Skip to content

Add optional AVX512-FP16 arithmetic for the scalar quantizer. #4225

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 10 commits into
base: main
Choose a base branch
from
Open

Add optional AVX512-FP16 arithmetic for the scalar quantizer. #4225

Show file tree
Hide file tree
Changes from all commits
Commits
Show all changes
10 commits
Select commit Hold shift + click to select a range
08faa85
Add optional AVX512-FP16 arithmetic for scalar quantizer.
mulugetam Mar 6, 2025
e9e69d9
Fix formatting issue in ScalarQuantizer.
mulugetam Mar 6, 2025
9dc2fd0
Merge branch 'main' into avx512-fp16
junjieqi Mar 7, 2025
456f335
Merge branch 'main' into avx512-fp16
junjieqi Mar 17, 2025
bcff377
Fix issue due operator precedence.
mulugetam Mar 17, 2025
f75a7fe
Fix template instantiation issue for arm.
mulugetam Mar 18, 2025
8b4ce8a
Merge branch 'main' into avx512-fp16
mulugetam Mar 18, 2025
6bd1e9f
Merge branch 'main' into avx512-fp16
junjieqi Mar 21, 2025
9eecfe4
Merge branch 'main' into avx512-fp16
junjieqi Apr 1, 2025
8e08aa0
Merge branch 'main' into avx512-fp16
junjieqi Apr 3, 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
1 change: 1 addition & 0 deletions CMakeLists.txt
View file
Open in desktop
Original file line number Diff line number Diff line change
Expand Up @@ -67,6 +67,7 @@ option(FAISS_ENABLE_PYTHON "Build Python extension." ON)
option(FAISS_ENABLE_C_API "Build C API." OFF)
option(FAISS_ENABLE_EXTRAS "Build extras like benchmarks and demos" ON)
option(FAISS_USE_LTO "Enable Link-Time optimization" OFF)
option(FAISS_ENABLE_AVX512_FP16 "Enable AVX512-FP16 arithmetic (for avx512_spr opt level)." OFF)
Copy link
Contributor

@mdouze mdouze Apr 1, 2025

Choose a reason for hiding this comment

The reason will be displayed to describe this comment to others. Learn more.

This should be a runtime option not a compile-time option.

Copy link
Contributor Author

@mulugetam mulugetam Apr 1, 2025

Choose a reason for hiding this comment

The reason will be displayed to describe this comment to others. Learn more.

Thanks @mdouze. The current functions, like reconstruct_8_components and reconstruct_16_components, are picked at compile time. I'm not exactly sure how to approach what you're suggesting. Could you explain a bit more?


if(FAISS_ENABLE_GPU)
if(FAISS_ENABLE_ROCM)
Expand Down
8 changes: 8 additions & 0 deletions faiss/CMakeLists.txt
View file
Open in desktop
Original file line number Diff line number Diff line change
Expand Up @@ -360,6 +360,14 @@ target_compile_definitions(faiss_avx512 PRIVATE FINTEGER=int)
target_compile_definitions(faiss_avx512_spr PRIVATE FINTEGER=int)
target_compile_definitions(faiss_sve PRIVATE FINTEGER=int)

if(FAISS_ENABLE_AVX512_FP16)
if (FAISS_OPT_LEVEL STREQUAL "avx512_spr")
target_compile_definitions(faiss_avx512_spr PRIVATE ENABLE_AVX512_FP16)
else()
message(STATUS "AVX512_FP16 not supported: requires FAISS_OPT_LEVEL=avx512_spr.")
endif()
endif()

if(FAISS_USE_LTO)
include(CheckIPOSupported)
check_ipo_supported(RESULT ipo_supported OUTPUT ipo_error)
Expand Down
Loading
Loading