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 7, 04:00 UTC
cpu: aarch64: Upgrade OneDNN for aarch64 to add support for int8 matmul (#36147)
Signed-off-by: Nikhil Gupta <nikhil.gupta2@arm.com>
Nikhil Gupta committed
Mar 6, 2026 at 03:48 UTC
0a49676fb0e54c9229a39f6304bc88b7d24e0355
1 file changed
+18
-7
cmake/cpu_extension.cmake
+18
-7
@@ -242,13 +242,24 @@ if (ENABLE_X86_ISA OR (ASIMD_FOUND AND NOT APPLE_SILICON_FOUND) OR POWER9_FOUND
242
)
243
else()
244
message(STATUS "Downloading oneDNN from GitHub")
245
- FetchContent_Declare(
246
- oneDNN
247
- GIT_REPOSITORY https://github.com/oneapi-src/oneDNN.git
248
- GIT_TAG v3.10
249
- GIT_PROGRESS TRUE
250
- GIT_SHALLOW TRUE
251
- )
245
+ if(ASIMD_FOUND AND NOT APPLE_SILICON_FOUND)
246
+ message(STATUS "aarch64 detected: using pinned oneDNN commit 9c5be1cc59e368aebf0909e6cf20f981ea61462a")
247
+ FetchContent_Declare(
248
+ oneDNN
249
+ GIT_REPOSITORY https://github.com/oneapi-src/oneDNN.git
250
+ GIT_TAG 9c5be1cc59e368aebf0909e6cf20f981ea61462a
251
+ GIT_PROGRESS TRUE
252
+ GIT_SHALLOW FALSE
253
+ )
254
+ else()
255
+ FetchContent_Declare(
256
+ oneDNN
257
+ GIT_REPOSITORY https://github.com/oneapi-src/oneDNN.git
258
+ GIT_TAG v3.10
259
+ GIT_PROGRESS TRUE
260
+ GIT_SHALLOW TRUE
261
+ )
262
+ endif()
263
endif()
264
265
set(ONEDNN_LIBRARY_TYPE "STATIC")