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

Update Makefile for the Cuda kernels

Browse source
This commit is contained in:
Henri Vasserman 2023-04-20 22:19:22 +03:00
parent 0fd8363adc
commit 54a63c10e8
1 changed files with 9 additions and 2 deletions
Download patch file Download diff file Expand all files Collapse all files

11
Makefile
View file

@ -108,8 +108,15 @@ ggml-cuda.o: ggml-cuda.cu ggml-cuda.h
nvcc -arch=native -c -o $@ $< nvcc -arch=native -c -o $@ $<
endif endif
ifdef LLAMA_HIPBLAS ifdef LLAMA_HIPBLAS
CFLAGS += -DGGML_USE_HIPBLAS -D__HIP_PLATFORM_AMD__ -I/opt/rocm/include ROCMPATH?= /opt/rocm
LDFLAGS += -lhipblas -lamdhip64 -L/opt/rocm/lib CFLAGS += -DGGML_USE_HIPBLAS -D__HIP_PLATFORM_AMD__ -I$(ROCMPATH)/include
CXXFLAGS+= -D__HIP_PLATFORM_AMD__ -I$(ROCMPATH)/include
HIPFLAGS?= -amdgpu-early-inline-all=true -amdgpu-function-calls=false -march=native
LDFLAGS += -lhipblas -lamdhip64 -L$(ROCMPATH)/lib
HIPCC ?= $(ROCMPATH)/bin/hipcc
OBJS += ggml-cuda.o
ggml-cuda.o: ggml-cuda.cu ggml-cuda.h
$(HIPCC) $(CXXFLAGS) -x hip $(HIPFLAGS) -c -o $@ $<
endif endif
ifdef LLAMA_GPROF ifdef LLAMA_GPROF
CFLAGS += -pg CFLAGS += -pg