + const std::vector& in, unsigned w, unsigned h, + LodePNGColorType colortype, unsigned bitdepth) +{ + if(lodepng_get_raw_size_lct(w, h, colortype, bitdepth) > in.size()) return 84; + return encode(out, in.empty() ? 0 : &in[0], w, h, colortype, bitdepth); +} + +unsigned encode(std::vector& out, + const unsigned char* in, unsigned w, unsigned h, + State& state) +{ + unsigned char* buffer; + size_t buffersize; + unsigned error = lodepng_encode(&buffer, &buffersize, in, w, h, &state); + if(buffer) + { + out.insert(out.end(), &buffer[0], &buffer[buffersize]); + lodepng_free(buffer); + } + return error; +} + +unsigned encode(std::vector& out, + const std::vector& in, unsigned w, unsigned h, + State& state) +{ + if(lodepng_get_raw_size(w, h, &state.info_raw) > in.size()) return 84; + return encode(out, in.empty() ? 0 : &in[0], w, h, state); +} + +#ifdef LODEPNG_COMPILE_DISK +unsigned encode(const std::string& filename, + const unsigned char* in, unsigned w, unsigned h, + LodePNGColorType colortype, unsigned bitdepth) +{ + std::vector buffer; + unsigned error = encode(buffer, in, w, h, colortype, bitdepth); + if(!error) error = save_file(buffer, filename); + return error; +} + +unsigned encode(const std::string& filename, + const std::vector& in, unsigned w, unsigned h, + LodePNGColorType colortype, unsigned bitdepth) +{ + if(lodepng_get_raw_size_lct(w, h, colortype, bitdepth) > in.size()) return 84; + return encode(filename, in.empty() ? 0 : &in[0], w, h, colortype, bitdepth); +} +#endif /* LODEPNG_COMPILE_DISK */ +#endif /* LODEPNG_COMPILE_ENCODER */ +#endif /* LODEPNG_COMPILE_PNG */ +} /* namespace lodepng */ +#endif /*LODEPNG_COMPILE_CPP*/ diff --git a/third_party/lodepng/lodepng.h b/third_party/lodepng/lodepng.h new file mode 100644 index 0000000..f121ab3 --- /dev/null +++ b/third_party/lodepng/lodepng.h @@ -0,0 +1,1756 @@ +/* +LodePNG version 20160124 + +Copyright (c) 2005-2016 Lode Vandevenne + +This software is provided 'as-is', without any express or implied +warranty. In no event will the authors be held liable for any damages +arising from the use of this software. + +Permission is granted to anyone to use this software for any purpose, +including commercial applications, and to alter it and redistribute it +freely, subject to the following restrictions: + + 1. The origin of this software must not be misrepresented; you must not + claim that you wrote the original software. If you use this software + in a product, an acknowledgment in the product documentation would be + appreciated but is not required. + + 2. Altered source versions must be plainly marked as such, and must not be + misrepresented as being the original software. + + 3. This notice may not be removed or altered from any source + distribution. +*/ + +#ifndef LODEPNG_H +#define LODEPNG_H + +#include /*for size_t*/ + +extern const char* LODEPNG_VERSION_STRING; + +/* +The following #defines are used to create code sections. They can be disabled +to disable code sections, which can give faster compile time and smaller binary. +The "NO_COMPILE" defines are designed to be used to pass as defines to the +compiler command to disable them without modifying this header, e.g. +-DLODEPNG_NO_COMPILE_ZLIB for gcc. +In addition to those below, you can also define LODEPNG_NO_COMPILE_CRC to +allow implementing a custom lodepng_crc32. +*/ +/*deflate & zlib. If disabled, you must specify alternative zlib functions in +the custom_zlib field of the compress and decompress settings*/ +#ifndef LODEPNG_NO_COMPILE_ZLIB +#define LODEPNG_COMPILE_ZLIB +#endif +/*png encoder and png decoder*/ +#ifndef LODEPNG_NO_COMPILE_PNG +#define LODEPNG_COMPILE_PNG +#endif +/*deflate&zlib decoder and png decoder*/ +#ifndef LODEPNG_NO_COMPILE_DECODER +#define LODEPNG_COMPILE_DECODER +#endif +/*deflate&zlib encoder and png encoder*/ +#ifndef LODEPNG_NO_COMPILE_ENCODER +#define LODEPNG_COMPILE_ENCODER +#endif +/*the optional built in harddisk file loading and saving functions*/ +#ifndef LODEPNG_NO_COMPILE_DISK +#define LODEPNG_COMPILE_DISK +#endif +/*support for chunks other than IHDR, IDAT, PLTE, tRNS, IEND: ancillary and unknown chunks*/ +#ifndef LODEPNG_NO_COMPILE_ANCILLARY_CHUNKS +#define LODEPNG_COMPILE_ANCILLARY_CHUNKS +#endif +/*ability to convert error numerical codes to English text string*/ +#ifndef LODEPNG_NO_COMPILE_ERROR_TEXT +#define LODEPNG_COMPILE_ERROR_TEXT +#endif +/*Compile the default allocators (C's free, malloc and realloc). If you disable this, +you can define the functions lodepng_free, lodepng_malloc and lodepng_realloc in your +source files with custom allocators.*/ +#ifndef LODEPNG_NO_COMPILE_ALLOCATORS +#define LODEPNG_COMPILE_ALLOCATORS +#endif +/*compile the C++ version (you can disable the C++ wrapper here even when compiling for C++)*/ +#ifdef __cplusplus +#ifndef LODEPNG_NO_COMPILE_CPP +#define LODEPNG_COMPILE_CPP +#endif +#endif + +#ifdef LODEPNG_COMPILE_CPP +#include +#include +#endif /*LODEPNG_COMPILE_CPP*/ + +#ifdef LODEPNG_COMPILE_PNG +/*The PNG color types (also used for raw).*/ +typedef enum LodePNGColorType +{ + LCT_GREY = 0, /*greyscale: 1,2,4,8,16 bit*/ + LCT_RGB = 2, /*RGB: 8,16 bit*/ + LCT_PALETTE = 3, /*palette: 1,2,4,8 bit*/ + LCT_GREY_ALPHA = 4, /*greyscale with alpha: 8,16 bit*/ + LCT_RGBA = 6 /*RGB with alpha: 8,16 bit*/ +} LodePNGColorType; + +#ifdef LODEPNG_COMPILE_DECODER +/* +Converts PNG data in memory to raw pixel data. +out: Output parameter. Pointer to buffer that will contain the raw pixel data. + After decoding, its size is w * h * (bytes per pixel) bytes larger than + initially. Bytes per pixel depends on colortype and bitdepth. + Must be freed after usage with free(*out). + Note: for 16-bit per channel colors, uses big endian format like PNG does. +w: Output parameter. Pointer to width of pixel data. +h: Output parameter. Pointer to height of pixel data. +in: Memory buffer with the PNG file. +insize: size of the in buffer. +colortype: the desired color type for the raw output image. See explanation on PNG color types. +bitdepth: the desired bit depth for the raw output image. See explanation on PNG color types. +Return value: LodePNG error code (0 means no error). +*/ +unsigned lodepng_decode_memory(unsigned char** out, unsigned* w, unsigned* h, + const unsigned char* in, size_t insize, + LodePNGColorType colortype, unsigned bitdepth); + +/*Same as lodepng_decode_memory, but always decodes to 32-bit RGBA raw image*/ +unsigned lodepng_decode32(unsigned char** out, unsigned* w, unsigned* h, + const unsigned char* in, size_t insize); + +/*Same as lodepng_decode_memory, but always decodes to 24-bit RGB raw image*/ +unsigned lodepng_decode24(unsigned char** out, unsigned* w, unsigned* h, + const unsigned char* in, size_t insize); + +#ifdef LODEPNG_COMPILE_DISK +/* +Load PNG from disk, from file with given name. +Same as the other decode functions, but instead takes a filename as input. +*/ +unsigned lodepng_decode_file(unsigned char** out, unsigned* w, unsigned* h, + const char* filename, + LodePNGColorType colortype, unsigned bitdepth); + +/*Same as lodepng_decode_file, but always decodes to 32-bit RGBA raw image.*/ +unsigned lodepng_decode32_file(unsigned char** out, unsigned* w, unsigned* h, + const char* filename); + +/*Same as lodepng_decode_file, but always decodes to 24-bit RGB raw image.*/ +unsigned lodepng_decode24_file(unsigned char** out, unsigned* w, unsigned* h, + const char* filename); +#endif /*LODEPNG_COMPILE_DISK*/ +#endif /*LODEPNG_COMPILE_DECODER*/ + + +#ifdef LODEPNG_COMPILE_ENCODER +/* +Converts raw pixel data into a PNG image in memory. The colortype and bitdepth + of the output PNG image cannot be chosen, they are automatically determined + by the colortype, bitdepth and content of the input pixel data. + Note: for 16-bit per channel colors, needs big endian format like PNG does. +out: Output parameter. Pointer to buffer that will contain the PNG image data. + Must be freed after usage with free(*out). +outsize: Output parameter. Pointer to the size in bytes of the out buffer. +image: The raw pixel data to encode. The size of this buffer should be + w * h * (bytes per pixel), bytes per pixel depends on colortype and bitdepth. +w: width of the raw pixel data in pixels. +h: height of the raw pixel data in pixels. +colortype: the color type of the raw input image. See explanation on PNG color types. +bitdepth: the bit depth of the raw input image. See explanation on PNG color types. +Return value: LodePNG error code (0 means no error). +*/ +unsigned lodepng_encode_memory(unsigned char** out, size_t* outsize, + const unsigned char* image, unsigned w, unsigned h, + LodePNGColorType colortype, unsigned bitdepth); + +/*Same as lodepng_encode_memory, but always encodes from 32-bit RGBA raw image.*/ +unsigned lodepng_encode32(unsigned char** out, size_t* outsize, + const unsigned char* image, unsigned w, unsigned h); + +/*Same as lodepng_encode_memory, but always encodes from 24-bit RGB raw image.*/ +unsigned lodepng_encode24(unsigned char** out, size_t* outsize, + const unsigned char* image, unsigned w, unsigned h); + +#ifdef LODEPNG_COMPILE_DISK +/* +Converts raw pixel data into a PNG file on disk. +Same as the other encode functions, but instead takes a filename as output. +NOTE: This overwrites existing files without warning! +*/ +unsigned lodepng_encode_file(const char* filename, + const unsigned char* image, unsigned w, unsigned h, + LodePNGColorType colortype, unsigned bitdepth); + +/*Same as lodepng_encode_file, but always encodes from 32-bit RGBA raw image.*/ +unsigned lodepng_encode32_file(const char* filename, + const unsigned char* image, unsigned w, unsigned h); + +/*Same as lodepng_encode_file, but always encodes from 24-bit RGB raw image.*/ +unsigned lodepng_encode24_file(const char* filename, + const unsigned char* image, unsigned w, unsigned h); +#endif /*LODEPNG_COMPILE_DISK*/ +#endif /*LODEPNG_COMPILE_ENCODER*/ + + +#ifdef LODEPNG_COMPILE_CPP +namespace lodepng +{ +#ifdef LODEPNG_COMPILE_DECODER +/*Same as lodepng_decode_memory, but decodes to an std::vector. The colortype +is the format to output the pixels to. Default is RGBA 8-bit per channel.*/ +unsigned decode(std::vector& out, unsigned& w, unsigned& h, + const unsigned char* in, size_t insize, + LodePNGColorType colortype = LCT_RGBA, unsigned bitdepth = 8); +unsigned decode(std::vector& out, unsigned& w, unsigned& h, + const std::vector& in, + LodePNGColorType colortype = LCT_RGBA, unsigned bitdepth = 8); +#ifdef LODEPNG_COMPILE_DISK +/* +Converts PNG file from disk to raw pixel data in memory. +Same as the other decode functions, but instead takes a filename as input. +*/ +unsigned decode(std::vector& out, unsigned& w, unsigned& h, + const std::string& filename, + LodePNGColorType colortype = LCT_RGBA, unsigned bitdepth = 8); +#endif /* LODEPNG_COMPILE_DISK */ +#endif /* LODEPNG_COMPILE_DECODER */ + +#ifdef LODEPNG_COMPILE_ENCODER +/*Same as lodepng_encode_memory, but encodes to an std::vector. colortype +is that of the raw input data. The output PNG color type will be auto chosen.*/ +unsigned encode(std::vector& out, + const unsigned char* in, unsigned w, unsigned h, + LodePNGColorType colortype = LCT_RGBA, unsigned bitdepth = 8); +unsigned encode(std::vector& out, + const std::vector& in, unsigned w, unsigned h, + LodePNGColorType colortype = LCT_RGBA, unsigned bitdepth = 8); +#ifdef LODEPNG_COMPILE_DISK +/* +Converts 32-bit RGBA raw pixel data into a PNG file on disk. +Same as the other encode functions, but instead takes a filename as output. +NOTE: This overwrites existing files without warning! +*/ +unsigned encode(const std::string& filename, + const unsigned char* in, unsigned w, unsigned h, + LodePNGColorType colortype = LCT_RGBA, unsigned bitdepth = 8); +unsigned encode(const std::string& filename, + const std::vector& in, unsigned w, unsigned h, + LodePNGColorType colortype = LCT_RGBA, unsigned bitdepth = 8); +#endif /* LODEPNG_COMPILE_DISK */ +#endif /* LODEPNG_COMPILE_ENCODER */ +} /* namespace lodepng */ +#endif /*LODEPNG_COMPILE_CPP*/ +#endif /*LODEPNG_COMPILE_PNG*/ + +#ifdef LODEPNG_COMPILE_ERROR_TEXT +/*Returns an English description of the numerical error code.*/ +const char* lodepng_error_text(unsigned code); +#endif /*LODEPNG_COMPILE_ERROR_TEXT*/ + +#ifdef LODEPNG_COMPILE_DECODER +/*Settings for zlib decompression*/ +typedef struct LodePNGDecompressSettings LodePNGDecompressSettings; +struct LodePNGDecompressSettings +{ + unsigned ignore_adler32; /*if 1, continue and don't give an error message if the Adler32 checksum is corrupted*/ + + /*use custom zlib decoder instead of built in one (default: null)*/ + unsigned (*custom_zlib)(unsigned char**, size_t*, + const unsigned char*, size_t, + const LodePNGDecompressSettings*); + /*use custom deflate decoder instead of built in one (default: null) + if custom_zlib is used, custom_deflate is ignored since only the built in + zlib function will call custom_deflate*/ + unsigned (*custom_inflate)(unsigned char**, size_t*, + const unsigned char*, size_t, + const LodePNGDecompressSettings*); + + const void* custom_context; /*optional custom settings for custom functions*/ +}; + +extern const LodePNGDecompressSettings lodepng_default_decompress_settings; +void lodepng_decompress_settings_init(LodePNGDecompressSettings* settings); +#endif /*LODEPNG_COMPILE_DECODER*/ + +#ifdef LODEPNG_COMPILE_ENCODER +/* +Settings for zlib compression. Tweaking these settings tweaks the balance +between speed and compression ratio. +*/ +typedef struct LodePNGCompressSettings LodePNGCompressSettings; +struct LodePNGCompressSettings /*deflate = compress*/ +{ + /*LZ77 related settings*/ + unsigned btype; /*the block type for LZ (0, 1, 2 or 3, see zlib standard). Should be 2 for proper compression.*/ + unsigned use_lz77; /*whether or not to use LZ77. Should be 1 for proper compression.*/ + unsigned windowsize; /*must be a power of two <= 32768. higher compresses more but is slower. Default value: 2048.*/ + unsigned minmatch; /*mininum lz77 length. 3 is normally best, 6 can be better for some PNGs. Default: 0*/ + unsigned nicematch; /*stop searching if >= this length found. Set to 258 for best compression. Default: 128*/ + unsigned lazymatching; /*use lazy matching: better compression but a bit slower. Default: true*/ + + /*use custom zlib encoder instead of built in one (default: null)*/ + unsigned (*custom_zlib)(unsigned char**, size_t*, + const unsigned char*, size_t, + const LodePNGCompressSettings*); + /*use custom deflate encoder instead of built in one (default: null) + if custom_zlib is used, custom_deflate is ignored since only the built in + zlib function will call custom_deflate*/ + unsigned (*custom_deflate)(unsigned char**, size_t*, + const unsigned char*, size_t, + const LodePNGCompressSettings*); + + const void* custom_context; /*optional custom settings for custom functions*/ +}; + +extern const LodePNGCompressSettings lodepng_default_compress_settings; +void lodepng_compress_settings_init(LodePNGCompressSettings* settings); +#endif /*LODEPNG_COMPILE_ENCODER*/ + +#ifdef LODEPNG_COMPILE_PNG +/* +Color mode of an image. Contains all information required to decode the pixel +bits to RGBA colors. This information is the same as used in the PNG file +format, and is used both for PNG and raw image data in LodePNG. +*/ +typedef struct LodePNGColorMode +{ + /*header (IHDR)*/ + LodePNGColorType colortype; /*color type, see PNG standard or documentation further in this header file*/ + unsigned bitdepth; /*bits per sample, see PNG standard or documentation further in this header file*/ + + /* + palette (PLTE and tRNS) + + Dynamically allocated with the colors of the palette, including alpha. + When encoding a PNG, to store your colors in the palette of the LodePNGColorMode, first use + lodepng_palette_clear, then for each color use lodepng_palette_add. + If you encode an image without alpha with palette, don't forget to put value 255 in each A byte of the palette. + + When decoding, by default you can ignore this palette, since LodePNG already + fills the palette colors in the pixels of the raw RGBA output. + + The palette is only supported for color type 3. + */ + unsigned char* palette; /*palette in RGBARGBA... order. When allocated, must be either 0, or have size 1024*/ + size_t palettesize; /*palette size in number of colors (amount of bytes is 4 * palettesize)*/ + + /* + transparent color key (tRNS) + + This color uses the same bit depth as the bitdepth value in this struct, which can be 1-bit to 16-bit. + For greyscale PNGs, r, g and b will all 3 be set to the same. + + When decoding, by default you can ignore this information, since LodePNG sets + pixels with this key to transparent already in the raw RGBA output. + + The color key is only supported for color types 0 and 2. + */ + unsigned key_defined; /*is a transparent color key given? 0 = false, 1 = true*/ + unsigned key_r; /*red/greyscale component of color key*/ + unsigned key_g; /*green component of color key*/ + unsigned key_b; /*blue component of color key*/ +} LodePNGColorMode; + +/*init, cleanup and copy functions to use with this struct*/ +void lodepng_color_mode_init(LodePNGColorMode* info); +void lodepng_color_mode_cleanup(LodePNGColorMode* info); +/*return value is error code (0 means no error)*/ +unsigned lodepng_color_mode_copy(LodePNGColorMode* dest, const LodePNGColorMode* source); + +void lodepng_palette_clear(LodePNGColorMode* info); +/*add 1 color to the palette*/ +unsigned lodepng_palette_add(LodePNGColorMode* info, + unsigned char r, unsigned char g, unsigned char b, unsigned char a); + +/*get the total amount of bits per pixel, based on colortype and bitdepth in the struct*/ +unsigned lodepng_get_bpp(const LodePNGColorMode* info); +/*get the amount of color channels used, based on colortype in the struct. +If a palette is used, it counts as 1 channel.*/ +unsigned lodepng_get_channels(const LodePNGColorMode* info); +/*is it a greyscale type? (only colortype 0 or 4)*/ +unsigned lodepng_is_greyscale_type(const LodePNGColorMode* info); +/*has it got an alpha channel? (only colortype 2 or 6)*/ +unsigned lodepng_is_alpha_type(const LodePNGColorMode* info); +/*has it got a palette? (only colortype 3)*/ +unsigned lodepng_is_palette_type(const LodePNGColorMode* info); +/*only returns true if there is a palette and there is a value in the palette with alpha < 255. +Loops through the palette to check this.*/ +unsigned lodepng_has_palette_alpha(const LodePNGColorMode* info); +/* +Check if the given color info indicates the possibility of having non-opaque pixels in the PNG image. +Returns true if the image can have translucent or invisible pixels (it still be opaque if it doesn't use such pixels). +Returns false if the image can only have opaque pixels. +In detail, it returns true only if it's a color type with alpha, or has a palette with non-opaque values, +or if "key_defined" is true. +*/ +unsigned lodepng_can_have_alpha(const LodePNGColorMode* info); +/*Returns the byte size of a raw image buffer with given width, height and color mode*/ +size_t lodepng_get_raw_size(unsigned w, unsigned h, const LodePNGColorMode* color); + +#ifdef LODEPNG_COMPILE_ANCILLARY_CHUNKS +/*The information of a Time chunk in PNG.*/ +typedef struct LodePNGTime +{ + unsigned year; /*2 bytes used (0-65535)*/ + unsigned month; /*1-12*/ + unsigned day; /*1-31*/ + unsigned hour; /*0-23*/ + unsigned minute; /*0-59*/ + unsigned second; /*0-60 (to allow for leap seconds)*/ +} LodePNGTime; +#endif /*LODEPNG_COMPILE_ANCILLARY_CHUNKS*/ + +/*Information about the PNG image, except pixels, width and height.*/ +typedef struct LodePNGInfo +{ + /*header (IHDR), palette (PLTE) and transparency (tRNS) chunks*/ + unsigned compression_method;/*compression method of the original file. Always 0.*/ + unsigned filter_method; /*filter method of the original file*/ + unsigned interlace_method; /*interlace method of the original file*/ + LodePNGColorMode color; /*color type and bits, palette and transparency of the PNG file*/ + +#ifdef LODEPNG_COMPILE_ANCILLARY_CHUNKS + /* + suggested background color chunk (bKGD) + This color uses the same color mode as the PNG (except alpha channel), which can be 1-bit to 16-bit. + + For greyscale PNGs, r, g and b will all 3 be set to the same. When encoding + the encoder writes the red one. For palette PNGs: When decoding, the RGB value + will be stored, not a palette index. But when encoding, specify the index of + the palette in background_r, the other two are then ignored. + + The decoder does not use this background color to edit the color of pixels. + */ + unsigned background_defined; /*is a suggested background color given?*/ + unsigned background_r; /*red component of suggested background color*/ + unsigned background_g; /*green component of suggested background color*/ + unsigned background_b; /*blue component of suggested background color*/ + + /* + non-international text chunks (tEXt and zTXt) + + The char** arrays each contain num strings. The actual messages are in + text_strings, while text_keys are keywords that give a short description what + the actual text represents, e.g. Title, Author, Description, or anything else. + + A keyword is minimum 1 character and maximum 79 characters long. It's + discouraged to use a single line length longer than 79 characters for texts. + + Don't allocate these text buffers yourself. Use the init/cleanup functions + correctly and use lodepng_add_text and lodepng_clear_text. + */ + size_t text_num; /*the amount of texts in these char** buffers (there may be more texts in itext)*/ + char** text_keys; /*the keyword of a text chunk (e.g. "Comment")*/ + char** text_strings; /*the actual text*/ + + /* + international text chunks (iTXt) + Similar to the non-international text chunks, but with additional strings + "langtags" and "transkeys". + */ + size_t itext_num; /*the amount of international texts in this PNG*/ + char** itext_keys; /*the English keyword of the text chunk (e.g. "Comment")*/ + char** itext_langtags; /*language tag for this text's language, ISO/IEC 646 string, e.g. ISO 639 language tag*/ + char** itext_transkeys; /*keyword translated to the international language - UTF-8 string*/ + char** itext_strings; /*the actual international text - UTF-8 string*/ + + /*time chunk (tIME)*/ + unsigned time_defined; /*set to 1 to make the encoder generate a tIME chunk*/ + LodePNGTime time; + + /*phys chunk (pHYs)*/ + unsigned phys_defined; /*if 0, there is no pHYs chunk and the values below are undefined, if 1 else there is one*/ + unsigned phys_x; /*pixels per unit in x direction*/ + unsigned phys_y; /*pixels per unit in y direction*/ + unsigned phys_unit; /*may be 0 (unknown unit) or 1 (metre)*/ + + /* + unknown chunks + There are 3 buffers, one for each position in the PNG where unknown chunks can appear + each buffer contains all unknown chunks for that position consecutively + The 3 buffers are the unknown chunks between certain critical chunks: + 0: IHDR-PLTE, 1: PLTE-IDAT, 2: IDAT-IEND + Do not allocate or traverse this data yourself. Use the chunk traversing functions declared + later, such as lodepng_chunk_next and lodepng_chunk_append, to read/write this struct. + */ + unsigned char* unknown_chunks_data[3]; + size_t unknown_chunks_size[3]; /*size in bytes of the unknown chunks, given for protection*/ +#endif /*LODEPNG_COMPILE_ANCILLARY_CHUNKS*/ +} LodePNGInfo; + +/*init, cleanup and copy functions to use with this struct*/ +void lodepng_info_init(LodePNGInfo* info); +void lodepng_info_cleanup(LodePNGInfo* info); +/*return value is error code (0 means no error)*/ +unsigned lodepng_info_copy(LodePNGInfo* dest, const LodePNGInfo* source); + +#ifdef LODEPNG_COMPILE_ANCILLARY_CHUNKS +void lodepng_clear_text(LodePNGInfo* info); /*use this to clear the texts again after you filled them in*/ +unsigned lodepng_add_text(LodePNGInfo* info, const char* key, const char* str); /*push back both texts at once*/ + +void lodepng_clear_itext(LodePNGInfo* info); /*use this to clear the itexts again after you filled them in*/ +unsigned lodepng_add_itext(LodePNGInfo* info, const char* key, const char* langtag, + const char* transkey, const char* str); /*push back the 4 texts of 1 chunk at once*/ +#endif /*LODEPNG_COMPILE_ANCILLARY_CHUNKS*/ + +/* +Converts raw buffer from one color type to another color type, based on +LodePNGColorMode structs to describe the input and output color type. +See the reference manual at the end of this header file to see which color conversions are supported. +return value = LodePNG error code (0 if all went ok, an error if the conversion isn't supported) +The out buffer must have size (w * h * bpp + 7) / 8, where bpp is the bits per pixel +of the output color type (lodepng_get_bpp). +For < 8 bpp images, there should not be padding bits at the end of scanlines. +For 16-bit per channel colors, uses big endian format like PNG does. +Return value is LodePNG error code +*/ +unsigned lodepng_convert(unsigned char* out, const unsigned char* in, + const LodePNGColorMode* mode_out, const LodePNGColorMode* mode_in, + unsigned w, unsigned h); + +#ifdef LODEPNG_COMPILE_DECODER +/* +Settings for the decoder. This contains settings for the PNG and the Zlib +decoder, but not the Info settings from the Info structs. +*/ +typedef struct LodePNGDecoderSettings +{ + LodePNGDecompressSettings zlibsettings; /*in here is the setting to ignore Adler32 checksums*/ + + unsigned ignore_crc; /*ignore CRC checksums*/ + + unsigned color_convert; /*whether to convert the PNG to the color type you want. Default: yes*/ + +#ifdef LODEPNG_COMPILE_ANCILLARY_CHUNKS + unsigned read_text_chunks; /*if false but remember_unknown_chunks is true, they're stored in the unknown chunks*/ + /*store all bytes from unknown chunks in the LodePNGInfo (off by default, useful for a png editor)*/ + unsigned remember_unknown_chunks; +#endif /*LODEPNG_COMPILE_ANCILLARY_CHUNKS*/ +} LodePNGDecoderSettings; + +void lodepng_decoder_settings_init(LodePNGDecoderSettings* settings); +#endif /*LODEPNG_COMPILE_DECODER*/ + +#ifdef LODEPNG_COMPILE_ENCODER +/*automatically use color type with less bits per pixel if losslessly possible. Default: AUTO*/ +typedef enum LodePNGFilterStrategy +{ + /*every filter at zero*/ + LFS_ZERO, + /*Use filter that gives minimum sum, as described in the official PNG filter heuristic.*/ + LFS_MINSUM, + /*Use the filter type that gives smallest Shannon entropy for this scanline. Depending + on the image, this is better or worse than minsum.*/ + LFS_ENTROPY, + /* + Brute-force-search PNG filters by compressing each filter for each scanline. + Experimental, very slow, and only rarely gives better compression than MINSUM. + */ + LFS_BRUTE_FORCE, + /*use predefined_filters buffer: you specify the filter type for each scanline*/ + LFS_PREDEFINED +} LodePNGFilterStrategy; + +/*Gives characteristics about the colors of the image, which helps decide which color model to use for encoding. +Used internally by default if "auto_convert" is enabled. Public because it's useful for custom algorithms.*/ +typedef struct LodePNGColorProfile +{ + unsigned colored; /*not greyscale*/ + unsigned key; /*if true, image is not opaque. Only if true and alpha is false, color key is possible.*/ + unsigned short key_r; /*these values are always in 16-bit bitdepth in the profile*/ + unsigned short key_g; + unsigned short key_b; + unsigned alpha; /*alpha channel or alpha palette required*/ + unsigned numcolors; /*amount of colors, up to 257. Not valid if bits == 16.*/ + unsigned char palette[1024]; /*Remembers up to the first 256 RGBA colors, in no particular order*/ + unsigned bits; /*bits per channel (not for palette). 1,2 or 4 for greyscale only. 16 if 16-bit per channel required.*/ +} LodePNGColorProfile; + +void lodepng_color_profile_init(LodePNGColorProfile* profile); + +/*Get a LodePNGColorProfile of the image.*/ +unsigned lodepng_get_color_profile(LodePNGColorProfile* profile, + const unsigned char* image, unsigned w, unsigned h, + const LodePNGColorMode* mode_in); +/*The function LodePNG uses internally to decide the PNG color with auto_convert. +Chooses an optimal color model, e.g. grey if only grey pixels, palette if < 256 colors, ...*/ +unsigned lodepng_auto_choose_color(LodePNGColorMode* mode_out, + const unsigned char* image, unsigned w, unsigned h, + const LodePNGColorMode* mode_in); + +/*Settings for the encoder.*/ +typedef struct LodePNGEncoderSettings +{ + LodePNGCompressSettings zlibsettings; /*settings for the zlib encoder, such as window size, ...*/ + + unsigned auto_convert; /*automatically choose output PNG color type. Default: true*/ + + /*If true, follows the official PNG heuristic: if the PNG uses a palette or lower than + 8 bit depth, set all filters to zero. Otherwise use the filter_strategy. Note that to + completely follow the official PNG heuristic, filter_palette_zero must be true and + filter_strategy must be LFS_MINSUM*/ + unsigned filter_palette_zero; + /*Which filter strategy to use when not using zeroes due to filter_palette_zero. + Set filter_palette_zero to 0 to ensure always using your chosen strategy. Default: LFS_MINSUM*/ + LodePNGFilterStrategy filter_strategy; + /*used if filter_strategy is LFS_PREDEFINED. In that case, this must point to a buffer with + the same length as the amount of scanlines in the image, and each value must <= 5. You + have to cleanup this buffer, LodePNG will never free it. Don't forget that filter_palette_zero + must be set to 0 to ensure this is also used on palette or low bitdepth images.*/ + const unsigned char* predefined_filters; + + /*force creating a PLTE chunk if colortype is 2 or 6 (= a suggested palette). + If colortype is 3, PLTE is _always_ created.*/ + unsigned force_palette; +#ifdef LODEPNG_COMPILE_ANCILLARY_CHUNKS + /*add LodePNG identifier and version as a text chunk, for debugging*/ + unsigned add_id; + /*encode text chunks as zTXt chunks instead of tEXt chunks, and use compression in iTXt chunks*/ + unsigned text_compression; +#endif /*LODEPNG_COMPILE_ANCILLARY_CHUNKS*/ +} LodePNGEncoderSettings; + +void lodepng_encoder_settings_init(LodePNGEncoderSettings* settings); +#endif /*LODEPNG_COMPILE_ENCODER*/ + + +#if defined(LODEPNG_COMPILE_DECODER) || defined(LODEPNG_COMPILE_ENCODER) +/*The settings, state and information for extended encoding and decoding.*/ +typedef struct LodePNGState +{ +#ifdef LODEPNG_COMPILE_DECODER + LodePNGDecoderSettings decoder; /*the decoding settings*/ +#endif /*LODEPNG_COMPILE_DECODER*/ +#ifdef LODEPNG_COMPILE_ENCODER + LodePNGEncoderSettings encoder; /*the encoding settings*/ +#endif /*LODEPNG_COMPILE_ENCODER*/ + LodePNGColorMode info_raw; /*specifies the format in which you would like to get the raw pixel buffer*/ + LodePNGInfo info_png; /*info of the PNG image obtained after decoding*/ + unsigned error; +#ifdef LODEPNG_COMPILE_CPP + /* For the lodepng::State subclass. */ + virtual ~LodePNGState(){} +#endif +} LodePNGState; + +/*init, cleanup and copy functions to use with this struct*/ +void lodepng_state_init(LodePNGState* state); +void lodepng_state_cleanup(LodePNGState* state); +void lodepng_state_copy(LodePNGState* dest, const LodePNGState* source); +#endif /* defined(LODEPNG_COMPILE_DECODER) || defined(LODEPNG_COMPILE_ENCODER) */ + +#ifdef LODEPNG_COMPILE_DECODER +/* +Same as lodepng_decode_memory, but uses a LodePNGState to allow custom settings and +getting much more information about the PNG image and color mode. +*/ +unsigned lodepng_decode(unsigned char** out, unsigned* w, unsigned* h, + LodePNGState* state, + const unsigned char* in, size_t insize); + +/* +Read the PNG header, but not the actual data. This returns only the information +that is in the header chunk of the PNG, such as width, height and color type. The +information is placed in the info_png field of the LodePNGState. +*/ +unsigned lodepng_inspect(unsigned* w, unsigned* h, + LodePNGState* state, + const unsigned char* in, size_t insize); +#endif /*LODEPNG_COMPILE_DECODER*/ + + +#ifdef LODEPNG_COMPILE_ENCODER +/*This function allocates the out buffer with standard malloc and stores the size in *outsize.*/ +unsigned lodepng_encode(unsigned char** out, size_t* outsize, + const unsigned char* image, unsigned w, unsigned h, + LodePNGState* state); +#endif /*LODEPNG_COMPILE_ENCODER*/ + +/* +The lodepng_chunk functions are normally not needed, except to traverse the +unknown chunks stored in the LodePNGInfo struct, or add new ones to it. +It also allows traversing the chunks of an encoded PNG file yourself. + +PNG standard chunk naming conventions: +First byte: uppercase = critical, lowercase = ancillary +Second byte: uppercase = public, lowercase = private +Third byte: must be uppercase +Fourth byte: uppercase = unsafe to copy, lowercase = safe to copy +*/ + +/* +Gets the length of the data of the chunk. Total chunk length has 12 bytes more. +There must be at least 4 bytes to read from. If the result value is too large, +it may be corrupt data. +*/ +unsigned lodepng_chunk_length(const unsigned char* chunk); + +/*puts the 4-byte type in null terminated string*/ +void lodepng_chunk_type(char type[5], const unsigned char* chunk); + +/*check if the type is the given type*/ +unsigned char lodepng_chunk_type_equals(const unsigned char* chunk, const char* type); + +/*0: it's one of the critical chunk types, 1: it's an ancillary chunk (see PNG standard)*/ +unsigned char lodepng_chunk_ancillary(const unsigned char* chunk); + +/*0: public, 1: private (see PNG standard)*/ +unsigned char lodepng_chunk_private(const unsigned char* chunk); + +/*0: the chunk is unsafe to copy, 1: the chunk is safe to copy (see PNG standard)*/ +unsigned char lodepng_chunk_safetocopy(const unsigned char* chunk); + +/*get pointer to the data of the chunk, where the input points to the header of the chunk*/ +unsigned char* lodepng_chunk_data(unsigned char* chunk); +const unsigned char* lodepng_chunk_data_const(const unsigned char* chunk); + +/*returns 0 if the crc is correct, 1 if it's incorrect (0 for OK as usual!)*/ +unsigned lodepng_chunk_check_crc(const unsigned char* chunk); + +/*generates the correct CRC from the data and puts it in the last 4 bytes of the chunk*/ +void lodepng_chunk_generate_crc(unsigned char* chunk); + +/*iterate to next chunks. don't use on IEND chunk, as there is no next chunk then*/ +unsigned char* lodepng_chunk_next(unsigned char* chunk); +const unsigned char* lodepng_chunk_next_const(const unsigned char* chunk); + +/* +Appends chunk to the data in out. The given chunk should already have its chunk header. +The out variable and outlength are updated to reflect the new reallocated buffer. +Returns error code (0 if it went ok) +*/ +unsigned lodepng_chunk_append(unsigned char** out, size_t* outlength, const unsigned char* chunk); + +/* +Appends new chunk to out. The chunk to append is given by giving its length, type +and data separately. The type is a 4-letter string. +The out variable and outlength are updated to reflect the new reallocated buffer. +Returne error code (0 if it went ok) +*/ +unsigned lodepng_chunk_create(unsigned char** out, size_t* outlength, unsigned length, + const char* type, const unsigned char* data); + + +/*Calculate CRC32 of buffer*/ +unsigned lodepng_crc32(const unsigned char* buf, size_t len); +#endif /*LODEPNG_COMPILE_PNG*/ + + +#ifdef LODEPNG_COMPILE_ZLIB +/* +This zlib part can be used independently to zlib compress and decompress a +buffer. It cannot be used to create gzip files however, and it only supports the +part of zlib that is required for PNG, it does not support dictionaries. +*/ + +#ifdef LODEPNG_COMPILE_DECODER +/*Inflate a buffer. Inflate is the decompression step of deflate. Out buffer must be freed after use.*/ +unsigned lodepng_inflate(unsigned char** out, size_t* outsize, + const unsigned char* in, size_t insize, + const LodePNGDecompressSettings* settings); + +/* +Decompresses Zlib data. Reallocates the out buffer and appends the data. The +data must be according to the zlib specification. +Either, *out must be NULL and *outsize must be 0, or, *out must be a valid +buffer and *outsize its size in bytes. out must be freed by user after usage. +*/ +unsigned lodepng_zlib_decompress(unsigned char** out, size_t* outsize, + const unsigned char* in, size_t insize, + const LodePNGDecompressSettings* settings); +#endif /*LODEPNG_COMPILE_DECODER*/ + +#ifdef LODEPNG_COMPILE_ENCODER +/* +Compresses data with Zlib. Reallocates the out buffer and appends the data. +Zlib adds a small header and trailer around the deflate data. +The data is output in the format of the zlib specification. +Either, *out must be NULL and *outsize must be 0, or, *out must be a valid +buffer and *outsize its size in bytes. out must be freed by user after usage. +*/ +unsigned lodepng_zlib_compress(unsigned char** out, size_t* outsize, + const unsigned char* in, size_t insize, + const LodePNGCompressSettings* settings); + +/* +Find length-limited Huffman code for given frequencies. This function is in the +public interface only for tests, it's used internally by lodepng_deflate. +*/ +unsigned lodepng_huffman_code_lengths(unsigned* lengths, const unsigned* frequencies, + size_t numcodes, unsigned maxbitlen); + +/*Compress a buffer with deflate. See RFC 1951. Out buffer must be freed after use.*/ +unsigned lodepng_deflate(unsigned char** out, size_t* outsize, + const unsigned char* in, size_t insize, + const LodePNGCompressSettings* settings); + +#endif /*LODEPNG_COMPILE_ENCODER*/ +#endif /*LODEPNG_COMPILE_ZLIB*/ + +#ifdef LODEPNG_COMPILE_DISK +/* +Load a file from disk into buffer. The function allocates the out buffer, and +after usage you should free it. +out: output parameter, contains pointer to loaded buffer. +outsize: output parameter, size of the allocated out buffer +filename: the path to the file to load +return value: error code (0 means ok) +*/ +unsigned lodepng_load_file(unsigned char** out, size_t* outsize, const char* filename); + +/* +Save a file from buffer to disk. Warning, if it exists, this function overwrites +the file without warning! +buffer: the buffer to write +buffersize: size of the buffer to write +filename: the path to the file to save to +return value: error code (0 means ok) +*/ +unsigned lodepng_save_file(const unsigned char* buffer, size_t buffersize, const char* filename); +#endif /*LODEPNG_COMPILE_DISK*/ + +#ifdef LODEPNG_COMPILE_CPP +/* The LodePNG C++ wrapper uses std::vectors instead of manually allocated memory buffers. */ +namespace lodepng +{ +#ifdef LODEPNG_COMPILE_PNG +class State : public LodePNGState +{ + public: + State(); + State(const State& other); + virtual ~State(); + State& operator=(const State& other); +}; + +#ifdef LODEPNG_COMPILE_DECODER +/* Same as other lodepng::decode, but using a State for more settings and information. */ +unsigned decode(std::vector& out, unsigned& w, unsigned& h, + State& state, + const unsigned char* in, size_t insize); +unsigned decode(std::vector& out, unsigned& w, unsigned& h, + State& state, + const std::vector& in); +#endif /*LODEPNG_COMPILE_DECODER*/ + +#ifdef LODEPNG_COMPILE_ENCODER +/* Same as other lodepng::encode, but using a State for more settings and information. */ +unsigned encode(std::vector& out, + const unsigned char* in, unsigned w, unsigned h, + State& state); +unsigned encode(std::vector& out, + const std::vector& in, unsigned w, unsigned h, + State& state); +#endif /*LODEPNG_COMPILE_ENCODER*/ + +#ifdef LODEPNG_COMPILE_DISK +/* +Load a file from disk into an std::vector. +return value: error code (0 means ok) +*/ +unsigned load_file(std::vector& buffer, const std::string& filename); + +/* +Save the binary data in an std::vector to a file on disk. The file is overwritten +without warning. +*/ +unsigned save_file(const std::vector& buffer, const std::string& filename); +#endif /* LODEPNG_COMPILE_DISK */ +#endif /* LODEPNG_COMPILE_PNG */ + +#ifdef LODEPNG_COMPILE_ZLIB +#ifdef LODEPNG_COMPILE_DECODER +/* Zlib-decompress an unsigned char buffer */ +unsigned decompress(std::vector& out, const unsigned char* in, size_t insize, + const LodePNGDecompressSettings& settings = lodepng_default_decompress_settings); + +/* Zlib-decompress an std::vector */ +unsigned decompress(std::vector& out, const std::vector& in, + const LodePNGDecompressSettings& settings = lodepng_default_decompress_settings); +#endif /* LODEPNG_COMPILE_DECODER */ + +#ifdef LODEPNG_COMPILE_ENCODER +/* Zlib-compress an unsigned char buffer */ +unsigned compress(std::vector& out, const unsigned char* in, size_t insize, + const LodePNGCompressSettings& settings = lodepng_default_compress_settings); + +/* Zlib-compress an std::vector */ +unsigned compress(std::vector& out, const std::vector& in, + const LodePNGCompressSettings& settings = lodepng_default_compress_settings); +#endif /* LODEPNG_COMPILE_ENCODER */ +#endif /* LODEPNG_COMPILE_ZLIB */ +} /* namespace lodepng */ +#endif /*LODEPNG_COMPILE_CPP*/ + +/* +TODO: +[.] test if there are no memory leaks or security exploits - done a lot but needs to be checked often +[.] check compatibility with various compilers - done but needs to be redone for every newer version +[X] converting color to 16-bit per channel types +[ ] read all public PNG chunk types (but never let the color profile and gamma ones touch RGB values) +[ ] make sure encoder generates no chunks with size > (2^31)-1 +[ ] partial decoding (stream processing) +[X] let the "isFullyOpaque" function check color keys and transparent palettes too +[X] better name for the variables "codes", "codesD", "codelengthcodes", "clcl" and "lldl" +[ ] don't stop decoding on errors like 69, 57, 58 (make warnings) +[ ] let the C++ wrapper catch exceptions coming from the standard library and return LodePNG error codes +[ ] allow user to provide custom color conversion functions, e.g. for premultiplied alpha, padding bits or not, ... +[ ] allow user to give data (void*) to custom allocator +*/ + +#endif /*LODEPNG_H inclusion guard*/ + +/* +LodePNG Documentation +--------------------- + +0. table of contents +-------------------- + + 1. about + 1.1. supported features + 1.2. features not supported + 2. C and C++ version + 3. security + 4. decoding + 5. encoding + 6. color conversions + 6.1. PNG color types + 6.2. color conversions + 6.3. padding bits + 6.4. A note about 16-bits per channel and endianness + 7. error values + 8. chunks and PNG editing + 9. compiler support + 10. examples + 10.1. decoder C++ example + 10.2. decoder C example + 11. state settings reference + 12. changes + 13. contact information + + +1. about +-------- + +PNG is a file format to store raster images losslessly with good compression, +supporting different color types and alpha channel. + +LodePNG is a PNG codec according to the Portable Network Graphics (PNG) +Specification (Second Edition) - W3C Recommendation 10 November 2003. + +The specifications used are: + +*) Portable Network Graphics (PNG) Specification (Second Edition): + http://www.w3.org/TR/2003/REC-PNG-20031110 +*) RFC 1950 ZLIB Compressed Data Format version 3.3: + http://www.gzip.org/zlib/rfc-zlib.html +*) RFC 1951 DEFLATE Compressed Data Format Specification ver 1.3: + http://www.gzip.org/zlib/rfc-deflate.html + +The most recent version of LodePNG can currently be found at +http://lodev.org/lodepng/ + +LodePNG works both in C (ISO C90) and C++, with a C++ wrapper that adds +extra functionality. + +LodePNG exists out of two files: +-lodepng.h: the header file for both C and C++ +-lodepng.c(pp): give it the name lodepng.c or lodepng.cpp (or .cc) depending on your usage + +If you want to start using LodePNG right away without reading this doc, get the +examples from the LodePNG website to see how to use it in code, or check the +smaller examples in chapter 13 here. + +LodePNG is simple but only supports the basic requirements. To achieve +simplicity, the following design choices were made: There are no dependencies +on any external library. There are functions to decode and encode a PNG with +a single function call, and extended versions of these functions taking a +LodePNGState struct allowing to specify or get more information. By default +the colors of the raw image are always RGB or RGBA, no matter what color type +the PNG file uses. To read and write files, there are simple functions to +convert the files to/from buffers in memory. + +This all makes LodePNG suitable for loading textures in games, demos and small +programs, ... It's less suitable for full fledged image editors, loading PNGs +over network (it requires all the image data to be available before decoding can +begin), life-critical systems, ... + +1.1. supported features +----------------------- + +The following features are supported by the decoder: + +*) decoding of PNGs with any color type, bit depth and interlace mode, to a 24- or 32-bit color raw image, + or the same color type as the PNG +*) encoding of PNGs, from any raw image to 24- or 32-bit color, or the same color type as the raw image +*) Adam7 interlace and deinterlace for any color type +*) loading the image from harddisk or decoding it from a buffer from other sources than harddisk +*) support for alpha channels, including RGBA color model, translucent palettes and color keying +*) zlib decompression (inflate) +*) zlib compression (deflate) +*) CRC32 and ADLER32 checksums +*) handling of unknown chunks, allowing making a PNG editor that stores custom and unknown chunks. +*) the following chunks are supported (generated/interpreted) by both encoder and decoder: + IHDR: header information + PLTE: color palette + IDAT: pixel data + IEND: the final chunk + tRNS: transparency for palettized images + tEXt: textual information + zTXt: compressed textual information + iTXt: international textual information + bKGD: suggested background color + pHYs: physical dimensions + tIME: modification time + +1.2. features not supported +--------------------------- + +The following features are _not_ supported: + +*) some features needed to make a conformant PNG-Editor might be still missing. +*) partial loading/stream processing. All data must be available and is processed in one call. +*) The following public chunks are not supported but treated as unknown chunks by LodePNG + cHRM, gAMA, iCCP, sRGB, sBIT, hIST, sPLT + Some of these are not supported on purpose: LodePNG wants to provide the RGB values + stored in the pixels, not values modified by system dependent gamma or color models. + + +2. C and C++ version +-------------------- + +The C version uses buffers allocated with alloc that you need to free() +yourself. You need to use init and cleanup functions for each struct whenever +using a struct from the C version to avoid exploits and memory leaks. + +The C++ version has extra functions with std::vectors in the interface and the +lodepng::State class which is a LodePNGState with constructor and destructor. + +These files work without modification for both C and C++ compilers because all +the additional C++ code is in "#ifdef __cplusplus" blocks that make C-compilers +ignore it, and the C code is made to compile both with strict ISO C90 and C++. + +To use the C++ version, you need to rename the source file to lodepng.cpp +(instead of lodepng.c), and compile it with a C++ compiler. + +To use the C version, you need to rename the source file to lodepng.c (instead +of lodepng.cpp), and compile it with a C compiler. + + +3. Security +----------- + +Even if carefully designed, it's always possible that LodePNG contains possible +exploits. If you discover one, please let me know, and it will be fixed. + +When using LodePNG, care has to be taken with the C version of LodePNG, as well +as the C-style structs when working with C++. The following conventions are used +for all C-style structs: + +-if a struct has a corresponding init function, always call the init function when making a new one +-if a struct has a corresponding cleanup function, call it before the struct disappears to avoid memory leaks +-if a struct has a corresponding copy function, use the copy function instead of "=". + The destination must also be inited already. + + +4. Decoding +----------- + +Decoding converts a PNG compressed image to a raw pixel buffer. + +Most documentation on using the decoder is at its declarations in the header +above. For C, simple decoding can be done with functions such as +lodepng_decode32, and more advanced decoding can be done with the struct +LodePNGState and lodepng_decode. For C++, all decoding can be done with the +various lodepng::decode functions, and lodepng::State can be used for advanced +features. + +When using the LodePNGState, it uses the following fields for decoding: +*) LodePNGInfo info_png: it stores extra information about the PNG (the input) in here +*) LodePNGColorMode info_raw: here you can say what color mode of the raw image (the output) you want to get +*) LodePNGDecoderSettings decoder: you can specify a few extra settings for the decoder to use + +LodePNGInfo info_png +-------------------- + +After decoding, this contains extra information of the PNG image, except the actual +pixels, width and height because these are already gotten directly from the decoder +functions. + +It contains for example the original color type of the PNG image, text comments, +suggested background color, etc... More details about the LodePNGInfo struct are +at its declaration documentation. + +LodePNGColorMode info_raw +------------------------- + +When decoding, here you can specify which color type you want +the resulting raw image to be. If this is different from the colortype of the +PNG, then the decoder will automatically convert the result. This conversion +always works, except if you want it to convert a color PNG to greyscale or to +a palette with missing colors. + +By default, 32-bit color is used for the result. + +LodePNGDecoderSettings decoder +------------------------------ + +The settings can be used to ignore the errors created by invalid CRC and Adler32 +chunks, and to disable the decoding of tEXt chunks. + +There's also a setting color_convert, true by default. If false, no conversion +is done, the resulting data will be as it was in the PNG (after decompression) +and you'll have to puzzle the colors of the pixels together yourself using the +color type information in the LodePNGInfo. + + +5. Encoding +----------- + +Encoding converts a raw pixel buffer to a PNG compressed image. + +Most documentation on using the encoder is at its declarations in the header +above. For C, simple encoding can be done with functions such as +lodepng_encode32, and more advanced decoding can be done with the struct +LodePNGState and lodepng_encode. For C++, all encoding can be done with the +various lodepng::encode functions, and lodepng::State can be used for advanced +features. + +Like the decoder, the encoder can also give errors. However it gives less errors +since the encoder input is trusted, the decoder input (a PNG image that could +be forged by anyone) is not trusted. + +When using the LodePNGState, it uses the following fields for encoding: +*) LodePNGInfo info_png: here you specify how you want the PNG (the output) to be. +*) LodePNGColorMode info_raw: here you say what color type of the raw image (the input) has +*) LodePNGEncoderSettings encoder: you can specify a few settings for the encoder to use + +LodePNGInfo info_png +-------------------- + +When encoding, you use this the opposite way as when decoding: for encoding, +you fill in the values you want the PNG to have before encoding. By default it's +not needed to specify a color type for the PNG since it's automatically chosen, +but it's possible to choose it yourself given the right settings. + +The encoder will not always exactly match the LodePNGInfo struct you give, +it tries as close as possible. Some things are ignored by the encoder. The +encoder uses, for example, the following settings from it when applicable: +colortype and bitdepth, text chunks, time chunk, the color key, the palette, the +background color, the interlace method, unknown chunks, ... + +When encoding to a PNG with colortype 3, the encoder will generate a PLTE chunk. +If the palette contains any colors for which the alpha channel is not 255 (so +there are translucent colors in the palette), it'll add a tRNS chunk. + +LodePNGColorMode info_raw +------------------------- + +You specify the color type of the raw image that you give to the input here, +including a possible transparent color key and palette you happen to be using in +your raw image data. + +By default, 32-bit color is assumed, meaning your input has to be in RGBA +format with 4 bytes (unsigned chars) per pixel. + +LodePNGEncoderSettings encoder +------------------------------ + +The following settings are supported (some are in sub-structs): +*) auto_convert: when this option is enabled, the encoder will +automatically choose the smallest possible color mode (including color key) that +can encode the colors of all pixels without information loss. +*) btype: the block type for LZ77. 0 = uncompressed, 1 = fixed huffman tree, + 2 = dynamic huffman tree (best compression). Should be 2 for proper + compression. +*) use_lz77: whether or not to use LZ77 for compressed block types. Should be + true for proper compression. +*) windowsize: the window size used by the LZ77 encoder (1 - 32768). Has value + 2048 by default, but can be set to 32768 for better, but slow, compression. +*) force_palette: if colortype is 2 or 6, you can make the encoder write a PLTE + chunk if force_palette is true. This can used as suggested palette to convert + to by viewers that don't support more than 256 colors (if those still exist) +*) add_id: add text chunk "Encoder: LodePNG " to the image. +*) text_compression: default 1. If 1, it'll store texts as zTXt instead of tEXt chunks. + zTXt chunks use zlib compression on the text. This gives a smaller result on + large texts but a larger result on small texts (such as a single program name). + It's all tEXt or all zTXt though, there's no separate setting per text yet. + + +6. color conversions +-------------------- + +An important thing to note about LodePNG, is that the color type of the PNG, and +the color type of the raw image, are completely independent. By default, when +you decode a PNG, you get the result as a raw image in the color type you want, +no matter whether the PNG was encoded with a palette, greyscale or RGBA color. +And if you encode an image, by default LodePNG will automatically choose the PNG +color type that gives good compression based on the values of colors and amount +of colors in the image. It can be configured to let you control it instead as +well, though. + +To be able to do this, LodePNG does conversions from one color mode to another. +It can convert from almost any color type to any other color type, except the +following conversions: RGB to greyscale is not supported, and converting to a +palette when the palette doesn't have a required color is not supported. This is +not supported on purpose: this is information loss which requires a color +reduction algorithm that is beyong the scope of a PNG encoder (yes, RGB to grey +is easy, but there are multiple ways if you want to give some channels more +weight). + +By default, when decoding, you get the raw image in 32-bit RGBA or 24-bit RGB +color, no matter what color type the PNG has. And by default when encoding, +LodePNG automatically picks the best color model for the output PNG, and expects +the input image to be 32-bit RGBA or 24-bit RGB. So, unless you want to control +the color format of the images yourself, you can skip this chapter. + +6.1. PNG color types +-------------------- + +A PNG image can have many color types, ranging from 1-bit color to 64-bit color, +as well as palettized color modes. After the zlib decompression and unfiltering +in the PNG image is done, the raw pixel data will have that color type and thus +a certain amount of bits per pixel. If you want the output raw image after +decoding to have another color type, a conversion is done by LodePNG. + +The PNG specification gives the following color types: + +0: greyscale, bit depths 1, 2, 4, 8, 16 +2: RGB, bit depths 8 and 16 +3: palette, bit depths 1, 2, 4 and 8 +4: greyscale with alpha, bit depths 8 and 16 +6: RGBA, bit depths 8 and 16 + +Bit depth is the amount of bits per pixel per color channel. So the total amount +of bits per pixel is: amount of channels * bitdepth. + +6.2. color conversions +---------------------- + +As explained in the sections about the encoder and decoder, you can specify +color types and bit depths in info_png and info_raw to change the default +behaviour. + +If, when decoding, you want the raw image to be something else than the default, +you need to set the color type and bit depth you want in the LodePNGColorMode, +or the parameters colortype and bitdepth of the simple decoding function. + +If, when encoding, you use another color type than the default in the raw input +image, you need to specify its color type and bit depth in the LodePNGColorMode +of the raw image, or use the parameters colortype and bitdepth of the simple +encoding function. + +If, when encoding, you don't want LodePNG to choose the output PNG color type +but control it yourself, you need to set auto_convert in the encoder settings +to false, and specify the color type you want in the LodePNGInfo of the +encoder (including palette: it can generate a palette if auto_convert is true, +otherwise not). + +If the input and output color type differ (whether user chosen or auto chosen), +LodePNG will do a color conversion, which follows the rules below, and may +sometimes result in an error. + +To avoid some confusion: +-the decoder converts from PNG to raw image +-the encoder converts from raw image to PNG +-the colortype and bitdepth in LodePNGColorMode info_raw, are those of the raw image +-the colortype and bitdepth in the color field of LodePNGInfo info_png, are those of the PNG +-when encoding, the color type in LodePNGInfo is ignored if auto_convert + is enabled, it is automatically generated instead +-when decoding, the color type in LodePNGInfo is set by the decoder to that of the original + PNG image, but it can be ignored since the raw image has the color type you requested instead +-if the color type of the LodePNGColorMode and PNG image aren't the same, a conversion + between the color types is done if the color types are supported. If it is not + supported, an error is returned. If the types are the same, no conversion is done. +-even though some conversions aren't supported, LodePNG supports loading PNGs from any + colortype and saving PNGs to any colortype, sometimes it just requires preparing + the raw image correctly before encoding. +-both encoder and decoder use the same color converter. + +Non supported color conversions: +-color to greyscale: no error is thrown, but the result will look ugly because +only the red channel is taken +-anything to palette when that palette does not have that color in it: in this +case an error is thrown + +Supported color conversions: +-anything to 8-bit RGB, 8-bit RGBA, 16-bit RGB, 16-bit RGBA +-any grey or grey+alpha, to grey or grey+alpha +-anything to a palette, as long as the palette has the requested colors in it +-removing alpha channel +-higher to smaller bitdepth, and vice versa + +If you want no color conversion to be done (e.g. for speed or control): +-In the encoder, you can make it save a PNG with any color type by giving the +raw color mode and LodePNGInfo the same color mode, and setting auto_convert to +false. +-In the decoder, you can make it store the pixel data in the same color type +as the PNG has, by setting the color_convert setting to false. Settings in +info_raw are then ignored. + +The function lodepng_convert does the color conversion. It is available in the +interface but normally isn't needed since the encoder and decoder already call +it. + +6.3. padding bits +----------------- + +In the PNG file format, if a less than 8-bit per pixel color type is used and the scanlines +have a bit amount that isn't a multiple of 8, then padding bits are used so that each +scanline starts at a fresh byte. But that is NOT true for the LodePNG raw input and output. +The raw input image you give to the encoder, and the raw output image you get from the decoder +will NOT have these padding bits, e.g. in the case of a 1-bit image with a width +of 7 pixels, the first pixel of the second scanline will the the 8th bit of the first byte, +not the first bit of a new byte. + +6.4. A note about 16-bits per channel and endianness +---------------------------------------------------- + +LodePNG uses unsigned char arrays for 16-bit per channel colors too, just like +for any other color format. The 16-bit values are stored in big endian (most +significant byte first) in these arrays. This is the opposite order of the +little endian used by x86 CPU's. + +LodePNG always uses big endian because the PNG file format does so internally. +Conversions to other formats than PNG uses internally are not supported by +LodePNG on purpose, there are myriads of formats, including endianness of 16-bit +colors, the order in which you store R, G, B and A, and so on. Supporting and +converting to/from all that is outside the scope of LodePNG. + +This may mean that, depending on your use case, you may want to convert the big +endian output of LodePNG to little endian with a for loop. This is certainly not +always needed, many applications and libraries support big endian 16-bit colors +anyway, but it means you cannot simply cast the unsigned char* buffer to an +unsigned short* buffer on x86 CPUs. + + +7. error values +--------------- + +All functions in LodePNG that return an error code, return 0 if everything went +OK, or a non-zero code if there was an error. + +The meaning of the LodePNG error values can be retrieved with the function +lodepng_error_text: given the numerical error code, it returns a description +of the error in English as a string. + +Check the implementation of lodepng_error_text to see the meaning of each code. + + +8. chunks and PNG editing +------------------------- + +If you want to add extra chunks to a PNG you encode, or use LodePNG for a PNG +editor that should follow the rules about handling of unknown chunks, or if your +program is able to read other types of chunks than the ones handled by LodePNG, +then that's possible with the chunk functions of LodePNG. + +A PNG chunk has the following layout: + +4 bytes length +4 bytes type name +length bytes data +4 bytes CRC + +8.1. iterating through chunks +----------------------------- + +If you have a buffer containing the PNG image data, then the first chunk (the +IHDR chunk) starts at byte number 8 of that buffer. The first 8 bytes are the +signature of the PNG and are not part of a chunk. But if you start at byte 8 +then you have a chunk, and can check the following things of it. + +NOTE: none of these functions check for memory buffer boundaries. To avoid +exploits, always make sure the buffer contains all the data of the chunks. +When using lodepng_chunk_next, make sure the returned value is within the +allocated memory. + +unsigned lodepng_chunk_length(const unsigned char* chunk): + +Get the length of the chunk's data. The total chunk length is this length + 12. + +void lodepng_chunk_type(char type[5], const unsigned char* chunk): +unsigned char lodepng_chunk_type_equals(const unsigned char* chunk, const char* type): + +Get the type of the chunk or compare if it's a certain type + +unsigned char lodepng_chunk_critical(const unsigned char* chunk): +unsigned char lodepng_chunk_private(const unsigned char* chunk): +unsigned char lodepng_chunk_safetocopy(const unsigned char* chunk): + +Check if the chunk is critical in the PNG standard (only IHDR, PLTE, IDAT and IEND are). +Check if the chunk is private (public chunks are part of the standard, private ones not). +Check if the chunk is safe to copy. If it's not, then, when modifying data in a critical +chunk, unsafe to copy chunks of the old image may NOT be saved in the new one if your +program doesn't handle that type of unknown chunk. + +unsigned char* lodepng_chunk_data(unsigned char* chunk): +const unsigned char* lodepng_chunk_data_const(const unsigned char* chunk): + +Get a pointer to the start of the data of the chunk. + +unsigned lodepng_chunk_check_crc(const unsigned char* chunk): +void lodepng_chunk_generate_crc(unsigned char* chunk): + +Check if the crc is correct or generate a correct one. + +unsigned char* lodepng_chunk_next(unsigned char* chunk): +const unsigned char* lodepng_chunk_next_const(const unsigned char* chunk): + +Iterate to the next chunk. This works if you have a buffer with consecutive chunks. Note that these +functions do no boundary checking of the allocated data whatsoever, so make sure there is enough +data available in the buffer to be able to go to the next chunk. + +unsigned lodepng_chunk_append(unsigned char** out, size_t* outlength, const unsigned char* chunk): +unsigned lodepng_chunk_create(unsigned char** out, size_t* outlength, unsigned length, + const char* type, const unsigned char* data): + +These functions are used to create new chunks that are appended to the data in *out that has +length *outlength. The append function appends an existing chunk to the new data. The create +function creates a new chunk with the given parameters and appends it. Type is the 4-letter +name of the chunk. + +8.2. chunks in info_png +----------------------- + +The LodePNGInfo struct contains fields with the unknown chunk in it. It has 3 +buffers (each with size) to contain 3 types of unknown chunks: +the ones that come before the PLTE chunk, the ones that come between the PLTE +and the IDAT chunks, and the ones that come after the IDAT chunks. +It's necessary to make the distionction between these 3 cases because the PNG +standard forces to keep the ordering of unknown chunks compared to the critical +chunks, but does not force any other ordering rules. + +info_png.unknown_chunks_data[0] is the chunks before PLTE +info_png.unknown_chunks_data[1] is the chunks after PLTE, before IDAT +info_png.unknown_chunks_data[2] is the chunks after IDAT + +The chunks in these 3 buffers can be iterated through and read by using the same +way described in the previous subchapter. + +When using the decoder to decode a PNG, you can make it store all unknown chunks +if you set the option settings.remember_unknown_chunks to 1. By default, this +option is off (0). + +The encoder will always encode unknown chunks that are stored in the info_png. +If you need it to add a particular chunk that isn't known by LodePNG, you can +use lodepng_chunk_append or lodepng_chunk_create to the chunk data in +info_png.unknown_chunks_data[x]. + +Chunks that are known by LodePNG should not be added in that way. E.g. to make +LodePNG add a bKGD chunk, set background_defined to true and add the correct +parameters there instead. + + +9. compiler support +------------------- + +No libraries other than the current standard C library are needed to compile +LodePNG. For the C++ version, only the standard C++ library is needed on top. +Add the files lodepng.c(pp) and lodepng.h to your project, include +lodepng.h where needed, and your program can read/write PNG files. + +It is compatible with C90 and up, and C++03 and up. + +If performance is important, use optimization when compiling! For both the +encoder and decoder, this makes a large difference. + +Make sure that LodePNG is compiled with the same compiler of the same version +and with the same settings as the rest of the program, or the interfaces with +std::vectors and std::strings in C++ can be incompatible. + +CHAR_BITS must be 8 or higher, because LodePNG uses unsigned chars for octets. + +*) gcc and g++ + +LodePNG is developed in gcc so this compiler is natively supported. It gives no +warnings with compiler options "-Wall -Wextra -pedantic -ansi", with gcc and g++ +version 4.7.1 on Linux, 32-bit and 64-bit. + +*) Clang + +Fully supported and warning-free. + +*) Mingw + +The Mingw compiler (a port of gcc for Windows) should be fully supported by +LodePNG. + +*) Visual Studio and Visual C++ Express Edition + +LodePNG should be warning-free with warning level W4. Two warnings were disabled +with pragmas though: warning 4244 about implicit conversions, and warning 4996 +where it wants to use a non-standard function fopen_s instead of the standard C +fopen. + +Visual Studio may want "stdafx.h" files to be included in each source file and +give an error "unexpected end of file while looking for precompiled header". +This is not standard C++ and will not be added to the stock LodePNG. You can +disable it for lodepng.cpp only by right clicking it, Properties, C/C++, +Precompiled Headers, and set it to Not Using Precompiled Headers there. + +NOTE: Modern versions of VS should be fully supported, but old versions, e.g. +VS6, are not guaranteed to work. + +*) Compilers on Macintosh + +LodePNG has been reported to work both with gcc and LLVM for Macintosh, both for +C and C++. + +*) Other Compilers + +If you encounter problems on any compilers, feel free to let me know and I may +try to fix it if the compiler is modern and standards complient. + + +10. examples +------------ + +This decoder example shows the most basic usage of LodePNG. More complex +examples can be found on the LodePNG website. + +10.1. decoder C++ example +------------------------- + +#include "lodepng.h" +#include + +int main(int argc, char *argv[]) +{ + const char* filename = argc > 1 ? argv[1] : "test.png"; + + //load and decode + std::vector image; + unsigned width, height; + unsigned error = lodepng::decode(image, width, height, filename); + + //if there's an error, display it + if(error) std::cout << "decoder error " << error << ": " << lodepng_error_text(error) << std::endl; + + //the pixels are now in the vector "image", 4 bytes per pixel, ordered RGBARGBA..., use it as texture, draw it, ... +} + +10.2. decoder C example +----------------------- + +#include "lodepng.h" + +int main(int argc, char *argv[]) +{ + unsigned error; + unsigned char* image; + size_t width, height; + const char* filename = argc > 1 ? argv[1] : "test.png"; + + error = lodepng_decode32_file(&image, &width, &height, filename); + + if(error) printf("decoder error %u: %s\n", error, lodepng_error_text(error)); + + / * use image here * / + + free(image); + return 0; +} + +11. state settings reference +---------------------------- + +A quick reference of some settings to set on the LodePNGState + +For decoding: + +state.decoder.zlibsettings.ignore_adler32: ignore ADLER32 checksums +state.decoder.zlibsettings.custom_...: use custom inflate function +state.decoder.ignore_crc: ignore CRC checksums +state.decoder.color_convert: convert internal PNG color to chosen one +state.decoder.read_text_chunks: whether to read in text metadata chunks +state.decoder.remember_unknown_chunks: whether to read in unknown chunks +state.info_raw.colortype: desired color type for decoded image +state.info_raw.bitdepth: desired bit depth for decoded image +state.info_raw....: more color settings, see struct LodePNGColorMode +state.info_png....: no settings for decoder but ouput, see struct LodePNGInfo + +For encoding: + +state.encoder.zlibsettings.btype: disable compression by setting it to 0 +state.encoder.zlibsettings.use_lz77: use LZ77 in compression +state.encoder.zlibsettings.windowsize: tweak LZ77 windowsize +state.encoder.zlibsettings.minmatch: tweak min LZ77 length to match +state.encoder.zlibsettings.nicematch: tweak LZ77 match where to stop searching +state.encoder.zlibsettings.lazymatching: try one more LZ77 matching +state.encoder.zlibsettings.custom_...: use custom deflate function +state.encoder.auto_convert: choose optimal PNG color type, if 0 uses info_png +state.encoder.filter_palette_zero: PNG filter strategy for palette +state.encoder.filter_strategy: PNG filter strategy to encode with +state.encoder.force_palette: add palette even if not encoding to one +state.encoder.add_id: add LodePNG identifier and version as a text chunk +state.encoder.text_compression: use compressed text chunks for metadata +state.info_raw.colortype: color type of raw input image you provide +state.info_raw.bitdepth: bit depth of raw input image you provide +state.info_raw: more color settings, see struct LodePNGColorMode +state.info_png.color.colortype: desired color type if auto_convert is false +state.info_png.color.bitdepth: desired bit depth if auto_convert is false +state.info_png.color....: more color settings, see struct LodePNGColorMode +state.info_png....: more PNG related settings, see struct LodePNGInfo + + +12. changes +----------- + +The version number of LodePNG is the date of the change given in the format +yyyymmdd. + +Some changes aren't backwards compatible. Those are indicated with a (!) +symbol. + +*) 08 dec 2015: Made load_file function return error if file can't be opened. +*) 24 okt 2015: Bugfix with decoding to palette output. +*) 18 apr 2015: Boundary PM instead of just package-merge for faster encoding. +*) 23 aug 2014: Reduced needless memory usage of decoder. +*) 28 jun 2014: Removed fix_png setting, always support palette OOB for + simplicity. Made ColorProfile public. +*) 09 jun 2014: Faster encoder by fixing hash bug and more zeros optimization. +*) 22 dec 2013: Power of two windowsize required for optimization. +*) 15 apr 2013: Fixed bug with LAC_ALPHA and color key. +*) 25 mar 2013: Added an optional feature to ignore some PNG errors (fix_png). +*) 11 mar 2013 (!): Bugfix with custom free. Changed from "my" to "lodepng_" + prefix for the custom allocators and made it possible with a new #define to + use custom ones in your project without needing to change lodepng's code. +*) 28 jan 2013: Bugfix with color key. +*) 27 okt 2012: Tweaks in text chunk keyword length error handling. +*) 8 okt 2012 (!): Added new filter strategy (entropy) and new auto color mode. + (no palette). Better deflate tree encoding. New compression tweak settings. + Faster color conversions while decoding. Some internal cleanups. +*) 23 sep 2012: Reduced warnings in Visual Studio a little bit. +*) 1 sep 2012 (!): Removed #define's for giving custom (de)compression functions + and made it work with function pointers instead. +*) 23 jun 2012: Added more filter strategies. Made it easier to use custom alloc + and free functions and toggle #defines from compiler flags. Small fixes. +*) 6 may 2012 (!): Made plugging in custom zlib/deflate functions more flexible. +*) 22 apr 2012 (!): Made interface more consistent, renaming a lot. Removed + redundant C++ codec classes. Reduced amount of structs. Everything changed, + but it is cleaner now imho and functionality remains the same. Also fixed + several bugs and shrunk the implementation code. Made new samples. +*) 6 nov 2011 (!): By default, the encoder now automatically chooses the best + PNG color model and bit depth, based on the amount and type of colors of the + raw image. For this, autoLeaveOutAlphaChannel replaced by auto_choose_color. +*) 9 okt 2011: simpler hash chain implementation for the encoder. +*) 8 sep 2011: lz77 encoder lazy matching instead of greedy matching. +*) 23 aug 2011: tweaked the zlib compression parameters after benchmarking. + A bug with the PNG filtertype heuristic was fixed, so that it chooses much + better ones (it's quite significant). A setting to do an experimental, slow, + brute force search for PNG filter types is added. +*) 17 aug 2011 (!): changed some C zlib related function names. +*) 16 aug 2011: made the code less wide (max 120 characters per line). +*) 17 apr 2011: code cleanup. Bugfixes. Convert low to 16-bit per sample colors. +*) 21 feb 2011: fixed compiling for C90. Fixed compiling with sections disabled. +*) 11 dec 2010: encoding is made faster, based on suggestion by Peter Eastman + to optimize long sequences of zeros. +*) 13 nov 2010: added LodePNG_InfoColor_hasPaletteAlpha and + LodePNG_InfoColor_canHaveAlpha functions for convenience. +*) 7 nov 2010: added LodePNG_error_text function to get error code description. +*) 30 okt 2010: made decoding slightly faster +*) 26 okt 2010: (!) changed some C function and struct names (more consistent). + Reorganized the documentation and the declaration order in the header. +*) 08 aug 2010: only changed some comments and external samples. +*) 05 jul 2010: fixed bug thanks to warnings in the new gcc version. +*) 14 mar 2010: fixed bug where too much memory was allocated for char buffers. +*) 02 sep 2008: fixed bug where it could create empty tree that linux apps could + read by ignoring the problem but windows apps couldn't. +*) 06 jun 2008: added more error checks for out of memory cases. +*) 26 apr 2008: added a few more checks here and there to ensure more safety. +*) 06 mar 2008: crash with encoding of strings fixed +*) 02 feb 2008: support for international text chunks added (iTXt) +*) 23 jan 2008: small cleanups, and #defines to divide code in sections +*) 20 jan 2008: support for unknown chunks allowing using LodePNG for an editor. +*) 18 jan 2008: support for tIME and pHYs chunks added to encoder and decoder. +*) 17 jan 2008: ability to encode and decode compressed zTXt chunks added + Also various fixes, such as in the deflate and the padding bits code. +*) 13 jan 2008: Added ability to encode Adam7-interlaced images. Improved + filtering code of encoder. +*) 07 jan 2008: (!) changed LodePNG to use ISO C90 instead of C++. A + C++ wrapper around this provides an interface almost identical to before. + Having LodePNG be pure ISO C90 makes it more portable. The C and C++ code + are together in these files but it works both for C and C++ compilers. +*) 29 dec 2007: (!) changed most integer types to unsigned int + other tweaks +*) 30 aug 2007: bug fixed which makes this Borland C++ compatible +*) 09 aug 2007: some VS2005 warnings removed again +*) 21 jul 2007: deflate code placed in new namespace separate from zlib code +*) 08 jun 2007: fixed bug with 2- and 4-bit color, and small interlaced images +*) 04 jun 2007: improved support for Visual Studio 2005: crash with accessing + invalid std::vector element [0] fixed, and level 3 and 4 warnings removed +*) 02 jun 2007: made the encoder add a tag with version by default +*) 27 may 2007: zlib and png code separated (but still in the same file), + simple encoder/decoder functions added for more simple usage cases +*) 19 may 2007: minor fixes, some code cleaning, new error added (error 69), + moved some examples from here to lodepng_examples.cpp +*) 12 may 2007: palette decoding bug fixed +*) 24 apr 2007: changed the license from BSD to the zlib license +*) 11 mar 2007: very simple addition: ability to encode bKGD chunks. +*) 04 mar 2007: (!) tEXt chunk related fixes, and support for encoding + palettized PNG images. Plus little interface change with palette and texts. +*) 03 mar 2007: Made it encode dynamic Huffman shorter with repeat codes. + Fixed a bug where the end code of a block had length 0 in the Huffman tree. +*) 26 feb 2007: Huffman compression with dynamic trees (BTYPE 2) now implemented + and supported by the encoder, resulting in smaller PNGs at the output. +*) 27 jan 2007: Made the Adler-32 test faster so that a timewaste is gone. +*) 24 jan 2007: gave encoder an error interface. Added color conversion from any + greyscale type to 8-bit greyscale with or without alpha. +*) 21 jan 2007: (!) Totally changed the interface. It allows more color types + to convert to and is more uniform. See the manual for how it works now. +*) 07 jan 2007: Some cleanup & fixes, and a few changes over the last days: + encode/decode custom tEXt chunks, separate classes for zlib & deflate, and + at last made the decoder give errors for incorrect Adler32 or Crc. +*) 01 jan 2007: Fixed bug with encoding PNGs with less than 8 bits per channel. +*) 29 dec 2006: Added support for encoding images without alpha channel, and + cleaned out code as well as making certain parts faster. +*) 28 dec 2006: Added "Settings" to the encoder. +*) 26 dec 2006: The encoder now does LZ77 encoding and produces much smaller files now. + Removed some code duplication in the decoder. Fixed little bug in an example. +*) 09 dec 2006: (!) Placed output parameters of public functions as first parameter. + Fixed a bug of the decoder with 16-bit per color. +*) 15 okt 2006: Changed documentation structure +*) 09 okt 2006: Encoder class added. It encodes a valid PNG image from the + given image buffer, however for now it's not compressed. +*) 08 sep 2006: (!) Changed to interface with a Decoder class +*) 30 jul 2006: (!) LodePNG_InfoPng , width and height are now retrieved in different + way. Renamed decodePNG to decodePNGGeneric. +*) 29 jul 2006: (!) Changed the interface: image info is now returned as a + struct of type LodePNG::LodePNG_Info, instead of a vector, which was a bit clumsy. +*) 28 jul 2006: Cleaned the code and added new error checks. + Corrected terminology "deflate" into "inflate". +*) 23 jun 2006: Added SDL example in the documentation in the header, this + example allows easy debugging by displaying the PNG and its transparency. +*) 22 jun 2006: (!) Changed way to obtain error value. Added + loadFile function for convenience. Made decodePNG32 faster. +*) 21 jun 2006: (!) Changed type of info vector to unsigned. + Changed position of palette in info vector. Fixed an important bug that + happened on PNGs with an uncompressed block. +*) 16 jun 2006: Internally changed unsigned into unsigned where + needed, and performed some optimizations. +*) 07 jun 2006: (!) Renamed functions to decodePNG and placed them + in LodePNG namespace. Changed the order of the parameters. Rewrote the + documentation in the header. Renamed files to lodepng.cpp and lodepng.h +*) 22 apr 2006: Optimized and improved some code +*) 07 sep 2005: (!) Changed to std::vector interface +*) 12 aug 2005: Initial release (C++, decoder only) + + +13. contact information +----------------------- + +Feel free to contact me with suggestions, problems, comments, ... concerning +LodePNG. If you encounter a PNG image that doesn't work properly with this +decoder, feel free to send it and I'll use it to find and fix the problem. + +My email address is (puzzle the account and domain together with an @ symbol): +Domain: gmail dot com. +Account: lode dot vandevenne. + + +Copyright (c) 2005-2016 Lode Vandevenne +*/