Skip to content

ggml-cpu: cmake add arm64 cpu feature check for macos #10487

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

Merged
merged 2 commits into from
Nov 26, 2024
Merged

ggml-cpu: cmake add arm64 cpu feature check for macos #10487

Changes from 1 commit
Commits
Show all changes
2 commits
Select commit Hold shift + click to select a range
8797958
ggml-cpu: cmake add arm64 cpu feature check for macos
chaxu01 Nov 25, 2024
ac90ee3
use vmmlaq_s32 for compile option i8mm check
chaxu01 Nov 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
18 changes: 18 additions & 0 deletions ggml/src/ggml-cpu/CMakeLists.txt
View file
Open in desktop
Original file line number Diff line number Diff line change
Expand Up @@ -97,6 +97,24 @@ if (CMAKE_OSX_ARCHITECTURES STREQUAL "arm64" OR
endif ()

set(CMAKE_REQUIRED_FLAGS ${CMAKE_REQUIRED_FLAGS_PREV})
elseif (APPLE)
if (GGML_NATIVE)
set(MARCH_FLAGS "-march=armv8.2a")

check_cxx_source_compiles("#include <arm_neon.h>\nint main() { int8x16_t _a, _b; int32x4_t _s = vdotq_s32(_s, _a, _b); return 0; }" GGML_COMPILER_SUPPORT_DOTPROD)
if (GGML_COMPILER_SUPPORT_DOTPROD)
set(MARCH_FLAGS "${MARCH_FLAGS}+dotprod")
add_compile_definitions(__ARM_FEATURE_DOTPROD)
endif ()

check_cxx_source_compiles("#include <arm_neon.h>\nint main() { int8x16_t _a, _b; int32x4_t _s = vmlaq_f32(_s, _a, _b); return 0; }" GGML_COMPILER_SUPPORT_MATMUL_INT8)
Copy link
Member

@ggerganov ggerganov Nov 25, 2024

Choose a reason for hiding this comment

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

How come the presence of vmlaq_f32 confirms __ARM_FEATURE_MATMUL_INT8?

Copy link
Collaborator Author

@chaxu01 chaxu01 Nov 25, 2024

Choose a reason for hiding this comment

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

Thanks for your question. The current check using vmlaq_f32 to confirm __ARM_FEATURE_MATMUL_INT8 is incorrect because vmlaq_f32 operates on floating-point values (float32x4_t), while __ARM_FEATURE_MATMUL_INT8 is specifically related to integer matrix multiplication using INT8 (8-bit signed integers).

I'll update the patch using an operation that deals with INT8 data types, such as vmlaq_s32.

if (GGML_COMPILER_SUPPORT_MATMUL_INT8)
add_compile_definitions(__ARM_FEATURE_MATMUL_INT8)
set(MARCH_FLAGS "${MARCH_FLAGS}+i8mm")
endif ()

list(APPEND ARCH_FLAGS "${MARCH_FLAGS}")
endif ()
else()
check_cxx_compiler_flag(-mfp16-format=ieee COMPILER_SUPPORTS_FP16_FORMAT_I3E)
if (NOT "${COMPILER_SUPPORTS_FP16_FORMAT_I3E}" STREQUAL "")
Expand Down
Loading