Skip to content

Fix cross-compiling librt by enabling x86_64 optimisations with pragmas #20815

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

Open
chewi wants to merge 1 commit into
base: master
Choose a base branch
from
+50 −16
Open

Fix cross-compiling librt by enabling x86_64 optimisations with pragmas #20815

Show file tree
Hide file tree
Changes from all commits
Commits
Show all changes
1 commit
Select commit
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
9 changes: 1 addition & 8 deletions mypyc/build_setup.py
View file
Open in desktop
Original file line number Diff line number Diff line change
Expand Up @@ -19,18 +19,11 @@
from distutils import ccompiler

EXTRA_FLAGS_PER_COMPILER_TYPE_PER_PATH_COMPONENT = {
"unix": {
"base64/arch/ssse3": ["-mssse3"],
"base64/arch/sse41": ["-msse4.1"],
"base64/arch/sse42": ["-msse4.2"],
"base64/arch/avx2": ["-mavx2"],
"base64/arch/avx": ["-mavx"],
},
"msvc": {
"base64/arch/sse42": ["/arch:SSE4.2"],
"base64/arch/avx2": ["/arch:AVX2"],
"base64/arch/avx": ["/arch:AVX"],
},
}
}

ccompiler.CCompiler.__spawn = ccompiler.CCompiler.spawn # type: ignore[attr-defined]
Expand Down
8 changes: 8 additions & 0 deletions mypyc/lib-rt/base64/arch/avx/codec.c
View file
Open in desktop
Original file line number Diff line number Diff line change
Expand Up @@ -9,6 +9,11 @@
#include "../../env.h"

#if HAVE_AVX
#if defined(__clang__)
#pragma clang attribute push (__attribute__((target("avx"))), apply_to=function)
#else
#pragma GCC target("avx")
#endif
#include <immintrin.h>

// Only enable inline assembly on supported compilers and on 64-bit CPUs.
Expand Down Expand Up @@ -62,6 +67,9 @@ base64_stream_decode_avx BASE64_DEC_PARAMS
#include "../generic/dec_head.c"
dec_loop_ssse3(&s, &slen, &o, &olen);
#include "../generic/dec_tail.c"
#if defined(__clang__)
#pragma clang attribute pop
#endif
#else
return base64_dec_stub(state, src, srclen, out, outlen);
#endif
Expand Down
8 changes: 8 additions & 0 deletions mypyc/lib-rt/base64/arch/avx2/codec.c
View file
Open in desktop
Original file line number Diff line number Diff line change
Expand Up @@ -9,6 +9,11 @@
#include "../../env.h"

#if HAVE_AVX2
#if defined(__clang__)
#pragma clang attribute push (__attribute__((target("avx2"))), apply_to=function)
#else
#pragma GCC target("avx2")
#endif
#include <immintrin.h>

// Only enable inline assembly on supported compilers and on 64-bit CPUs.
Expand Down Expand Up @@ -52,6 +57,9 @@ base64_stream_decode_avx2 BASE64_DEC_PARAMS
#include "../generic/dec_head.c"
dec_loop_avx2(&s, &slen, &o, &olen);
#include "../generic/dec_tail.c"
#if defined(__clang__)
#pragma clang attribute pop
#endif
#else
return base64_dec_stub(state, src, srclen, out, outlen);
#endif
Expand Down
8 changes: 8 additions & 0 deletions mypyc/lib-rt/base64/arch/avx512/codec.c
View file
Open in desktop
Original file line number Diff line number Diff line change
Expand Up @@ -9,6 +9,11 @@
#include "../../env.h"

#if HAVE_AVX512
#if defined(__clang__)
#pragma clang attribute push (__attribute__((target("avx512vbmi"))), apply_to=function)
#else
#pragma GCC target("avx512vbmi")
#endif
#include <immintrin.h>

#include "../avx2/dec_reshuffle.c"
Expand Down Expand Up @@ -38,6 +43,9 @@ base64_stream_decode_avx512 BASE64_DEC_PARAMS
#include "../generic/dec_head.c"
dec_loop_avx2(&s, &slen, &o, &olen);
#include "../generic/dec_tail.c"
#if defined(__clang__)
#pragma clang attribute pop
#endif
#else
return base64_dec_stub(state, src, srclen, out, outlen);
#endif
Expand Down
8 changes: 8 additions & 0 deletions mypyc/lib-rt/base64/arch/sse41/codec.c
View file
Open in desktop
Original file line number Diff line number Diff line change
Expand Up @@ -9,6 +9,11 @@
#include "../../env.h"

#if HAVE_SSE41
#if defined(__clang__)
#pragma clang attribute push (__attribute__((target("sse4.1"))), apply_to=function)
#else
#pragma GCC target("sse4.1")
#endif
#include <smmintrin.h>

// Only enable inline assembly on supported compilers and on 64-bit CPUs.
Expand Down Expand Up @@ -52,6 +57,9 @@ base64_stream_decode_sse41 BASE64_DEC_PARAMS
#include "../generic/dec_head.c"
dec_loop_ssse3(&s, &slen, &o, &olen);
#include "../generic/dec_tail.c"
#if defined(__clang__)
#pragma clang attribute pop
#endif
#else
return base64_dec_stub(state, src, srclen, out, outlen);
#endif
Expand Down
8 changes: 8 additions & 0 deletions mypyc/lib-rt/base64/arch/sse42/codec.c
View file
Open in desktop
Original file line number Diff line number Diff line change
Expand Up @@ -9,6 +9,11 @@
#include "../../env.h"

#if HAVE_SSE42
#if defined(__clang__)
#pragma clang attribute push (__attribute__((target("sse4.2"))), apply_to=function)
#else
#pragma GCC target("sse4.2")
#endif
#include <nmmintrin.h>

// Only enable inline assembly on supported compilers and on 64-bit CPUs.
Expand Down Expand Up @@ -52,6 +57,9 @@ base64_stream_decode_sse42 BASE64_DEC_PARAMS
#include "../generic/dec_head.c"
dec_loop_ssse3(&s, &slen, &o, &olen);
#include "../generic/dec_tail.c"
#if defined(__clang__)
#pragma clang attribute pop
#endif
#else
return base64_dec_stub(state, src, srclen, out, outlen);
#endif
Expand Down
8 changes: 8 additions & 0 deletions mypyc/lib-rt/base64/arch/ssse3/codec.c
View file
Open in desktop
Original file line number Diff line number Diff line change
Expand Up @@ -9,6 +9,11 @@
#include "../../env.h"

#if HAVE_SSSE3
#if defined(__clang__)
#pragma clang attribute push (__attribute__((target("ssse3"))), apply_to=function)
#else
#pragma GCC target("ssse3")
#endif
#include <tmmintrin.h>

// Only enable inline assembly on supported compilers and on 64-bit CPUs.
Expand Down Expand Up @@ -54,6 +59,9 @@ base64_stream_decode_ssse3 BASE64_DEC_PARAMS
#include "../generic/dec_head.c"
dec_loop_ssse3(&s, &slen, &o, &olen);
#include "../generic/dec_tail.c"
#if defined(__clang__)
#pragma clang attribute pop
#endif
#else
return base64_dec_stub(state, src, srclen, out, outlen);
#endif
Expand Down
9 changes: 1 addition & 8 deletions mypyc/lib-rt/build_setup.py
View file
Open in desktop
Original file line number Diff line number Diff line change
Expand Up @@ -19,18 +19,11 @@
from distutils import ccompiler

EXTRA_FLAGS_PER_COMPILER_TYPE_PER_PATH_COMPONENT = {
"unix": {
"base64/arch/ssse3": ["-mssse3"],
"base64/arch/sse41": ["-msse4.1"],
"base64/arch/sse42": ["-msse4.2"],
"base64/arch/avx2": ["-mavx2"],
"base64/arch/avx": ["-mavx"],
},
"msvc": {
"base64/arch/sse42": ["/arch:SSE4.2"],
"base64/arch/avx2": ["/arch:AVX2"],
"base64/arch/avx": ["/arch:AVX"],
},
}
}

ccompiler.CCompiler.__spawn = ccompiler.CCompiler.spawn # type: ignore[attr-defined]
Expand Down