A high-throughput and memory-efficient inference and serving engine for LLMs
Mirrored from https://github.com/tduncan-lum/Nvidia-Optimization-Stack---VLLM.git
Synced Nov 14, 05:30 UTC
[Hardware][Apple-CPU] Disable OneDNN build for Apple Silicon (#24200)
Signed-off-by: ignaciosica <mignacio.sica@gmail.com> Co-authored-by: Li, Jiang <jiang1.li@intel.com>
Ignacio Sica committed
Sep 4, 2025 at 06:48 UTC
369a079568f8e8ced4d9dac4da2fbfdd451e778e
1 file changed
+2
-1
cmake/cpu_extension.cmake
+2
-1
@@ -88,6 +88,7 @@ is_avx512_disabled(AVX512_DISABLED)
88
89
if (MACOSX_FOUND AND CMAKE_SYSTEM_PROCESSOR STREQUAL "arm64")
90
message(STATUS "Apple Silicon Detected")
91
+ set(APPLE_SILICON_FOUND TRUE)
92
set(ENABLE_NUMA OFF)
93
check_sysctl(hw.optional.neon ASIMD_FOUND)
94
check_sysctl(hw.optional.arm.FEAT_BF16 ARM_BF16_FOUND)
@@ -189,7 +190,7 @@ else()
190
set(USE_ACL OFF)
191
endif()
192
192
-if ((AVX512_FOUND AND NOT AVX512_DISABLED) OR ASIMD_FOUND OR POWER9_FOUND OR POWER10_FOUND OR POWER11_FOUND)
193
+if ((AVX512_FOUND AND NOT AVX512_DISABLED) OR (ASIMD_FOUND AND NOT APPLE_SILICON_FOUND) OR POWER9_FOUND OR POWER10_FOUND OR POWER11_FOUND)
194
FetchContent_Declare(
195
oneDNN
196
GIT_REPOSITORY https://github.com/oneapi-src/oneDNN.git