Skip to content

arm64 improvements for x86 clmul & fma #1350

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

Draft
mr-c wants to merge 3 commits into
base: master
Choose a base branch
from
Draft

arm64 improvements for x86 clmul & fma #1350

Show file tree
Hide file tree
Changes from all commits
Commits
Show all changes
3 commits
Select commit Hold shift + click to select a range
2f62f4e
x86 clmul: remove unneeded AES & and put an upper-bounds on the
mr-c Oct 17, 2025
1f21d9d
x86 fma: improve fallback for simde_mm256_fmadd_pd
mr-c Oct 17, 2025
b047a45
gh-actions macos: drop deprecated macos-13 runner, add xcode 16.3 and…
mr-c Oct 17, 2025
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
16 changes: 13 additions & 3 deletions .github/workflows/ci.yml
View file
Open in desktop
Original file line number Diff line number Diff line change
Expand Up @@ -900,16 +900,26 @@ jobs:
arch_flags: -march=native -Wno-poison-system-directories
# "-Wno-poison-system-directories": due to meson test
# "-march=native": without, or with -mcpu=apple-m1 then unavailable __ARM_FEATURE_s get enabled
- xcode: "15.4"
os: macos-14 # arm64
- xcode: "15.4" # clang 15?
os: macos-14 # arm64
arch_flags: -march=native -Wno-poison-system-directories
# "-Wno-poison-system-directories": due to meson test
# "-march=native": without, or with -mcpu=apple-m1 then unavailable __ARM_FEATURE_s get enabled
- xcode: "16.0"
- xcode: "16.0" # clang 16?
os: macos-15 # arm64
arch_flags: -march=native -Wno-poison-system-directories -U_LIBCPP_ENABLE_ASSERTIONS
# "-Wno-poison-system-directories": due to meson test
# "-march=native": without, or with -mcpu=apple-m1 then unavailable __ARM_FEATURE_s get enabled
- xcode: "16.3" # clang 17?
os: macos-15 # arm64
arch_flags: -march=native -Wno-poison-system-directories -U_LIBCPP_ENABLE_ASSERTIONS -Wno-switch-default
# "-Wno-poison-system-directories": due to meson test
# "-march=native": without, or with -mcpu=apple-m1 then unavailable __ARM_FEATURE_s get enabled
- xcode: "26.0.1" # clang 17?
os: macos-26 # arm64
arch_flags: -march=native -Wno-poison-system-directories -U_LIBCPP_ENABLE_ASSERTIONS -Wno-switch-default
# "-Wno-poison-system-directories": due to meson test
# "-march=native": without, or with -mcpu=apple-m1 then unavailable __ARM_FEATURE_s get enabled
runs-on: ${{ matrix.os }}
env:
DEVELOPER_DIR: /Applications/Xcode_${{ matrix.xcode }}.app
Expand Down
2 changes: 1 addition & 1 deletion simde/x86/clmul.h
View file
Open in desktop
Original file line number Diff line number Diff line change
Expand Up @@ -203,7 +203,7 @@ simde_mm_clmulepi64_si128 (simde__m128i a, simde__m128i b, const int imm8)
#else
#define simde_mm_clmulepi64_si128(a, b, imm8) _mm_clmulepi64_si128(a, b, imm8)
#endif
#elif defined(SIMDE_ARM_NEON_A64V8_NATIVE) && defined(SIMDE_ARCH_ARM_AES) && !defined(__clang__)
#elif defined(SIMDE_ARM_NEON_A64V8_NATIVE) && !(SIMDE_DETECT_CLANG_VERSION_NOT(22,0,0))
#define simde_mm_clmulepi64_si128(a, b, imm8) \
simde__m128i_from_neon_u64( \
vreinterpretq_u64_p128( \
Expand Down
10 changes: 10 additions & 0 deletions simde/x86/fma.h
View file
Open in desktop
Original file line number Diff line number Diff line change
Expand Up @@ -82,6 +82,16 @@ simde_mm256_fmadd_pd (simde__m256d a, simde__m256d b, simde__m256d c) {
return _mm256_fmadd_pd(a, b, c);
#elif defined(SIMDE_LOONGARCH_LASX_NATIVE)
return __lasx_xvfmadd_d(a, b, c);
#elif SIMDE_NATURAL_VECTOR_SIZE_LE(128)
simde__m256d_private
r_,
a_ = simde__m256d_to_private(a),
b_ = simde__m256d_to_private(b),
c_ = simde__m256d_to_private(c);

r_.m128d[0] = simde_mm_fmadd_pd(a_.m128d[0], b_.m128d[0], c_.m128d[0]);
r_.m128d[1] = simde_mm_fmadd_pd(a_.m128d[1], b_.m128d[1], c_.m128d[1]);
return simde__m256d_from_private(r_);
#else
return simde_mm256_add_pd(simde_mm256_mul_pd(a, b), c);
#endif
Expand Down
Loading