-
Notifications
You must be signed in to change notification settings - Fork 978
New issue
Have a question about this project? Sign up for a free GitHub account to open an issue and contact its maintainers and the community.
By clicking “Sign up for GitHub”, you agree to our terms of service and privacy statement. We’ll occasionally send you account related emails.
Already on GitHub? Sign in to your account
Add CUDA & OpenCL support #227
base: master
Are you sure you want to change the base?
Conversation
cuDiffmapOpsinDynamicsImage cuComputeBlockZeroingOrder cuMask
继续简化代码
目前速度比opencl略差,待分析优化
肯定是opencl生成代码时有bug!
Conflicts: guetzli/butteraugli_comparator.cc third_party/butteraugli/butteraugli/butteraugli.h
#ifdef __cplusplus | ||
#ifndef __CUDACC__ | ||
#include "CL/cl.h" | ||
#include "cuda.h" |
There was a problem hiding this comment.
Choose a reason for hiding this comment
The reason will be displayed to describe this comment to others. Learn more.
I'm having trouble building OpenCL only with the Intel SDK, this seems to be part of it.
size_t len, size_t offset, | ||
const float* __restrict__ multipliers, | ||
const float* __restrict__ inp, | ||
double border_ratio, |
There was a problem hiding this comment.
Choose a reason for hiding this comment
The reason will be displayed to describe this comment to others. Learn more.
float
here, otherwise linking error.
There was a problem hiding this comment.
Choose a reason for hiding this comment
The reason will be displayed to describe this comment to others. Learn more.
Try OpenCL 2.0 or larger version in OpenCL Code Builder setting.
Any news on this? |
``` | ||
You can pass a `--c` parameter to enable the procedure optimization or `--cuda` parameter to use the CUDA acceleration or `--opencl` to use the OpenCL acceleration. | ||
|
||
If you have any question about CUDA/OpenCL support, please contact strongtu@tencent.com, ianhuang@tencent.com or chriskzhou@tencent.com. |
There was a problem hiding this comment.
Choose a reason for hiding this comment
The reason will be displayed to describe this comment to others. Learn more.
Maybe create a dropbox email like guetzli-support@tencent.com that would go to all 3 of you (and could be adjusted on your end to add/remove people as necessary without having to update these docs)
Could someone of you provide a working binary file for windows with cuda support? This would be awesome. |
Will this require non-free CUDA (cuda.h) libraries for compilation? If yes, probably a compilation flag is needed to disable them... |
where is cu_mem defined? cuda or opencl can not find that. |
@leafjungle It's defined in clguetzli.cl.h, not an original CUDA definition. |
which GPU was used in your environment? |
Waiting impatiently for this to be merged )) |
Yes, we all do! |
@joyjoker2017 Tesla M40 |
🙏 👍 Fingers crossed 🤞 😄 |
Bump... |
😴 |
Any news on this? |
Is this ever going to be merged ? :) |
Any update on this? Please :) |
Did anyone ever succeed in building those binaries and want to share them? ;-) |
This looks incredible! @crazyks could you resolve the conflicts, update the dependencies and rebase? I will check with someone from Google if we can merge into master. If not, we can create a fork. |
Hi there. I have built Guetzli with CUDA for Windows. You can download binaries from here. |
@crazyks could you resolve the conflict? |
Guetzli is an awesome jpeg encoder, however, it works a liitle bit slow. In order to speed it up, we have added CUDA & OpenCL support for Guetzli and we also optimized some procedure and added full jpeg format support.
This work is made by strongtu, ianhuang, tongzhan and me.
We had it tested on our GPU server, and here is the test statistical data for one of the sample pictures.
We hope that our codes can be merged into guetzli/master branch:)