- search rocm hipblas with the find_package config method
 - use the official target name from rocm: roc::hipblas
 - export CMAKE_PREFIX_PATH in CHAMELEONConfig.cmake to let users find rocm hipblas the same way

gitlab-ci-initial-cache.cmake: CMAKE_C_FLAGS must be set as a cache variable to be used, and thus activate -Werror

cmake: fix this if when MPI_C_LIBRARIES is made of a list of several libraries. This fixes cmake with nmad/madmpi.

pkg-config: remove the cuda.pc and mpi.pc required dependencies, these files are not common and not installed usually
ChameleonPkgConfig.cmake: find the precise mpi .pc file to add in chameleon.pc required packages

Abel Calluaud authored 1 year ago and Mathieu Faverge committed 1 year ago

Remove the use of CHAMELEON_PRECISIONS that is an upper duplicate of CHAMELEON_PRECISION provided by cmake_morse (Fix #124)

PRUVOST Florent authored 1 year ago and Mathieu Faverge committed 1 year ago

Change name of internal libraries cudablas and hipblas to gpucublas and gpuhipblas to avoid confusion with external libraries that can have the same name

Force to use Netlib blas/lapack (instead of openblas also available and use by default in the docker image) for all builds

CMake modifications to avoid cmake projects dependents to rediscover all our dependencies at configure time