root/llama.cpp
1
0
Fork 0
mirror of https://github.com/ggerganov/llama.cpp.git synced 2024-12-26 03:14:35 +00:00
Code Issues Actions 7 Packages Projects Releases Wiki Activity

mtgpu: Add MUSA_DOCKER_ARCH in Dockerfiles && update cmake and make (#10516)

Browse Source 
Signed-off-by: Xiaodong Ye <xiaodong.ye@mthreads.com>
This commit is contained in:
R0CKSTAR 2024-11-27 00:00:41 +08:00 committed by GitHub
parent 904109ed0d
commit 249cd93da3
No known key found for this signature in database
GPG Key ID: B5690EEEBB952194
5 changed files with 39 additions and 8 deletions
Show Stats Download Patch File Download Diff File Expand all files Collapse all files

9
.devops/full-musa.Dockerfile
View File

@ -6,6 +6,9 @@ ARG BASE_MUSA_DEV_CONTAINER=mthreads/musa:${MUSA_VERSION}-devel-ubuntu${UBUNTU_V
FROM ${BASE_MUSA_DEV_CONTAINER} AS build
# MUSA architecture to build for (defaults to all supported archs)
ARG MUSA_DOCKER_ARCH=default
RUN apt-get update && \
apt-get install -y build-essential cmake python3 python3-pip git libcurl4-openssl-dev libgomp1
@ -19,7 +22,11 @@ WORKDIR /app
COPY . .
RUN cmake -B build -DGGML_NATIVE=OFF -DGGML_MUSA=ON -DLLAMA_CURL=ON ${CMAKE_ARGS} -DCMAKE_EXE_LINKER_FLAGS=-Wl,--allow-shlib-undefined . && \
# Use the default MUSA archs if not specified
RUN if [ "${MUSA_DOCKER_ARCH}" != "default" ]; then \
export CMAKE_ARGS="-DMUSA_ARCHITECTURES=${MUSA_DOCKER_ARCH}"; \
fi && \
cmake -B build -DGGML_NATIVE=OFF -DGGML_MUSA=ON -DLLAMA_CURL=ON ${CMAKE_ARGS} -DCMAKE_EXE_LINKER_FLAGS=-Wl,--allow-shlib-undefined . && \
cmake --build build --config Release -j$(nproc) && \
cp build/bin/* .

9
.devops/llama-cli-musa.Dockerfile
View File

@ -8,6 +8,9 @@ ARG BASE_MUSA_RUN_CONTAINER=mthreads/musa:${MUSA_VERSION}-runtime-ubuntu${UBUNTU
FROM ${BASE_MUSA_DEV_CONTAINER} AS build
# MUSA architecture to build for (defaults to all supported archs)
ARG MUSA_DOCKER_ARCH=default
RUN apt-get update && \
apt-get install -y build-essential git cmake
@ -15,7 +18,11 @@ WORKDIR /app
COPY . .
RUN cmake -B build -DGGML_NATIVE=OFF -DGGML_MUSA=ON ${CMAKE_ARGS} -DCMAKE_EXE_LINKER_FLAGS=-Wl,--allow-shlib-undefined . && \
# Use the default MUSA archs if not specified
RUN if [ "${MUSA_DOCKER_ARCH}" != "default" ]; then \
export CMAKE_ARGS="-DMUSA_ARCHITECTURES=${MUSA_DOCKER_ARCH}"; \
fi && \
cmake -B build -DGGML_NATIVE=OFF -DGGML_MUSA=ON ${CMAKE_ARGS} -DCMAKE_EXE_LINKER_FLAGS=-Wl,--allow-shlib-undefined . && \
cmake --build build --config Release --target llama-cli -j$(nproc) && \
mkdir -p /app/lib && \
find build -name "*.so" -exec cp {} /app/lib \;

9
.devops/llama-server-musa.Dockerfile
View File

@ -8,6 +8,9 @@ ARG BASE_MUSA_RUN_CONTAINER=mthreads/musa:${MUSA_VERSION}-runtime-ubuntu${UBUNTU
FROM ${BASE_MUSA_DEV_CONTAINER} AS build
# MUSA architecture to build for (defaults to all supported archs)
ARG MUSA_DOCKER_ARCH=default
RUN apt-get update && \
apt-get install -y build-essential git cmake libcurl4-openssl-dev
@ -15,7 +18,11 @@ WORKDIR /app
COPY . .
RUN cmake -B build -DGGML_NATIVE=OFF -DGGML_MUSA=ON -DLLAMA_CURL=ON ${CMAKE_ARGS} -DCMAKE_EXE_LINKER_FLAGS=-Wl,--allow-shlib-undefined . && \
# Use the default MUSA archs if not specified
RUN if [ "${MUSA_DOCKER_ARCH}" != "default" ]; then \
export CMAKE_ARGS="-DMUSA_ARCHITECTURES=${MUSA_DOCKER_ARCH}"; \
fi && \
cmake -B build -DGGML_NATIVE=OFF -DGGML_MUSA=ON -DLLAMA_CURL=ON ${CMAKE_ARGS} -DCMAKE_EXE_LINKER_FLAGS=-Wl,--allow-shlib-undefined . && \
cmake --build build --config Release --target llama-server -j$(nproc) && \
mkdir -p /app/lib && \
find build -name "*.so" -exec cp {} /app/lib \;

9
Makefile
View File

@ -815,7 +815,7 @@ ifdef GGML_MUSA
else
MUSA_PATH ?= /opt/musa
endif
MTGPU_TARGETS ?= mp_21 mp_22
MUSA_ARCHITECTURES ?= 21;22
MK_CPPFLAGS += -DGGML_USE_MUSA -DGGML_USE_CUDA
MK_LDFLAGS += -L$(MUSA_PATH)/lib -Wl,-rpath=$(MUSA_PATH)/lib
@ -834,7 +834,8 @@ ifdef GGML_MUSA
CXX := $(MUSA_PATH)/bin/clang++
MCC := $(CCACHE) $(MUSA_PATH)/bin/mcc
MUSAFLAGS += $(addprefix --cuda-gpu-arch=, $(MTGPU_TARGETS))
MUSAFLAGS = -x musa -mtgpu
MUSAFLAGS += $(foreach arch,$(subst ;, ,$(MUSA_ARCHITECTURES)),--cuda-gpu-arch=mp_$(arch))
ifdef GGML_CUDA_FORCE_MMQ
MUSAFLAGS += -DGGML_CUDA_FORCE_MMQ
@ -878,14 +879,14 @@ ggml/src/ggml-cuda/ggml-cuda.o: \
ggml/src/ggml-backend-impl.h \
ggml/src/ggml-common.h \
$(wildcard ggml/src/ggml-cuda/*.cuh)
$(MCC) $(CXXFLAGS) $(MUSAFLAGS) -x musa -mtgpu -c -o $@ $<
$(MCC) $(CXXFLAGS) $(MUSAFLAGS) -c -o $@ $<
ggml/src/ggml-cuda/%.o: \
ggml/src/ggml-cuda/%.cu \
ggml/include/ggml.h \
ggml/src/ggml-common.h \
ggml/src/ggml-cuda/common.cuh
$(MCC) $(CXXFLAGS) $(MUSAFLAGS) -x musa -mtgpu -c -o $@ $<
$(MCC) $(CXXFLAGS) $(MUSAFLAGS) -c -o $@ $<
endif # GGML_MUSA
ifdef GGML_METAL

11
ggml/src/ggml-musa/CMakeLists.txt
View File

@ -20,6 +20,11 @@ find_package(MUSAToolkit)
if (MUSAToolkit_FOUND)
message(STATUS "MUSA Toolkit found")
if (NOT DEFINED MUSA_ARCHITECTURES)
set(MUSA_ARCHITECTURES "21;22")
endif()
message(STATUS "Using MUSA architectures: ${MUSA_ARCHITECTURES}")
file(GLOB GGML_HEADERS_MUSA "../ggml-cuda/*.cuh")
list(APPEND GGML_HEADERS_MUSA "../../include/ggml-cuda.h")
@ -44,7 +49,11 @@ if (MUSAToolkit_FOUND)
set_source_files_properties(${GGML_SOURCES_MUSA} PROPERTIES LANGUAGE CXX)
foreach(SOURCE ${GGML_SOURCES_MUSA})
set_property(SOURCE ${SOURCE} PROPERTY COMPILE_FLAGS "-x musa -mtgpu --cuda-gpu-arch=mp_21 --cuda-gpu-arch=mp_22")
set(COMPILE_FLAGS "-x musa -mtgpu")
foreach(ARCH ${MUSA_ARCHITECTURES})
set(COMPILE_FLAGS "${COMPILE_FLAGS} --cuda-gpu-arch=mp_${ARCH}")
endforeach()
set_property(SOURCE ${SOURCE} PROPERTY COMPILE_FLAGS ${COMPILE_FLAGS})
endforeach()
ggml_add_backend_library(ggml-musa