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

Add GGML_MUSA in Makefile

Browse source 
Signed-off-by: Xiaodong Ye <xiaodong.ye@mthreads.com>
This commit is contained in:
Xiaodong Ye 2024-07-14 15:59:20 +08:00
parent cb2c688648
commit 779c920b88
1 changed files with 47 additions and 10 deletions
Download patch file Download diff file Expand all files Collapse all files

57
Makefile
View file

@ -526,10 +526,21 @@ ifndef GGML_NO_ACCELERATE
endif endif
endif # GGML_NO_ACCELERATE endif # GGML_NO_ACCELERATE
ifdef GGML_MUSA
CC := clang
CXX := clang++
GGML_CUDA := 1
MK_CPPFLAGS += -DGGML_USE_MUSA
endif
ifndef GGML_NO_OPENMP ifndef GGML_NO_OPENMP
MK_CPPFLAGS += -DGGML_USE_OPENMP MK_CPPFLAGS += -DGGML_USE_OPENMP
MK_CFLAGS += -fopenmp MK_CFLAGS += -fopenmp
MK_CXXFLAGS += -fopenmp MK_CXXFLAGS += -fopenmp
ifdef GGML_MUSA
MK_CPPFLAGS += -I/usr/lib/llvm-10/include/openmp
MK_LDFLAGS += -L/usr/lib/llvm-10/lib
endif # GGML_MUSA
endif # GGML_NO_OPENMP endif # GGML_NO_OPENMP
ifdef GGML_OPENBLAS ifdef GGML_OPENBLAS
@ -574,15 +585,27 @@ else
endif # GGML_CUDA_FA_ALL_QUANTS endif # GGML_CUDA_FA_ALL_QUANTS
ifdef GGML_CUDA ifdef GGML_CUDA
ifneq ('', '$(wildcard /opt/cuda)') ifdef GGML_MUSA
CUDA_PATH ?= /opt/cuda ifneq ('', '$(wildcard /opt/musa)')
else CUDA_PATH ?= /opt/musa
CUDA_PATH ?= /usr/local/cuda else
endif CUDA_PATH ?= /usr/local/musa
endif
MK_CPPFLAGS += -DGGML_USE_CUDA -I$(CUDA_PATH)/include -I$(CUDA_PATH)/targets/$(UNAME_M)-linux/include -DGGML_CUDA_USE_GRAPHS MK_CPPFLAGS += -DGGML_USE_CUDA -I$(CUDA_PATH)/include
MK_LDFLAGS += -lcuda -lcublas -lculibos -lcudart -lcublasLt -lpthread -ldl -lrt -L$(CUDA_PATH)/lib64 -L/usr/lib64 -L$(CUDA_PATH)/targets/$(UNAME_M)-linux/lib -L$(CUDA_PATH)/lib64/stubs -L/usr/lib/wsl/lib MK_LDFLAGS += -lmusa -lmublas -lmusart -lpthread -ldl -lrt -L$(CUDA_PATH)/lib -L/usr/lib64
MK_NVCCFLAGS += -use_fast_math MK_NVCCFLAGS += -x musa -mtgpu --cuda-gpu-arch=mp_22
else
ifneq ('', '$(wildcard /opt/cuda)')
CUDA_PATH ?= /opt/cuda
else
CUDA_PATH ?= /usr/local/cuda
endif
MK_CPPFLAGS += -DGGML_USE_CUDA -I$(CUDA_PATH)/include -I$(CUDA_PATH)/targets/$(UNAME_M)-linux/include -DGGML_CUDA_USE_GRAPHS
MK_LDFLAGS += -lcuda -lcublas -lculibos -lcudart -lcublasLt -lpthread -ldl -lrt -L$(CUDA_PATH)/lib64 -L/usr/lib64 -L$(CUDA_PATH)/targets/$(UNAME_M)-linux/lib -L$(CUDA_PATH)/lib64/stubs -L/usr/lib/wsl/lib
MK_NVCCFLAGS += -use_fast_math
endif # GGML_MUSA
OBJ_GGML += ggml/src/ggml-cuda.o OBJ_GGML += ggml/src/ggml-cuda.o
OBJ_GGML += $(patsubst %.cu,%.o,$(wildcard ggml/src/ggml-cuda/*.cu)) OBJ_GGML += $(patsubst %.cu,%.o,$(wildcard ggml/src/ggml-cuda/*.cu))
@ -592,9 +615,11 @@ ifdef LLAMA_FATAL_WARNINGS
MK_NVCCFLAGS += -Werror all-warnings MK_NVCCFLAGS += -Werror all-warnings
endif # LLAMA_FATAL_WARNINGS endif # LLAMA_FATAL_WARNINGS
ifndef GGML_MUSA
ifndef JETSON_EOL_MODULE_DETECT ifndef JETSON_EOL_MODULE_DETECT
MK_NVCCFLAGS += --forward-unknown-to-host-compiler MK_NVCCFLAGS += --forward-unknown-to-host-compiler
endif # JETSON_EOL_MODULE_DETECT endif # JETSON_EOL_MODULE_DETECT
endif # GGML_MUSA
ifdef LLAMA_DEBUG ifdef LLAMA_DEBUG
MK_NVCCFLAGS += -lineinfo MK_NVCCFLAGS += -lineinfo
@ -607,8 +632,12 @@ endif # GGML_CUDA_DEBUG
ifdef GGML_CUDA_NVCC ifdef GGML_CUDA_NVCC
NVCC = $(CCACHE) $(GGML_CUDA_NVCC) NVCC = $(CCACHE) $(GGML_CUDA_NVCC)
else else
NVCC = $(CCACHE) nvcc ifdef GGML_MUSA
endif #GGML_CUDA_NVCC NVCC = $(CCACHE) mcc
else
NVCC = $(CCACHE) nvcc
endif # GGML_MUSA
endif # GGML_CUDA_NVCC
ifdef CUDA_DOCKER_ARCH ifdef CUDA_DOCKER_ARCH
MK_NVCCFLAGS += -Wno-deprecated-gpu-targets -arch=$(CUDA_DOCKER_ARCH) MK_NVCCFLAGS += -Wno-deprecated-gpu-targets -arch=$(CUDA_DOCKER_ARCH)
@ -679,9 +708,15 @@ define NVCC_COMPILE
$(NVCC) -I. -Icommon -D_XOPEN_SOURCE=600 -D_GNU_SOURCE -DNDEBUG -DGGML_USE_CUDA -I/usr/local/cuda/include -I/opt/cuda/include -I/usr/local/cuda/targets/aarch64-linux/include -std=c++11 -O3 $(NVCCFLAGS) $(CPPFLAGS) -Xcompiler "$(CUDA_CXXFLAGS)" -c $< -o $@ $(NVCC) -I. -Icommon -D_XOPEN_SOURCE=600 -D_GNU_SOURCE -DNDEBUG -DGGML_USE_CUDA -I/usr/local/cuda/include -I/opt/cuda/include -I/usr/local/cuda/targets/aarch64-linux/include -std=c++11 -O3 $(NVCCFLAGS) $(CPPFLAGS) -Xcompiler "$(CUDA_CXXFLAGS)" -c $< -o $@
endef # NVCC_COMPILE endef # NVCC_COMPILE
else else
ifdef GGML_MUSA
define NVCC_COMPILE
$(NVCC) $(NVCCFLAGS) $(CPPFLAGS) -c $< -o $@
endef # NVCC_COMPILE
else
define NVCC_COMPILE define NVCC_COMPILE
$(NVCC) $(NVCCFLAGS) $(CPPFLAGS) -Xcompiler "$(CUDA_CXXFLAGS)" -c $< -o $@ $(NVCC) $(NVCCFLAGS) $(CPPFLAGS) -Xcompiler "$(CUDA_CXXFLAGS)" -c $< -o $@
endef # NVCC_COMPILE endef # NVCC_COMPILE
endif # GGML_MUSA
endif # JETSON_EOL_MODULE_DETECT endif # JETSON_EOL_MODULE_DETECT
ggml/src/ggml-cuda/%.o: \ ggml/src/ggml-cuda/%.o: \
@ -907,6 +942,7 @@ $(info I CXX: $(shell $(CXX) --version | head -n 1))
ifdef GGML_CUDA ifdef GGML_CUDA
$(info I NVCC: $(shell $(NVCC) --version | tail -n 1)) $(info I NVCC: $(shell $(NVCC) --version | tail -n 1))
CUDA_VERSION := $(shell $(NVCC) --version | grep -oP 'release (\K[0-9]+\.[0-9])') CUDA_VERSION := $(shell $(NVCC) --version | grep -oP 'release (\K[0-9]+\.[0-9])')
ifndef GGML_MUSA
ifeq ($(shell awk -v "v=$(CUDA_VERSION)" 'BEGIN { print (v < 11.7) }'),1) ifeq ($(shell awk -v "v=$(CUDA_VERSION)" 'BEGIN { print (v < 11.7) }'),1)
ifndef CUDA_DOCKER_ARCH ifndef CUDA_DOCKER_ARCH
@ -916,6 +952,7 @@ endif # CUDA_POWER_ARCH
endif # CUDA_DOCKER_ARCH endif # CUDA_DOCKER_ARCH
endif # eq ($(shell echo "$(CUDA_VERSION) < 11.7" | bc),1) endif # eq ($(shell echo "$(CUDA_VERSION) < 11.7" | bc),1)
endif # GGML_MUSA
endif # GGML_CUDA endif # GGML_CUDA
$(info ) $(info )