vbatts/llama.cpp
1
0
Fork 0
Code Issues Pull requests Projects Releases Packages Wiki Activity Actions

HIP: force max threads per block to be 1024

Browse source 
Some old compilers still use 256. Explicitly set it to 1024 to get correct
result from ops like ARGMAX and GROUP_NORM.

Related: #10610, #11619
Signed-off-by: fxzjshm <fxzjshm@163.com>
This commit is contained in:
fxzjshm 2025-02-03 22:33:38 +08:00
parent d92cb67e37
commit 59ad593a95
No known key found for this signature in database
GPG key ID: 7638FA33A7259C00
1 changed files with 3 additions and 0 deletions
Download patch file Download diff file Expand all files Collapse all files

3
ggml/src/ggml-hip/CMakeLists.txt
View file

@ -46,6 +46,9 @@ endif()
message(STATUS "HIP and hipBLAS found")
# Workaround old compilers
set(CMAKE_HIP_FLAGS "${CMAKE_HIP_FLAGS} --gpu-max-threads-per-block=1024")
file(GLOB GGML_HEADERS_ROCM "../ggml-cuda/*.cuh")
list(APPEND GGML_HEADERS_ROCM "../../include/ggml-cuda.h")