root/llama.cpp
1
0
Fork 0
mirror of https://github.com/ggerganov/llama.cpp.git synced 2024-12-25 02:44:36 +00:00
Code Issues Actions 9 Packages Projects Releases Wiki Activity

Fixes for building SYCL backend for AMD GPUs

Browse Source 
- cmake has a hardcoded `onemkl` that will cause linker failure for AMD
  path and needs to be changed
- docfix: oneMKL compile needs `-DHIP_TARGETS` not `-DHIPTARGETS`
- docfix: modified cmake build parameters to pass in more linker flags,
  otherwise linking will fail (might be slight overkill, but without
  adding `-DCMAKE_SHARED_LINKER_FLAGS` it fails
This commit is contained in:
lhl 2024-12-16 16:44:57 +09:00
parent 4ddd199f6f
commit 5cb6209de5
2 changed files with 8 additions and 6 deletions
Show Stats Download Patch File Download Diff File Expand all files Collapse all files

12
docs/backend/SYCL.md
View File

@ -237,7 +237,7 @@ cmake --build buildWithCublas --config Release
git clone https://github.com/oneapi-src/oneMKL
cd oneMKL
# Find your HIPTARGET with rocminfo, under the key 'Name:'
cmake -B buildWithrocBLAS -DCMAKE_CXX_COMPILER=icpx -DCMAKE_C_COMPILER=icx -DENABLE_MKLGPU_BACKEND=OFF -DENABLE_MKLCPU_BACKEND=OFF -DENABLE_ROCBLAS_BACKEND=ON -DHIPTARGETS=${HIPTARGET} -DTARGET_DOMAINS=blas
cmake -B buildWithrocBLAS -DCMAKE_CXX_COMPILER=icpx -DCMAKE_C_COMPILER=icx -DENABLE_MKLGPU_BACKEND=OFF -DENABLE_MKLCPU_BACKEND=OFF -DENABLE_ROCBLAS_BACKEND=ON -DHIP_TARGETS=${HIPTARGET} -DTARGET_DOMAINS=blas
cmake --build buildWithrocBLAS --config Release
```
@ -330,9 +330,10 @@ cmake --build build --config Release -j -v
```sh
# Export relevant ENV variables
export LD_LIBRARY_PATH=/path/to/oneMKL/buildWithrocBLAS/lib:$LD_LIBRARY_PATH
export LIBRARY_PATH=/path/to/oneMKL/buildWithrocBLAS/lib:$LIBRARY_PATH
export CPLUS_INCLUDE_DIR=/path/to/oneMKL/buildWithrocBLAS/include:$CPLUS_INCLUDE_DIR
export MKL_ROCBLAS=/path/to/oneMKL/buildWithrocmBLAS/lib
export LD_LIBRARY_PATH=$MKL_ROCBLAS:$LD_LIBRARY_PATH
export LIBRARY_PATH=$MKL_ROCBLAS:$LIBRARY_PATH
export CPLUS_INCLUDE_DIR=$MKL_ROCBLAS:$CPLUS_INCLUDE_DIR
# Build LLAMA with rocBLAS acceleration through SYCL
@ -340,7 +341,8 @@ export CPLUS_INCLUDE_DIR=/path/to/oneMKL/buildWithrocBLAS/include:$CPLUS_INCLUDE
# Use FP32, FP16 is not supported
# Find your GGML_SYCL_DEVICE_ARCH with rocminfo, under the key 'Name:'
GGML_SYCL_DEVICE_ARCH=gfx90a # Example architecture
cmake -B build -DGGML_SYCL=ON -DGGML_SYCL_TARGET=AMD -DGGML_SYCL_DEVICE_ARCH=${GGML_SYCL_DEVICE_ARCH} -DCMAKE_C_COMPILER=icx -DCMAKE_CXX_COMPILER=icpx
cmake -B build -DGGML_SYCL=ON -DGGML_SYCL_TARGET=AMD -DGGML_SYCL_DEVICE_ARCH=${GGML_SYCL_DEVICE_ARCH} -DCMAKE_C_COMPILER=icx -DCMAKE_CXX_COMPILER=icpx -DCMAKE_SHARED_LINKER_FLAGS="-L$MKL_ROCBLAS -L/opt/intel/oneapi/mkl/latest/lib/intel64 -lonemath_blas_rocblas -Wl,--no-as-needed -lmkl_sycl -lmkl_intel_ilp64 -lmkl_sequential -lmkl_core"
# build all binary
cmake --build build --config Release -j -v

2
ggml/src/ggml-sycl/CMakeLists.txt
View File

@ -75,7 +75,7 @@ else()
message(ERROR "Can't enable SYCL hip backend, GGML_SYCL_DEVICE_ARCH has not been set.")
endif()
set(CMAKE_CXX_FLAGS "${CMAKE_CXX_FLAGS} -fsycl-targets=amdgcn-amd-amdhsa")
target_link_libraries(ggml-sycl PRIVATE sycl pthread m dl onemkl)
target_link_libraries(ggml-sycl PRIVATE sycl pthread m dl onemath_blas_rocblas)
endif()
if (GGML_SYCL_DEVICE_ARCH)