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

CUDA: mul_mat_q RDNA2 tunings (#2910)

Browse source 
* CUDA: mul_mat_q RDNA2 tunings

* Update ggml-cuda.cu

Co-authored-by: Henri Vasserman <henv@hot.ee>

---------

Co-authored-by: Henri Vasserman <henv@hot.ee>
This commit is contained in:
Johannes Gäßler 2023-09-13 11:20:24 +02:00 • committed by GitHub
parent 84e723653c
commit 0a5eebb45d
No known key found for this signature in database
GPG key ID: 4AEE18F83AFDEB23
3 changed files with 426 additions and 45 deletions
Download patch file Download diff file Expand all files Collapse all files

1
CMakeLists.txt
View file

@ -388,7 +388,6 @@ if (LLAMA_HIPBLAS)
target_compile_definitions(ggml-rocm PRIVATE GGML_CUDA_DMMV_X=${LLAMA_CUDA_DMMV_X})
target_compile_definitions(ggml-rocm PRIVATE GGML_CUDA_MMV_Y=${LLAMA_CUDA_MMV_Y})
target_compile_definitions(ggml-rocm PRIVATE K_QUANTS_PER_ITERATION=${LLAMA_CUDA_KQUANTS_ITER})
target_compile_definitions(ggml-rocm PRIVATE CC_TURING=1000000000)
set_source_files_properties(ggml-cuda.cu PROPERTIES LANGUAGE CXX)
target_link_libraries(ggml-rocm PRIVATE hip::device PUBLIC hip::host roc::rocblas roc::hipblas)