Merge pull request #20107 from Flamefire/20240313135157_new_pr_CUDA1240

{lib}[GCCcore/13.2.0] CUDA v12.4.0, NCCL v2.20.5, magma v2.7.2, ... w/ CUDA 12.4.0

easybuild/easyconfigs/c/CUDA/CUDA-12.4.0.eb

@@ -0,0 +1,24 @@
+name = 'CUDA'
+version = '12.4.0'
+local_nv_version = '550.54.14'
+
+homepage = 'https://developer.nvidia.com/cuda-toolkit'
+description = """CUDA (formerly Compute Unified Device Architecture) is a parallel
+ computing platform and programming model created by NVIDIA and implemented by the
+ graphics processing units (GPUs) that they produce. CUDA gives developers access
+ to the virtual instruction set and memory of the parallel computational elements in CUDA GPUs."""
+
+toolchain = SYSTEM
+
+source_urls = ['https://developer.download.nvidia.com/compute/cuda/%(version)s/local_installers/']
+sources = ['cuda_%%(version)s_%s_linux%%(cudaarch)s.run' % local_nv_version]
+checksums = [{
+    'cuda_%%(version)s_%s_linux.run' % local_nv_version:
+        'e6a842f4eca9490575cdb68b6b1bb78d47b95a897de48dee292c431892e57d17',
+    'cuda_%%(version)s_%s_linux_ppc64le.run' % local_nv_version:
+        'ef9a712daccf2805b4422f2301ff0eaa5c3ad41ef5d64b8626773bce7d1f41fe',
+    'cuda_%%(version)s_%s_linux_sbsa.run' % local_nv_version:
+        'b12bfe6c36d32ecf009a6efb0024325c5fc389fca1143f5f377ae2555936e803'
+}]
+
+moduleclass = 'system'

easybuild/easyconfigs/m/magma/magma-2.7.2-foss-2023b-CUDA-12.4.0.eb

@@ -0,0 +1,44 @@
+easyblock = 'CMakeMake'
+
+name = 'magma'
+version = '2.7.2'
+versionsuffix = '-CUDA-%(cudaver)s'
+
+homepage = 'https://icl.cs.utk.edu/magma/'
+description = """The MAGMA project aims to develop a dense linear algebra library similar to
+ LAPACK but for heterogeneous/hybrid architectures, starting with current Multicore+GPU systems."""
+
+toolchain = {'name': 'foss', 'version': '2023b'}
+toolchainopts = {'pic': True, 'openmp': True}
+
+source_urls = ['https://icl.cs.utk.edu/projectsfiles/%(name)s/downloads/']
+sources = [SOURCE_TAR_GZ]
+patches = ['%(name)s-2.7.2_allow-all-sms.patch']
+checksums = [
+    {'magma-2.7.2.tar.gz': '729bc1a70e518a7422fe7a3a54537a4741035a77be3349f66eac5c362576d560'},
+    {'magma-2.7.2_allow-all-sms.patch': '2996485844d7e199ea57eaebf59903d3514a03536f82b655ce30de480e28b828'},
+]
+
+builddependencies = [
+    ('CMake', '3.27.6'),
+]
+dependencies = [
+    ('CUDA', '12.4.0', '', SYSTEM),
+    ('UCX-CUDA', '1.15.0', versionsuffix),
+]
+
+# default CUDA compute capabilities to use (override via --cuda-compute-capabilities)
+cuda_compute_capabilities = ['5.0', '6.0', '7.0', '7.5', '8.0', '8.6', '9.0']
+
+# make sure both static and shared libs are built
+configopts = [
+    '-DBUILD_SHARED_LIBS=%s \
+     -DGPU_TARGET="%%(cuda_sm_space_sep)s" ' % local_shared for local_shared in ('ON', 'OFF')
+]
+
+sanity_check_paths = {
+    'files': ['lib/libmagma.%s' % SHLIB_EXT, 'lib/libmagma.a'],
+    'dirs': ['include'],
+}
+
+moduleclass = 'math'

easybuild/easyconfigs/n/NCCL/NCCL-2.20.5-GCCcore-13.2.0-CUDA-12.4.0.eb

@@ -0,0 +1,26 @@
+name = 'NCCL'
+version = '2.20.5'
+versionsuffix = '-CUDA-%(cudaver)s'
+
+homepage = 'https://developer.nvidia.com/nccl'
+description = """The NVIDIA Collective Communications Library (NCCL) implements multi-GPU and multi-node collective
+communication primitives that are performance optimized for NVIDIA GPUs."""
+
+toolchain = {'name': 'GCCcore', 'version': '13.2.0'}
+
+github_account = 'NVIDIA'
+source_urls = [GITHUB_SOURCE]
+sources = ['v%(version)s-1.tar.gz']
+checksums = ['d11ad65c1df3cbe4447eaddceec71569f5c0497e27b3b8369cf79f18d2b2ad8c']
+
+builddependencies = [('binutils', '2.40')]
+
+dependencies = [
+    ('CUDA', '12.4.0', '', SYSTEM),
+    ('UCX-CUDA', '1.15.0', versionsuffix),
+]
+
+# default CUDA compute capabilities to use (override via --cuda-compute-capabilities)
+cuda_compute_capabilities = ['5.0', '6.0', '7.0', '7.5', '8.0', '8.6', '9.0']
+
+moduleclass = 'lib'

easybuild/easyconfigs/u/UCX-CUDA/UCX-CUDA-1.15.0-GCCcore-13.2.0-CUDA-12.4.0.eb

@@ -0,0 +1,41 @@
+easyblock = 'EB_UCX_Plugins'
+
+name = 'UCX-CUDA'
+version = '1.15.0'
+versionsuffix = '-CUDA-%(cudaver)s'
+
+homepage = 'http://www.openucx.org/'
+description = """Unified Communication X
+An open-source production grade communication framework for data centric
+and high-performance applications
+
+This module adds the UCX CUDA support.
+"""
+
+toolchain = {'name': 'GCCcore', 'version': '13.2.0'}
+toolchainopts = {'pic': True}
+
+source_urls = ['https://github.com/openucx/ucx/releases/download/v%(version)s']
+sources = [{'filename': 'ucx-%(version)s.tar.gz', 'alt_location': 'UCX'}]
+patches = ['%(name)s-1.11.0_link_against_existing_UCX_libs.patch']
+checksums = [
+    {'ucx-1.15.0.tar.gz': '4b202087076bc1c98f9249144f0c277a8ea88ad4ca6f404f94baa9cb3aebda6d'},
+    {'UCX-CUDA-1.11.0_link_against_existing_UCX_libs.patch':
+     '457187fa020e526609ba91e7750c9941d57bd57d60d6eed317b40ad8824aca93'},
+]
+
+builddependencies = [
+    ('binutils', '2.40'),
+    ('Autotools', '20220317'),
+    ('pkgconf', '2.0.3'),
+]
+
+dependencies = [
+    ('zlib', '1.2.13'),
+    ('UCX', version),
+    ('CUDA', '12.4.0', '', SYSTEM),
+    ('GDRCopy', '2.4'),
+]
+
+
+moduleclass = 'lib'