CMake Error at CMakeLists.txt:379 (include): include could not find requested file: .../magma/CMake.src.cuda

[jeffhammond]

In response to #21, I am trying to build the git version (2716e3f), but it fails as follows.

~/magma/build$ git clean -dfx ; cmake .. -DCMAKE_CUDA_ARCHITECTURES=80
warning: failed to remove ./: Invalid argument
-- The C compiler identification is GNU 11.4.0
-- The CXX compiler identification is GNU 11.4.0
-- The Fortran compiler identification is GNU 11.4.0
-- Detecting C compiler ABI info
-- Detecting C compiler ABI info - done
-- Check for working C compiler: /usr/bin/cc - skipped
-- Detecting C compile features
-- Detecting C compile features - done
-- Detecting CXX compiler ABI info
-- Detecting CXX compiler ABI info - done
-- Check for working CXX compiler: /usr/bin/c++ - skipped
-- Detecting CXX compile features
-- Detecting CXX compile features - done
-- Detecting Fortran compiler ABI info
-- Detecting Fortran compiler ABI info - done
-- Check for working Fortran compiler: /usr/bin/gfortran - skipped
-- MAGMA requires one enabled backend!
-- Building CUDA backend
-- Performing Test COMPILER_SUPPORTS_CXX11
-- Performing Test COMPILER_SUPPORTS_CXX11 - Success
-- Performing Test COMPILER_SUPPORTS_CXX0X
-- Performing Test COMPILER_SUPPORTS_CXX0X - Success
-- Performing Test COMPILER_SUPPORTS_FPIC
-- Performing Test COMPILER_SUPPORTS_FPIC - Success
-- Performing Test COMPILER_SUPPORTS_C99
-- Performing Test COMPILER_SUPPORTS_C99 - Success
-- Detecting Fortran/C Interface
-- Detecting Fortran/C Interface - Found GLOBAL and MODULE mangling
-- Found OpenMP_C: -fopenmp (found version "4.5")
-- Found OpenMP_CXX: -fopenmp (found version "4.5")
-- Found OpenMP_Fortran: -fopenmp (found version "4.5")
-- Found OpenMP: TRUE (found version "4.5")
-- Found OpenMP
--     OpenMP_C_FLAGS   -fopenmp
--     OpenMP_CXX_FLAGS -fopenmp
-- The CUDA compiler identification is NVIDIA 12.6.85
-- Detecting CUDA compiler ABI info
-- Detecting CUDA compiler ABI info - done
-- Check for working CUDA compiler: /usr/local/cuda/bin/nvcc - skipped
-- Detecting CUDA compile features
-- Detecting CUDA compile features - done
-- Found CUDAToolkit: /usr/local/cuda/targets/x86_64-linux/include (found version "12.6.85")
-- Performing Test CMAKE_HAVE_LIBC_PTHREAD
-- Performing Test CMAKE_HAVE_LIBC_PTHREAD - Success
-- Found Threads: TRUE
-- Found CUDA 
--     CUDA_CUDART_LIBRARY: CUDA::cudart
--     Compile for CMAKE_CUDA_ARCHITECTURES=80
-- Define -DMAGMA_HAVE_CUDA -DMAGMA_CUDA_ARCH_MIN=800
-- Searching for BLAS and LAPACK. To override, set LAPACK_LIBRARIES using ccmake.
-- Looking for Fortran sgemm
-- Looking for Fortran sgemm - not found
-- Looking for Fortran sgemm
-- Looking for Fortran sgemm - found
-- Found BLAS: /usr/lib/x86_64-linux-gnu/libopenblas.so
-- Looking for Fortran cheev
-- Looking for Fortran cheev - found
-- Found LAPACK: /usr/lib/x86_64-linux-gnu/libopenblas.so;-lm;-ldl
CMake Error at CMakeLists.txt:379 (include):
  include could not find requested file:

    ~/magma/CMake.src.cuda


-- Looking for sys/types.h
-- Looking for sys/types.h - found
-- Looking for stdint.h
-- Looking for stdint.h - found
-- Looking for stddef.h
-- Looking for stddef.h - found
-- Check size of void*
-- Check size of void* - done
CMake Deprecation Warning at CMakeLists.txt:657 (cmake_policy):
  The OLD behavior for policy CMP0037 will be removed from a future version
  of CMake.

  The cmake-policies(7) manual explains that the OLD behaviors of all
  policies are deprecated and that a policy should be set to OLD only under
  specific short-term circumstances.  Projects should be ported to the NEW
  behavior and not rely on setting a policy to OLD.


-- pkgconfig lib/pkgconfig/magma.pc
-- Flags
--     CMAKE_INSTALL_PREFIX:  /usr/local/magma
--     CFLAGS:                 -std=c99 -fopenmp -Wall -Wno-unused-function
--     CXXFLAGS:               -std=c++11 -fopenmp -Wall -Wno-unused-function
--     NVCCFLAGS:             
--     FFLAGS:                 -Dmagma_devptr_t="integer(kind=8)"
--     LIBS:                   /usr/lib/x86_64-linux-gnu/libopenblas.so -lm -ldl -L/usr/local/cuda/lib64 -lcudart -lcublas -lcusparse
--     blas_fix:                (MacOS Accelerate only)
--     LAPACK_LIBRARIES:      /usr/lib/x86_64-linux-gnu/libopenblas.so;-lm;-ldl
--     INCLUDE_DIRECTORIES:   -I/usr/local/cuda/targets/x86_64-linux/include -I~/magma/build/include -I~/magma/include -I~/magma/control -I~/magma/magmablas -I~/magma/sparse/include -I~/magma/sparse/control -I~/cuEST/magma/testing 
--     CUDA_CUDART_LIBRARY:   CUDA::cudart
--     CUDA_CUBLAS_LIBRARIES: CUDA::cublas
--     CUDA_cusparse_LIBRARY: CUDA::cusparse
--     Fortran modules:       
-- Configuring incomplete, errors occurred!

[jeffhammond]

I don't know if CMake is supported, because I had to generate this file using the Makefile build system, which I thought was an alternative to CMake. Anyways, I am building the non-CMake way now.

[nbeams]

The CMake support is indeed a bit weird -- you still have to do the code generation step first, as mentioned under Quick start (CMake) in the README. Did you do that and still see this error? (Edit: I didn't read your second comment closely enough. I see you did generate it!) Regardless, doing the Makefile build is probably just as easy for MAGMA.

[nbeams]

So just to confirm, did you try the CMake configure again after the make generate step, or go straight to building with the Makefile?

[jeffhammond]

i just used the Makefile. i dislike CMake but it seems to be the preferred way for a lot of projects, so i started there. i am very happy to avoid it and am glad that UTK still invests in pure Make.

[nbeams]

We do generally recommend the Makefile build. I get the impression that the CMake option is more commonly used by other projects who use CMake and want to include MAGMA. But the non-standard CMake use is a bit of a problem.