Skip to content

Fix: static SIMD dispatch falls to scalar for avx512_spr/avx512/arm_sve builds #5057

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
mulugetam wants to merge 1 commit into
base: main
Choose a base branch
from
Open

Fix: static SIMD dispatch falls to scalar for avx512_spr/avx512/arm_sve builds #5057

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
27 changes: 24 additions & 3 deletions faiss/impl/simd_dispatch.h
View file
Open in desktop
Original file line number Diff line number Diff line change
Expand Up @@ -106,11 +106,32 @@ inline auto with_selected_simd_levels(LambdaType&& action) {
}
#else // static dispatch
// In static mode, SINGLE_SIMD_LEVEL is a constexpr resolved at compile
// time. If the compiled level is not in the available set, fall through
// to NONE (mirroring the DD fallthrough behavior). Only SINGLE_SIMD_LEVEL
// and NONE have compiled specializations.
// time. We mirror the DD fallthrough behavior at compile time:
// x86: AVX512_SPR -> AVX512 -> AVX2 -> NONE
// ARM: ARM_SVE -> ARM_NEON -> NONE
// This ensures that e.g. an AVX512_SPR build can use AVX512
// specializations when no AVX512_SPR-specific implementation exists,
// and an AVX512 build can use AVX2 for 256-bit operations.

if constexpr (available_levels & (1 << int(SINGLE_SIMD_LEVEL))) {
// Exact match — use the compiled-in level directly.
return action.template operator()<SINGLE_SIMD_LEVEL>();
} else if constexpr (
SINGLE_SIMD_LEVEL == SIMDLevel::AVX512_SPR &&
(available_levels & (1 << int(SIMDLevel::AVX512)))) {
// AVX512_SPR -> AVX512 fallthrough.
return action.template operator()<SIMDLevel::AVX512>();
} else if constexpr (
(SINGLE_SIMD_LEVEL == SIMDLevel::AVX512_SPR ||
SINGLE_SIMD_LEVEL == SIMDLevel::AVX512) &&
(available_levels & (1 << int(SIMDLevel::AVX2)))) {
// AVX512_SPR/AVX512 -> AVX2 fallthrough
return action.template operator()<SIMDLevel::AVX2>();
} else if constexpr (
SINGLE_SIMD_LEVEL == SIMDLevel::ARM_SVE &&
(available_levels & (1 << int(SIMDLevel::ARM_NEON)))) {
// ARM_SVE -> ARM_NEON fallthrough.
return action.template operator()<SIMDLevel::ARM_NEON>();
} else {
return action.template operator()<SIMDLevel::NONE>();
}
Expand Down
Loading