third-party-mirrors/ollama
1
0
Fork 1
Code Issues Pull Requests Actions Packages Projects Releases Wiki Activity

fix Makefile on windows

Browse Source
This commit is contained in:
jmorganca 2024-06-20 21:52:10 -04:00
parent b3c62dcafd
commit 8dda9293fa
2 changed files with 8 additions and 6 deletions
Show Stats Download Patch File Download Diff File Expand all files Collapse all files

10
llama/Makefile
View File

@ -1,6 +1,8 @@
OS := $(shell uname -s)
ARCH := $(or $(ARCH), $(shell uname -m))
NVCC := nvcc
HIPCC := "$(HIP_PATH)/bin/hipcc.bin.exe"
export CGO_CFLAGS_ALLOW = -mfma|-mf16c
export CGO_CXXFLAGS_ALLOW = -mfma|-mf16c
@ -12,7 +14,8 @@ endif
ifneq (,$(findstring MINGW,$(OS)))
OBJ_EXT := obj
SHARED_EXT := dll
HIP_PATH := $(shell cygpath -w -s "$(HIP_PATH)")
CUDA_LIB := $(shell cygpath -w -s "$(CUDA_PATH)\lib\x64")
HIP_LIB := $(shell cygpath -w -s "$(HIP_PATH)\lib")
else
OBJ_EXT := o
SHARED_EXT := so
@ -62,7 +65,6 @@ CUDA_FLAGS := \
HIP_ARCHS := gfx900 gfx940 gfx941 gfx942 gfx1010 gfx1012 gfx1030 gfx1100 gfx1101 gfx1102
LINUX_HIP_ARCHS := gfx906:xnack- gfx908:xnack- gfx90a:xnack+ gfx90a:xnack-
HIPCC := "$(HIP_PATH)/bin/hipcc.bin.exe"
HIP_FLAGS := \
-c \
-O3 \
@ -142,10 +144,10 @@ ollama_runner_avx2:
CGO_ENABLED=1 GOARCH=$(ARCH) go build -ldflags "-s -w" -tags avx,avx2 -o $@ ./runner
ollama_runner_cuda: ggml_cuda.dll
CGO_ENABLED=1 GOARCH=$(ARCH) go build -ldflags "-s -w" -tags avx,cuda -o $@ ./runner
CGO_ENABLED=1 GOARCH=$(ARCH) CGO_LDFLAGS=-L"$(CUDA_LIB)" go build -ldflags "-s -w" -tags avx,cuda -o $@ ./runner
ollama_runner_rocm: ggml_hipblas.dll
CGO_ENABLED=1 GOARCH=$(ARCH) go build -ldflags "-s -w" -tags avx,rocm -o $@ ./runner
CGO_ENABLED=1 GOARCH=$(ARCH) CGO_LDFLAGS=-L"$(HIP_LIB)" go build -ldflags "-s -w" -tags avx,rocm -o $@ ./runner
clean:
rm -f $(CUDA_OBJS) $(HIP_OBJS) ggml_cuda.$(SHARED_EXT) ggml_cuda.* ggml_hipblas.* ollama_runner*

4
llama/llama.go
View File

@ -25,8 +25,8 @@ package llama
#cgo rocm CFLAGS: -DGGML_USE_CUDA -DGGML_USE_HIPBLAS -DGGML_CUDA_DMMV_X=32 -DGGML_CUDA_PEER_MAX_BATCH_SIZE=128 -DGGML_CUDA_MMV_Y=1 -DGGML_BUILD=1
#cgo rocm CXXFLAGS: -DGGML_USE_CUDA -DGGML_USE_HIPBLAS -DGGML_CUDA_DMMV_X=32 -DGGML_CUDA_PEER_MAX_BATCH_SIZE=128 -DGGML_CUDA_MMV_Y=1 -DGGML_BUILD=1
#cgo rocm LDFLAGS: -L${SRCDIR} -lggml_hipblas -lhipblas -lamdhip64 -lrocblas
#cgo windows,cuda LDFLAGS: -L${SRCDIR} -L"C:/Program Files/NVIDIA GPU Computing Toolkit/CUDA/v11.3/lib/x64" -lggml_cuda -lcuda -lcudart -lcublas -lcublasLt
#cgo windows,rocm LDFLAGS: -L${SRCDIR} -L"C:/Program Files/AMD/ROCm/5.7/lib" -lggml_hipblas -lhipblas -lamdhip64 -lrocblas
#cgo windows,cuda LDFLAGS: -L${SRCDIR} -lggml_cuda -lcuda -lcudart -lcublas -lcublasLt
#cgo windows,rocm LDFLAGS: -L${SRCDIR} -lggml_hipblas -lhipblas -lamdhip64 -lrocblas
#cgo linux,cuda LDFLAGS: -L${SRCDIR} -L/usr/local/cuda/lib64 -lggml_cuda -lcuda -lcudart -lcublas -lcublasLt -lpthread -ldl -lrt
#cgo linux,rocm LDFLAGS: -L/opt/rocm/lib