Hi! The suitesparse library has been uptated to version 4.0.0: Major changes: added GPU support for CHOLMOD. Renamed UFconfig to SuiteSparse_config. In particular the cholmod library now is able to use the CUBLAS on NVIDIA GPUs and cuda should be a new USE flag for this package. As I do not unterstand enough of autotools based ebuilds I request a version bump here. However, I would be glad to help testing possible new ebuilds. Greets Janick
now in cvs. (cuda part is still to be tested)
Thanks for updating all the suitesparse packages! I wanted to test the cholmod cuda integration but I'm unable to compile sci-libs/cholmod with the cuda use flag: checking for SUITESPARSECONFIG... yes checking SuiteSparse_config.h usability... yes checking SuiteSparse_config.h presence... yes checking for SuiteSparse_config.h... yes checking for SuiteSparse_malloc in -lsuitesparseconfig... yes checking whether to build user guide... no checking for CUDA... no configure: error: Package requirements (cuda) were not met: No package 'cuda' found Consider adjusting the PKG_CONFIG_PATH environment variable if you installed software in a non-standard prefix. Alternatively, you may set the environment variables CUDA_CFLAGS and CUDA_LIBS to avoid the need to call pkg-config. See the pkg-config man page for more details. !!! Please attach the following file when seeking support: !!! /var/tmp/portage/sci-libs/cholmod-2.0.1/work/cholmod-2.0.1_build/config.log * ERROR: sci-libs/cholmod-2.0.1 failed (configure phase): * econf failed * * Call stack: * ebuild.sh, line 93: Called src_configure * environment, line 3079: Called autotools-utils_src_configure * environment, line 612: Called econf '--docdir=/usr/share/doc/cholmod-2.0.1' '--enable-shared' '--disable-static' '--with-blas=-Wl,--start-group -Wl,--end-group -fopenmp -L/opt/intel/composerxe-2013.0.079/mkl/lib/intel64 -lmkl_gf_lp64 -lmkl_gnu_thread -lmkl_core -lpthread -lm -lgomp ' '--with-lapack=-Wl,--start-group -Wl,--end-group -fopenmp -L/opt/intel/composerxe-2013.0.079/mkl/lib/intel64 -lmkl_gf_lp64 -lmkl_gnu_thread -lmkl_core -lpthread -lm -lgomp ' '--with-cuda' '--without-doc' '--with-modify' '--with-matrixops' '--with-partition' '--with-supernodal' * phase-helpers.sh, line 501: Called die * The specific snippet of code: * die "econf failed" * * If you need support, post the output of `emerge --info '=sci-libs/cholmod-2.0.1'`, * the complete build log and the output of `emerge -pqv '=sci-libs/cholmod-2.0.1'`. * The complete build log is located at '/var/tmp/portage/sci-libs/cholmod-2.0.1/temp/build.log'. * The ebuild environment file is located at '/var/tmp/portage/sci-libs/cholmod-2.0.1/temp/environment'. * Working directory: '/var/tmp/portage/sci-libs/cholmod-2.0.1/work/cholmod-2.0.1_build' * S: '/var/tmp/portage/sci-libs/cholmod-2.0.1/work/cholmod-2.0.1' >>> Failed to emerge sci-libs/cholmod-2.0.1, Log file: >>> '/var/tmp/portage/sci-libs/cholmod-2.0.1/temp/build.log' I have nvidia-cuda-toolkit-4.2 installed. Is this package supposed to provide a 'cuda' pkgconfig file?
the version bump is done. you are now hitting bug #442858, which is another issue. i'm working on it without any graphic card to test, so the road to a proper fix might be bumpy.