search

ggerganov / llama.cpp

LLM inference in C/C++
MIT License
65.7k stars 9.43k forks source link

ROCm 5.6: make: /opt/rocm/llvm/bin/amdgpu-arch: No such file or directory #2787

Closed grigio closed 1 year ago

grigio commented 1 year ago

UPDATE: Can you specify with version of ROCm did you use? I tried to installa latest ROCm and now I get:

make: /opt/rocm/llvm/bin/amdgpu-arch: No such file or directory
/opt/rocm/bin/hipcc -I. -I./common -O3 -std=c++11 -fPIC -DNDEBUG -Wall -Wextra -Wpedantic -Wcast-qual -Wno-unused-function -Wno-multichar -pthread -march=native -mtune=native -DGGML_USE_K_QUANTS -DGGML_USE_HIPBLAS -DGGML_USE_CUBLAS  -DGGML_CUDA_DMMV_X=32 -DGGML_CUDA_MMV_Y=1 -DK_QUANTS_PER_ITERATION=2 -DCC_TURING=1000000000 -x hip -c -o ggml-cuda.o ggml-cuda.cu
make: /opt/rocm/bin/hipcc: No such file or directory
make: *** [Makefile:304: ggml-cuda.o] Error 127

make LLAMA_HIPBLAS=1
I llama.cpp build info: 
I UNAME_S:  Linux
I UNAME_P:  unknown
I UNAME_M:  x86_64
I CFLAGS:   -I.            -O3 -std=c11   -fPIC -DNDEBUG -Wall -Wextra -Wpedantic -Wcast-qual -Wdouble-promotion -Wshadow -Wstrict-prototypes -Wpointer-arith -Wmissing-prototypes -pthread -march=native -mtune=native -DGGML_USE_K_QUANTS -DGGML_USE_HIPBLAS -DGGML_USE_CUBLAS
I CXXFLAGS: -I. -I./common -O3 -std=c++11 -fPIC -DNDEBUG -Wall -Wextra -Wpedantic -Wcast-qual -Wno-unused-function -Wno-multichar -pthread -march=native -mtune=native -DGGML_USE_K_QUANTS -DGGML_USE_HIPBLAS -DGGML_USE_CUBLAS
I LDFLAGS:  -L/opt/rocm/lib -Wl,-rpath=/opt/rocm/lib -lhipblas -lamdhip64 -lrocblas
I CC:       cc (Debian 12.2.0-14) 12.2.0
I CXX:      g++ (Debian 12.2.0-14) 12.2.0

/opt/rocm/bin/hipcc -I. -I./common -O3 -std=c++11 -fPIC -DNDEBUG -Wall -Wextra -Wpedantic -Wcast-qual -Wno-unused-function -Wno-multichar -pthread -march=native -mtune=native -DGGML_USE_K_QUANTS -DGGML_USE_HIPBLAS -DGGML_USE_CUBLAS --offload-arch=gfx1036 -DGGML_CUDA_DMMV_X=32 -DGGML_CUDA_MMV_Y=1 -DK_QUANTS_PER_ITERATION=2 -DCC_TURING=1000000000 -x hip -c -o ggml-cuda.o ggml-cuda.cu
In file included from <built-in>:1:
/opt/rocm-5.4.3/llvm/lib/clang/15.0.0/include/__clang_hip_runtime_wrapper.h:50:10: fatal error: 'cmath' file not found
#include <cmath>
         ^~~~~~~
1 error generated when compiling for gfx1036.
make: *** [Makefile:304: ggml-cuda.o] Error 1

i tried to reinstall libstdc++-12-dev as suggested here but I still get that error

find /opt/rocm/ | grep cmath
/opt/rocm/include/hipify/__clang_hip_cmath.h
/opt/rocm/include/hipify/__clang_cuda_cmath.h
/opt/rocm/llvm/lib/clang/15.0.0/include/openmp_wrappers/cmath
/opt/rocm/llvm/lib/clang/15.0.0/include/openmp_wrappers/complex_cmath.h
/opt/rocm/llvm/lib/clang/15.0.0/include/__clang_hip_cmath.h
/opt/rocm/llvm/lib/clang/15.0.0/include/__clang_cuda_cmath.h
grigio commented 1 year ago

OK, it seems this is the proper way #1087

arch-btw commented 11 months ago

What specifically fixed it?