diff --git a/n/NVHPC/NVHPC-24.11.eb b/n/NVHPC/NVHPC-24.11.eb new file mode 100644 index 00000000..e646844d --- /dev/null +++ b/n/NVHPC/NVHPC-24.11.eb @@ -0,0 +1,41 @@ +# IT4Innovations +# PH 2024 + +name = 'NVHPC' +version = '24.11' + +homepage = 'https://developer.nvidia.com/hpc-sdk/' +description = """C, C++ and Fortran compilers included with the NVIDIA HPC SDK (previously: PGI)""" + +toolchain = SYSTEM + +local_tarball_tmpl = 'nvhpc_2024_%%(version_major)s%%(version_minor)s_Linux_%s_cuda_multi.tar.gz' +# By downloading, you accept the HPC SDK Software License Agreement +# https://docs.nvidia.com/hpc-sdk/eula/index.html +accept_eula = True +source_urls = ['https://developer.download.nvidia.com/hpc-sdk/%(version)s/'] +sources = [local_tarball_tmpl % '%(arch)s'] +checksums = [ + { + local_tarball_tmpl % 'aarch64': + 'f2f64e5dec5e90dad5e12a31a992172b0aa19abf872ef1c54a1a437c7008eefb', +# local_tarball_tmpl % 'ppc64le': +# 'e7330eb35e23dcd9b0b3bedc67c0d5443c4fd76b59caa894a76ecb0d17f71f43', + local_tarball_tmpl % 'x86_64': + '0c27d66ed0e2d3007d30ac904922a9abf96475197dc0f4dcc6316d235a1dc0c3', + } +] + +default_cuda_version = "11.0" + +local_gccver = '13.3.0' +dependencies = [ + ('GCCcore', local_gccver), + ('binutils', '2.42', '', ('GCCcore', local_gccver)), + # This is necessary to avoid cases where just libnuma.so.1 is present in the system and -lnuma fails + ('numactl', '2.0.18', '', ('GCCcore', local_gccver)), +] + +module_add_cuda = False + +moduleclass = 'compiler' diff --git a/o/OpenMPI/OpenMPI-4.1.6-NVHPC-24.11.eb b/o/OpenMPI/OpenMPI-4.1.6-NVHPC-24.11.eb new file mode 100644 index 00000000..dc5d6452 --- /dev/null +++ b/o/OpenMPI/OpenMPI-4.1.6-NVHPC-24.11.eb @@ -0,0 +1,112 @@ +# IT4Innovations +# PH 2024 + +name = 'OpenMPI' +version = '4.1.6' + +homepage = 'https://www.open-mpi.org/' +description = """The Open MPI Project is an open source MPI-3 implementation.""" + +toolchain = {'name': 'NVHPC', 'version': '24.11'} + +source_urls = ['https://www.open-mpi.org/software/ompi/v%(version_major_minor)s/downloads'] +sources = [SOURCELOWER_TAR_BZ2] +patches = [ + 'OpenMPI-4.1.1_build-with-internal-cuda-header.patch', + 'OpenMPI-4.1.1_opal-datatype-cuda-performance.patch', + 'OpenMPI-4.1.6_disable_opal_path_nfs_test.patch', +] +checksums = [ + 'f740994485516deb63b5311af122c265179f5328a0d857a567b85db00b11e415', # openmpi-4.1.6.tar.bz2 + # OpenMPI-4.1.1_build-with-internal-cuda-header.patch + '63eac52736bdf7644c480362440a7f1f0ae7c7cae47b7565f5635c41793f8c83', + # OpenMPI-4.1.1_opal-datatype-cuda-performance.patch + 'b767c7166cf0b32906132d58de5439c735193c9fd09ec3c5c11db8d5fa68750e', + # OpenMPI-4.1.6_disable_opal_path_nfs_test.patch + 'f9647cef5b3e0469573bc93936655acfc9a2a925c813108115e0b38264e1114d', +] + +builddependencies = [ + ('pkgconf', '2.2.0'), + ('Perl', '5.38.2'), + ('Autotools', '20231222'), +] + +dependencies = [ + ('zlib', '1.3.1'), + ('hwloc', '2.10.0'), + ('libevent', '2.1.12'), + ('UCX', '1.16.0'), + ('libfabric', '1.21.0'), + ('PMIx', '5.0.2'), + ('UCC', '1.3.0'), +] + +# Update configure to include changes from the "internal-cuda" patch +# by running a subset of autogen.pl sufficient to achieve this +# without doing the full, long-running regeneration. +preconfigopts = ' && '.join([ + 'cd config', + 'autom4te --language=m4sh opal_get_version.m4sh -o opal_get_version.sh', + 'cd ..', + 'autoconf', + 'autoheader', + 'aclocal', + 'automake', + '' +]) + +# IT4I-specific settings + +#configopts += '--enable-shared ' +configopts = ' --enable-mpi-thread-multiple' +configopts += ' --with-verbs' +configopts += ' --enable-mpirun-prefix-by-default' +configopts += ' --with-hwloc=$EBROOTHWLOC' # hwloc support +configopts += ' --with-slurm' # Enable slurm +configopts += ' --enable-mpi-cxx' # Enable building the C++ MPI bindings +configopts += ' --with-ucx=$EBROOTUCX' + +import os +if os.environ.get("CLUSTERNAME") in ["BARBORA"]: + configopts += ' --with-knem=/opt/knem-1.1.4.90mlnx1' +if os.environ.get("CLUSTERNAME") in ["KAROLINA"]: + configopts += ' --with-knem=/opt/knem-1.1.4.90mlnx3' + +osdependencies = [('libibverbs-dev', 'libibverbs-devel', 'rdma-core-devel')] + +postinstallcmds = [ + 'echo "# By default, for Open MPI 4.0 and later, infiniband ports on a device are not used by default." >> %(installdir)s/etc/openmpi-mca-params.conf', + 'echo "btl_openib_allow_ib = true" >> %(installdir)s/etc/openmpi-mca-params.conf', +] + +local_libs = ["mpi_mpifh", "mpi", "ompitrace", "open-pal", "open-rte"] +sanity_check_paths = { + 'files': [ + "bin/%s" % + binfile for binfile in [ + "ompi_info", "opal_wrapper", "orterun"]] + [ + "lib/lib%s.%s" % + (libfile, SHLIB_EXT) for libfile in local_libs] + [ + "include/%s.h" % + x for x in [ + "mpi-ext", "mpif-config", "mpif", "mpi", "mpi_portable_platform"]], 'dirs': [], } + +if os.environ.get("CLUSTERNAME") in ["BARBORA"]: + modextravars = {'OMPI_MCA_btl_openib_if_include': 'mlx5_0', + 'OMPI_MCA_btl_tcp_if_include': '10.33.4.0/24', + 'OMPI_MCA_orte_base_help_aggregate': '0', + 'SLURM_MPI_TYPE': 'pmix_v4', + } +elif os.environ.get("CLUSTERNAME") in ["KAROLINA"]: + modextravars = {'OMPI_MCA_btl_openib_if_include': 'mlx5_0', + 'OMPI_MCA_orte_base_help_aggregate': '0', + 'SLURM_MPI_TYPE': 'pmix_v4', + } +else: + modextravars = {'OMPI_MCA_btl_openib_if_include': 'mlx4_0', + 'OMPI_MCA_oob_tcp_if_include': '10.0.0.0/8', + 'SLURM_MPI_TYPE': 'pmix_v4', + } + +moduleclass = 'mpi' diff --git a/o/OpenMPI/OpenMPI-4.1.6_disable_opal_path_nfs_test.patch b/o/OpenMPI/OpenMPI-4.1.6_disable_opal_path_nfs_test.patch new file mode 100644 index 00000000..73c2dc47 --- /dev/null +++ b/o/OpenMPI/OpenMPI-4.1.6_disable_opal_path_nfs_test.patch @@ -0,0 +1,28 @@ +diff -ruN openmpi-4.1.6.orig/test/util/Makefile.am openmpi-4.1.6/test/util/Makefile.am +--- openmpi-4.1.6.orig/test/util/Makefile.am 2023-09-30 17:23:17.000000000 +0200 ++++ openmpi-4.1.6/test/util/Makefile.am 2024-12-12 15:31:51.665278259 +0100 +@@ -38,7 +38,6 @@ + + check_PROGRAMS = \ + opal_bit_ops \ +- opal_path_nfs \ + bipartite_graph + + TESTS = \ +@@ -79,11 +78,11 @@ + $(top_builddir)/test/support/libsupport.a + opal_bit_ops_DEPENDENCIES = $(opal_path_nfs_LDADD) + +-opal_path_nfs_SOURCES = opal_path_nfs.c +-opal_path_nfs_LDADD = \ +- $(top_builddir)/opal/lib@OPAL_LIB_PREFIX@open-pal.la \ +- $(top_builddir)/test/support/libsupport.a +-opal_path_nfs_DEPENDENCIES = $(opal_path_nfs_LDADD) ++#opal_path_nfs_SOURCES = opal_path_nfs.c ++#opal_path_nfs_LDADD = \ ++# $(top_builddir)/opal/lib@OPAL_LIB_PREFIX@open-pal.la \ ++# $(top_builddir)/test/support/libsupport.a ++#opal_path_nfs_DEPENDENCIES = $(opal_path_nfs_LDADD) + + #opal_os_path_SOURCES = opal_os_path.c + #opal_os_path_LDADD = \