tqcq/llama.cpp
0
0
Fork 0
mirror of https://github.com/ggml-org/llama.cpp.git synced 2025-06-26 19:55:04 +00:00
Code Issues Packages Projects Releases Wiki Activity

CUDA: Quantized matrix matrix multiplication (#2160)

Browse Source 
* mmq implementation for non k-quants

* q6_K

* q2_K

* q3_k

* q4_K

* vdr

* q5_K

* faster q8_1 loading

* loop unrolling

* add __restrict__

* q2_K sc_high

* GGML_CUDA_MMQ_Y

* Updated Makefile

* Update Makefile

* DMMV_F16 -> F16

* Updated README, CMakeLists

* Fix CMakeLists.txt

* Fix CMakeLists.txt

* Fix multi GPU out-of-bounds
This commit is contained in:
Johannes Gäßler
2023-07-29 23:04:44 +02:00
committed by GitHub
parent 9baf9ef304
commit 11f3ca06b8
4 changed files with 1295 additions and 324 deletions
Download Patch File Download Diff File Expand all files Collapse all files

8
CMakeLists.txt
View File

@ -67,7 +67,9 @@ endif()
option(LLAMA_ACCELERATE "llama: enable Accelerate framework" ON)
option(LLAMA_BLAS "llama: use BLAS" OFF)
set(LLAMA_BLAS_VENDOR "Generic" CACHE STRING "llama: BLAS library vendor")
option(LLAMA_CUBLAS "llama: use cuBLAS" OFF)
option(LLAMA_CUBLAS "llama: use CUDA" OFF)
option(LLAMA_CUDA_CUBLAS "llama: use cuBLAS for prompt processing" OFF)
set(LLAMA_CUDA_MMQ_Y "64" CACHE STRING "llama: y tile size for mmq CUDA kernels")
option(LLAMA_CUDA_FORCE_DMMV "llama: use dmmv instead of mmvq CUDA kernels" OFF)
set(LLAMA_CUDA_DMMV_X "32" CACHE STRING "llama: x stride for dmmv CUDA kernels")
set(LLAMA_CUDA_MMV_Y "1" CACHE STRING "llama: y block size for mmv CUDA kernels")
@ -251,6 +253,10 @@ if (LLAMA_CUBLAS)
set(GGML_SOURCES_CUDA ggml-cuda.cu ggml-cuda.h)
add_compile_definitions(GGML_USE_CUBLAS)
if (LLAMA_CUDA_CUBLAS)
add_compile_definitions(GGML_CUDA_CUBLAS)
endif()
add_compile_definitions(GGML_CUDA_MMQ_Y=${LLAMA_CUDA_MMQ_Y})
if (LLAMA_CUDA_FORCE_DMMV)
add_compile_definitions(GGML_CUDA_FORCE_DMMV)
endif()