You signed in with another tab or window. Reload to refresh your session.You signed out in another tab or window. Reload to refresh your session.You switched accounts on another tab or window. Reload to refresh your session.Dismiss alert
@ndellingwood
I felt adventurous and tried to do a tpl build on kokkos-dev with clang, here is what my configure script looks like:
#!/usr/bin/bash
module load sems-env
module load kokkos-env
module load sems-cmake/3.12.2
module load sems-clang/7.0.1
module load sems-cuda/9.2
KOKKOS_PATH="/ascldap/users/lberge/kokkos"
KOKKOS_DEVICES="Serial,Cuda"
KOKKOS_ARCHS="Kepler35"
KOKKOSKERNELS_PATH="/ascldap/users/lberge/kk_lucbv"
KOKKOSKERNELS_LAYOUTS="LayoutLeft" # the layout types to instantiate.
KOKKOSKERNELS_SCALARS="double" # the scalar types to instantiate =double,float...
KOKKOSKERNELS_ORDINALS="int" # ordinal types to instantiate
KOKKOSKERNELS_OFFSETS="int" # offset types to instantiate
KOKKOSKERNELS_OPTIONS="eti-only" # options for kokkoskernels
KOKKOSKERNELS_TPLS="blas,cublas,cusparse" # tpls to be used as backend when possible
KOKKOSKERNELS_INSTALL="/ascldap/users/lberge/kk_lucbv_build/clang_7.0.1_cuda_9.2_tpls"
KOKKOS_INSTALL="${KOKKOSKERNELS_INSTALL}/kokkos-install"
CUDA_PATH="${SEMS_CUDA_ROOT}"
CUDA_OPTIONS=""
rm -rf ${KOKKOS_INSTALL}/CMake*
rm -rf ${KOKKOSKERNELS_INSTALL}/CMake*
cd ${KOKKOSKERNELS_INSTALL}
${KOKKOSKERNELS_PATH}/cm_generate_makefile.bash \
--kokkos-path="${KOKKOS_PATH}" \
--kokkoskernels-path="${KOKKOSKERNELS_PATH}" \
--prefix="${KOKKOSKERNELS_INSTALL}" \
--with-devices="${KOKKOS_DEVICES}" \
--arch="${KOKKOS_ARCHS}" \
--compiler="${SEMS_CLANG_ROOT}/bin/clang++" \
--cxxflags="-O3 -Wall -Wshadow -pedantic -Werror -Wsign-compare -Wtype-limits -Wuninitialized -std=c++11" \
--cxxstandard="11" \
--ldflags="" \
--with-cuda="${CUDA_PATH}" \
--with-scalars="${KOKKOSKERNELS_SCALARS}" \
--with-ordinals="${KOKKOSKERNELS_ORDINALS}" \
--with-offsets="${KOKKOSKERNELS_OFFSETS}" \
--with-layouts="${KOKKOSKERNELS_LAYOUTS}" \
--with-tpls="${KOKKOSKERNELS_TPLS}" \
--no-examples \
--with-options=enable_large_mem_tests |& tee configure_clang_cuda_tpls.log
It actually mostly works but there are a few issues with sptrsv and spgemm, I will start working on this a bit... it would be nice for that build to work so we can increase the coverage we have with "cm_test_all_sandia --spot-check-tpls" make_clang_cuda_tpls.log
The text was updated successfully, but these errors were encountered:
@lucbv errors look pretty straightforward (remove unused vars in sptrsv, likely reorder of initialization or listing of member variables in the spgemm handle?), let me know if you want me to put in the sptrsv fix. Once this is cleaned up, we can add this to cm_test_all_sandia for kokkos-dev-2 and I'll set up a nightly build there.
Yeah, I'm looking into a quick clean-up of the easy things and if I stumble upon something hard I might ask you about it, but that does not seem to likely, hopefully I can make a PR for that quickly.
@ndellingwood
I felt adventurous and tried to do a tpl build on kokkos-dev with clang, here is what my configure script looks like:
It actually mostly works but there are a few issues with
sptrsv
andspgemm
, I will start working on this a bit... it would be nice for that build to work so we can increase the coverage we have with "cm_test_all_sandia --spot-check-tpls"make_clang_cuda_tpls.log
The text was updated successfully, but these errors were encountered: