cicc not found when using mexcuda

Hi,

I am trying to compile a cuda file into a mex file via the command:

mexcuda -v -g test.cu

where test.cu is a simple hello world program.

I am using Ubuntu 20.04

I get the following error messages:

Trying MEX options '/usr/local/MATLAB/R2020a/toolbox/parallel/gpu/extern/src/mex/glnxa64/nvcc_g++.xml'...SUCCESS

mex -largeArrayDims -f /usr/local/MATLAB/R2020a/toolbox/parallel/gpu/extern/src/mex/glnxa64/nvcc_g++.xml NVCC_FLAGS="" -v -g test.cu

Verbose mode is on.

... Looking for compiler 'nvcc' ...

... Looking for environment variable 'MW_NVCC_PATH' ...No.

... Looking for environment variable 'CUDA_BIN_PATH' ...No.

... Looking for environment variable 'CUDA_PATH' ...No.

... Executing command 'which nvcc' ...Yes ('/usr/bin/nvcc').

... Executing command '(((if [ -z ${MW_ALLOW_ANY_CUDA} ]; then false; else true; fi;) || /usr/bin/nvcc --version | egrep -E 'release 10.1') > /dev/null) && echo /usr/bin/nvcc' ...Yes ('/usr/bin/nvcc').

... Executing command '(((if [ -z ${MW_ALLOW_ANY_CUDA} ]; then false; else true; fi;) || /usr/bin/nvcc --version | egrep -E 'release 10.1') > /dev/null) && echo /usr/bin/nvcc' ...Yes ('/usr/bin/nvcc').

... Looking for folder '/usr' ...Yes.

... Looking for environment variable 'MW_NVCC_PATH' ...No.

... Looking for environment variable 'CUDA_BIN_PATH' ...No.

... Looking for environment variable 'CUDA_PATH' ...No.

... Executing command 'which nvcc' ...Yes ('/usr/bin/nvcc').

... Executing command '(((if [ -z ${MW_ALLOW_ANY_CUDA} ]; then false; else true; fi;) || /usr/bin/nvcc --version | egrep -E 'release 10.1') > /dev/null) && echo /usr/bin/nvcc' ...Yes ('/usr/bin/nvcc').

... Executing command '(((if [ -z ${MW_ALLOW_ANY_CUDA} ]; then false; else true; fi;) || /usr/bin/nvcc --version | egrep -E 'release 10.1') > /dev/null) && echo /usr/bin/nvcc' ...Yes ('/usr/bin/nvcc').

... Looking for folder '/usr/nvvm/libdevice' ...No.

... Looking for file '/usr/local/MATLAB/R2020a/sys/cuda/glnxa64/cuda/bin/nvcc' ...Yes.

... Looking for folder '/usr/local/MATLAB/R2020a/bin/glnxa64' ...Yes.

... Looking for environment variable 'MW_NVCC_PATH' ...No.

... Looking for environment variable 'CUDA_BIN_PATH' ...No.

... Looking for environment variable 'CUDA_PATH' ...No.

... Executing command 'which nvcc' ...Yes ('/usr/bin/nvcc').

... Executing command '(((if [ -z ${MW_ALLOW_ANY_CUDA} ]; then false; else true; fi;) || /usr/bin/nvcc --version | egrep -E 'release 10.1') > /dev/null) && echo /usr/bin/nvcc' ...Yes ('/usr/bin/nvcc').

... Executing command '(((if [ -z ${MW_ALLOW_ANY_CUDA} ]; then false; else true; fi;) || /usr/bin/nvcc --version | egrep -E 'release 10.1') > /dev/null) && echo /usr/bin/nvcc' ...Yes ('/usr/bin/nvcc').

... Looking for folder '/usr/lib64' ...Yes.

... Executing command 'which g++' ...Yes ('/usr/bin/g++').

... Executing command 'g++ -print-file-name=libstdc++.so' ...Yes ('/usr/lib/gcc/x86_64-linux-gnu/8/libstdc++.so').

... Executing command 'g++ -dumpversion' ...Yes ('8').

... Executing command 'which g++' ...Yes ('/usr/bin/g++').

... Looking for folder '/usr' ...Yes.

... Executing command 'g++ -dumpmachine' ...Yes ('x86_64-linux-gnu').

Found installed compiler 'nvcc'.

Set PATH = :/usr/nvvm/bin:/usr/bin:;/usr/local/cuda-8.0/bin:/home/simon/.local/bin:/usr/local/sbin:/usr/local/bin:/usr/sbin:/usr/bin:/sbin:/bin:/usr/games:/usr/local/games:/snap/bin:/usr/local/cuda-8.0/bin

Set INCLUDE = :/usr/lib/gcc/x86_64-linux-gnu/8/include;/usr/include/c++/8;/usr/include/c++/8/x86_64-linux-gnu;/usr/include/c++/8/backward;:/usr/lib/gcc/x86_64-linux-gnu/8/include;/usr/include/c++/8;/usr/include/c++/8/x86_64-linux-gnu;/usr/include/c++/8/backward;

Options file details

-------------------------------------------------------------------

Compiler location: /usr/bin/nvcc

Options file: /usr/local/MATLAB/R2020a/toolbox/parallel/gpu/extern/src/mex/glnxa64/nvcc_g++.xml

CMDLINE2 : /usr/bin/g++ -pthread -Wl,--no-undefined -shared -g -Wl,--version-script,"/usr/local/MATLAB/R2020a/extern/lib/glnxa64/c_exportsmexfileversion.map" /tmp/mex_22266400607270_177416/test.o /tmp/mex_22266400607270_177416/c_mexapi_version.o -Wl,--as-needed -Wl,-rpath-link,/usr/lib64 -L"/usr/lib64" -Wl,-rpath-link,/usr/lib64 -L"/usr/lib64" -Wl,-rpath-link,/usr/local/MATLAB/R2020a/bin/glnxa64 -L"/usr/local/MATLAB/R2020a/bin/glnxa64" -Wl,-rpath-link,/usr/local/MATLAB/R2020a/extern/bin/glnxa64 -L"/usr/local/MATLAB/R2020a/extern/bin/glnxa64" -lMatlabDataArray -lmx -lmex -lmat -lm -lc -lstdc++ -lmwgpu /usr/lib64/libcudart.so.* -o test.mexa64

CXX : /usr/bin/nvcc

DEFINES : -DMX_COMPAT_64 -DMATLAB_DEFAULT_RELEASE=R2017b -DUSE_MEX_CMD --compiler-options=-D_GNU_SOURCE,-DMATLAB_MEX_FILE

MATLABMEX : -DMATLAB_MEX_FILE

NVCCFLAGS : -gencode=arch=compute_30,code=sm_30 -gencode=arch=compute_50,code=sm_50 -gencode=arch=compute_60,code=sm_60 -gencode=arch=compute_70,code=\"sm_70,compute_70\" -std=c++11 --dont-use-profile -ldir /usr/local/MATLAB/R2020a/bin/glnxa64

CFLAGS : --compiler-options=-fexceptions,-fPIC,-fno-omit-frame-pointer,-pthread,-fwrapv

CXXFLAGS : --compiler-options=-fexceptions,-fPIC,-fno-omit-frame-pointer,-pthread,-fwrapv

INCLUDE : -I"/usr/local/MATLAB/R2020a/extern/include" -I"/usr/local/MATLAB/R2020a/simulink/include" -I"/usr/local/MATLAB/R2020a/toolbox/parallel/gpu/extern/include/" -I"/usr/include"

CXXOPTIMFLAGS : -O2 -DNDEBUG

CXXDEBUGFLAGS : -g

LDXX : /usr/bin/g++

LDFLAGS : -pthread -Wl,--no-undefined

LDTYPE : -shared

LINKEXPORT : -Wl,--version-script,"/usr/local/MATLAB/R2020a/extern/lib/glnxa64/mexFunction.map"

LINKEXPORTVER : -Wl,--version-script,"/usr/local/MATLAB/R2020a/extern/lib/glnxa64/c_exportsmexfileversion.map"

LINKLIBS : -Wl,--as-needed -Wl,-rpath-link,/usr/lib64 -L"/usr/lib64" -Wl,-rpath-link,/usr/lib64 -L"/usr/lib64" -Wl,-rpath-link,/usr/local/MATLAB/R2020a/bin/glnxa64 -L"/usr/local/MATLAB/R2020a/bin/glnxa64" -Wl,-rpath-link,/usr/local/MATLAB/R2020a/extern/bin/glnxa64 -L"/usr/local/MATLAB/R2020a/extern/bin/glnxa64" -lMatlabDataArray -lmx -lmex -lmat -lm -lc -lstdc++ -lmwgpu /usr/lib64/libcudart.so.*

NVVMIR_LIBRARY : --dont-use-profile -ldir /usr/local/MATLAB/R2020a/bin/glnxa64

LDOPTIMFLAGS : -O

LDDEBUGFLAGS : -g

MWCPPLIB : "/usr/local/MATLAB/R2020a/sys/os/glnxa64/libstdc++.so.6"

OBJEXT : .o

LDEXT : .mexa64

SETENV : CC="gcc"

CXX="/usr/bin/nvcc"

CFLAGS="--compiler-options=-fexceptions,-fPIC,-fno-omit-frame-pointer,-pthread,-fwrapv -DMX_COMPAT_64 -DMATLAB_DEFAULT_RELEASE=R2017b -DUSE_MEX_CMD --compiler-options=-D_GNU_SOURCE,-DMATLAB_MEX_FILE"

CXXFLAGS="--compiler-options=-fexceptions,-fPIC,-fno-omit-frame-pointer,-pthread,-fwrapv -DMX_COMPAT_64 -DMATLAB_DEFAULT_RELEASE=R2017b -DUSE_MEX_CMD --compiler-options=-D_GNU_SOURCE,-DMATLAB_MEX_FILE"

COPTIMFLAGS="-O2 -DNDEBUG"

CXXOPTIMFLAGS="-O2 -DNDEBUG"

CDEBUGFLAGS="-g"

CXXDEBUGFLAGS="-g"

LD="gcc"

LDXX="/usr/bin/g++"

LDFLAGS="-pthread -Wl,--no-undefined -shared -Wl,--as-needed -Wl,-rpath-link,/usr/lib64 -L"/usr/lib64" -Wl,-rpath-link,/usr/lib64 -L"/usr/lib64" -Wl,-rpath-link,/usr/local/MATLAB/R2020a/bin/glnxa64 -L"/usr/local/MATLAB/R2020a/bin/glnxa64" -Wl,-rpath-link,/usr/local/MATLAB/R2020a/extern/bin/glnxa64 -L"/usr/local/MATLAB/R2020a/extern/bin/glnxa64" -lMatlabDataArray -lmx -lmex -lmat -lm -lc -lstdc++ -lmwgpu /usr/lib64/libcudart.so.* -Wl,--version-script,"/usr/local/MATLAB/R2020a/extern/lib/glnxa64/mexFunction.map""

LDDEBUGFLAGS="-g"

CUDA_ROOT : /usr

LIBDEVICE : /usr/local/MATLAB/R2020a/bin/glnxa64

CUDA_LIBS : /usr/lib64

GCC : /usr/bin/g++

CPPLIB_DIR : /usr/lib/gcc/x86_64-linux-gnu/8/libstdc++.so

VER : 8

GCCDIR : /usr

GCC_TARGET : x86_64-linux-gnu

MATLABROOT : /usr/local/MATLAB/R2020a

ARCH : glnxa64

SRC : "/home/simon/temp/temp_mexcuda/test.cu";"/usr/local/MATLAB/R2020a/extern/version/c_mexapi_version.c"

OBJ : /tmp/mex_22266400607270_177416/test.o;/tmp/mex_22266400607270_177416/c_mexapi_version.o

OBJS : /tmp/mex_22266400607270_177416/test.o /tmp/mex_22266400607270_177416/c_mexapi_version.o

SRCROOT : /home/simon/temp/temp_mexcuda/test

DEF : /tmp/mex_22266400607270_177416/test.def

EXP : "test.exp"

LIB : "test.lib"

EXE : test.mexa64

ILK : "test.ilk"

MANIFEST : "test.mexa64.manifest"

TEMPNAME : test

EXEDIR :

EXENAME : test

NVCC_FLAGS :

OPTIM : -g

LINKOPTIM : -g

CMDLINE1_0 : /usr/bin/nvcc -c -DMX_COMPAT_64 -DMATLAB_DEFAULT_RELEASE=R2017b -DUSE_MEX_CMD --compiler-options=-D_GNU_SOURCE,-DMATLAB_MEX_FILE -I"/usr/local/MATLAB/R2020a/extern/include" -I"/usr/local/MATLAB/R2020a/simulink/include" -I"/usr/local/MATLAB/R2020a/toolbox/parallel/gpu/extern/include/" -I"/usr/include" -gencode=arch=compute_30,code=sm_30 -gencode=arch=compute_50,code=sm_50 -gencode=arch=compute_60,code=sm_60 -gencode=arch=compute_70,code=\"sm_70,compute_70\" -std=c++11 --dont-use-profile -ldir /usr/local/MATLAB/R2020a/bin/glnxa64 --compiler-options=-fexceptions,-fPIC,-fno-omit-frame-pointer,-pthread,-fwrapv -g "/home/simon/temp/temp_mexcuda/test.cu" -o /tmp/mex_22266400607270_177416/test.o

CMDLINE1_1 : /usr/bin/nvcc -c -DMX_COMPAT_64 -DMATLAB_DEFAULT_RELEASE=R2017b -DUSE_MEX_CMD --compiler-options=-D_GNU_SOURCE,-DMATLAB_MEX_FILE -I"/usr/local/MATLAB/R2020a/extern/include" -I"/usr/local/MATLAB/R2020a/simulink/include" -I"/usr/local/MATLAB/R2020a/toolbox/parallel/gpu/extern/include/" -I"/usr/include" -gencode=arch=compute_30,code=sm_30 -gencode=arch=compute_50,code=sm_50 -gencode=arch=compute_60,code=sm_60 -gencode=arch=compute_70,code=\"sm_70,compute_70\" -std=c++11 --dont-use-profile -ldir /usr/local/MATLAB/R2020a/bin/glnxa64 --compiler-options=-fexceptions,-fPIC,-fno-omit-frame-pointer,-pthread,-fwrapv -g "/usr/local/MATLAB/R2020a/extern/version/c_mexapi_version.c" -o /tmp/mex_22266400607270_177416/c_mexapi_version.o

CMDLINE3_0 : rm -f /tmp/mex_22266400607270_177416/test.o

CMDLINE3_1 : rm -f /tmp/mex_22266400607270_177416/c_mexapi_version.o

-------------------------------------------------------------------

Building with 'nvcc'.

/usr/bin/nvcc -c -DMX_COMPAT_64 -DMATLAB_DEFAULT_RELEASE=R2017b -DUSE_MEX_CMD --compiler-options=-D_GNU_SOURCE,-DMATLAB_MEX_FILE -I"/usr/local/MATLAB/R2020a/extern/include" -I"/usr/local/MATLAB/R2020a/simulink/include" -I"/usr/local/MATLAB/R2020a/toolbox/parallel/gpu/extern/include/" -I"/usr/include" -gencode=arch=compute_30,code=sm_30 -gencode=arch=compute_50,code=sm_50 -gencode=arch=compute_60,code=sm_60 -gencode=arch=compute_70,code=\"sm_70,compute_70\" -std=c++11 --dont-use-profile -ldir /usr/local/MATLAB/R2020a/bin/glnxa64 --compiler-options=-fexceptions,-fPIC,-fno-omit-frame-pointer,-pthread,-fwrapv -g "/home/simon/temp/temp_mexcuda/test.cu" -o /tmp/mex_22266400607270_177416/test.o

Error using mex

sh: 1: cicc: not found

Error in mexcuda (line 166)

[varargout{1:nargout}] = mex(mexArguments{:});

Error in untitled (line 1)

mexcuda -v -g test.cu

[Linda Koletsou Soulti]

Hello Simon,

I can see from the output you attached that you are using a local installation of the CUDA Toolkit. Could you please use the environment variable MW_NVCC_PATH to specify the location of the NVCC?

For example:

setenv('MW_NVCC_PATH','/usr/local/CUDA/bin')

Let us know if that works.

Cheers,

Linda

[Simon Zhang]

whereis nvcc

tells me nvcc is located at /usr/bin/nvcc. I tried:

setenv('MW_NVCC_PATH','/usr/bin')

and am getting the same error. This is the only place nvcc is located.

I originally installed nvcc for Ubuntu via

sudo apt-get install cuda 

my cuda toolkit version is 10.1 and gcc version is 8.4.0. I have no problems compiling with nvcc alone (without any MATLAB dependencies) and mex seems to work fine.

[Linda Koletsou Soulti]

Hello Simon,

have you also updated PATH as described in the post-installation actions in NVIDIA's installation guide:

https://docs.nvidia.com/cuda/cuda-installation-guide-linux/index.html#ubuntu-installation

Cheers,

Linda

[Simon Zhang]

yes, in fact I install CUDA 11 and updated PATH as in the postinstallation instructions:

https://docs.nvidia.com/cuda/cuda-installation-guide-linux/index.html#post-installation-actions

I am still getting the same error as before: cicc not found.

Answer 1

[Linda Koletsou Soulti]

Hello Simon,

better install the version of the CUDA toolkit consistent with the ToolkitVersion property of the gpuDevice object for the MATLAB version you use.

Can you also set MW_NVCC_PATH to point to the path you use to update PATH?

For example:

setenv('MW_NVCC_PATH','/usr/local/cuda-10.1/bin')

Cheers,

Linda

[Simon Zhang]

Hi Linda,

The solution for me was as follows:

$whereis nvidia-cuda-toolkit

/usr/lib/nvidia-cuda-toolkit

in matlab:

>> setenv('MW_NVCC_PATH', '/usr/lib/nvidia-cuda-toolkit/bin')
>> mexcuda helloworld.cu
Building with 'nvcc'.
MEX completed successfully.