Skip to content
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.

Sign up for GitHub

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

Fix x86 compilation #376

Merged
merged 8 commits into from
Jul 4, 2024
Merged

Fix x86 compilation #376

Show file tree
Hide file tree
Changes from 2 commits
Commits
Show all changes
8 commits
Select commit Hold shift + click to select a range
6e1aea8
Fix x86 standard library imports
PigeonF Jun 26, 2024
07cc010
Add fallback implementation of `compute_quote_mask` for x86
PigeonF Jun 26, 2024
9097323
Fix unused field warnings for nightly compiler
PigeonF Jun 26, 2024
e93edd1
Fix unexpected config error on nightly
PigeonF Jun 26, 2024
c2a0d92
Fix `clippy::doc_markdown` lints
PigeonF Jun 26, 2024
0cc31fe
Fix `clippy::unconditional_recursion` lints
PigeonF Jun 26, 2024
67bbe6a
Fix `clippy::multiple_bound_locations` lints
PigeonF Jun 26, 2024
0351a7b
Fix nightly `clippy::legacy_numeric_constants` lints
PigeonF Jun 26, 2024
File filter

Filter by extension

Filter by extension
Conversations
Failed to load comments.
Loading
Jump to
Jump to file
Failed to load files.
Loading
Diff view
Diff view
24 changes: 8 additions & 16 deletions src/impls/avx2/deser.rs
View file
Open in desktop
Original file line number Diff line number Diff line change
@@ -1,10 +1,10 @@
#[cfg(target_arch = "x86")]
use std::arch::x86::{
__m256i, _mm256_cmpeq_epi8, _mm256_loadu_si256, _mm256_movemask_epi8, _mm256_set1_epi8,
_mm256_storeu_si256,
};
use std::arch::x86 as arch;

#[cfg(target_arch = "x86_64")]
use std::arch::x86_64::{
use std::arch::x86_64 as arch;

use arch::{
__m256i, _mm256_cmpeq_epi8, _mm256_loadu_si256, _mm256_movemask_epi8, _mm256_set1_epi8,
_mm256_storeu_si256,
};
Expand Down Expand Up @@ -46,8 +46,7 @@ pub(crate) unsafe fn parse_str<'invoke, 'de>(
loop {
// _mm256_loadu_si256 does not require alignment
#[allow(clippy::cast_ptr_alignment)]
let v: __m256i =
_mm256_loadu_si256(src.as_ptr().add(src_i).cast::<std::arch::x86_64::__m256i>());
let v: __m256i = _mm256_loadu_si256(src.as_ptr().add(src_i).cast::<__m256i>());

// store to dest unconditionally - we can overwrite the bits we don't like
// later
Expand Down Expand Up @@ -97,18 +96,11 @@ pub(crate) unsafe fn parse_str<'invoke, 'de>(
loop {
// _mm256_loadu_si256 does not require alignment
#[allow(clippy::cast_ptr_alignment)]
let v: __m256i =
_mm256_loadu_si256(src.as_ptr().add(src_i).cast::<std::arch::x86_64::__m256i>());
let v: __m256i = _mm256_loadu_si256(src.as_ptr().add(src_i).cast::<__m256i>());

// _mm256_storeu_si256 does not require alignment
#[allow(clippy::cast_ptr_alignment)]
_mm256_storeu_si256(
buffer
.as_mut_ptr()
.add(dst_i)
.cast::<std::arch::x86_64::__m256i>(),
v,
);
_mm256_storeu_si256(buffer.as_mut_ptr().add(dst_i).cast::<__m256i>(), v);

// store to dest unconditionally - we can overwrite the bits we don't like
// later
Expand Down
38 changes: 23 additions & 15 deletions src/impls/avx2/stage1.rs
View file
Open in desktop
Original file line number Diff line number Diff line change
@@ -1,18 +1,16 @@
#![allow(dead_code)]
use crate::{static_cast_i32, static_cast_i64, static_cast_u32, Stage1Parse};
#[cfg(target_arch = "x86")]
use std::arch::x86::{
__m256i, _mm256_add_epi32, _mm256_and_si256, _mm256_cmpeq_epi8, _mm256_loadu_si256,
_mm256_max_epu8, _mm256_movemask_epi8, _mm256_set1_epi8, _mm256_set_epi32, _mm256_setr_epi8,
_mm256_setzero_si256, _mm256_shuffle_epi8, _mm256_srli_epi32, _mm256_storeu_si256,
_mm_clmulepi64_si128, _mm_cvtsi128_si64, _mm_set1_epi8, _mm_set_epi64x,
};
use std::arch::x86 as arch;

#[cfg(target_arch = "x86_64")]
use std::arch::x86_64::{
use std::arch::x86_64 as arch;

use arch::{
__m256i, _mm256_add_epi32, _mm256_and_si256, _mm256_cmpeq_epi8, _mm256_loadu_si256,
_mm256_max_epu8, _mm256_movemask_epi8, _mm256_set1_epi8, _mm256_set_epi32, _mm256_setr_epi8,
_mm256_setzero_si256, _mm256_shuffle_epi8, _mm256_srli_epi32, _mm256_storeu_si256,
_mm_clmulepi64_si128, _mm_cvtsi128_si64, _mm_set1_epi8, _mm_set_epi64x,
_mm_clmulepi64_si128, _mm_set1_epi8, _mm_set_epi64x,
};

macro_rules! low_nibble_mask {
Expand Down Expand Up @@ -56,14 +54,29 @@ impl Stage1Parse for SimdInput {
#[cfg_attr(not(feature = "no-inline"), inline)]
#[allow(clippy::cast_sign_loss)]
#[target_feature(enable = "avx2")]
#[cfg(target_arch = "x86_64")]
unsafe fn compute_quote_mask(quote_bits: u64) -> u64 {
_mm_cvtsi128_si64(_mm_clmulepi64_si128(
std::arch::x86_64::_mm_cvtsi128_si64(_mm_clmulepi64_si128(
_mm_set_epi64x(0, static_cast_i64!(quote_bits)),
_mm_set1_epi8(-1_i8 /* 0xFF */),
0,
)) as u64
}

#[cfg_attr(not(feature = "no-inline"), inline)]
#[allow(clippy::cast_sign_loss)]
#[target_feature(enable = "avx2")]
#[cfg(target_arch = "x86")]
unsafe fn compute_quote_mask(quote_bits: u64) -> u64 {
let mut quote_mask: u64 = quote_bits ^ (quote_bits << 1);
quote_mask = quote_mask ^ (quote_mask << 2);
quote_mask = quote_mask ^ (quote_mask << 4);
quote_mask = quote_mask ^ (quote_mask << 8);
quote_mask = quote_mask ^ (quote_mask << 16);
quote_mask = quote_mask ^ (quote_mask << 32);
quote_mask
}

/// a straightforward comparison of a mask against input
#[cfg_attr(not(feature = "no-inline"), inline)]
#[allow(clippy::cast_possible_wrap, clippy::cast_sign_loss)]
Expand Down Expand Up @@ -215,12 +228,7 @@ impl Stage1Parse for SimdInput {

let v: __m256i = _mm256_set_epi32(v7, v6, v5, v4, v3, v2, v1, v0);
let v: __m256i = _mm256_add_epi32(idx_64_v, v);
_mm256_storeu_si256(
base.as_mut_ptr()
.add(l)
.cast::<std::arch::x86_64::__m256i>(),
v,
);
_mm256_storeu_si256(base.as_mut_ptr().add(l).cast::<__m256i>(), v);
l += 8;
}
// We have written all the data
Expand Down
Loading