NCCL-2.18.3-GCCcore-12.3.0-CUDA-12.1.1.eb

name = 'NCCL'
version = '2.18.3'
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': '12.3.0'}

github_account = 'NVIDIA'
source_urls = [GITHUB_SOURCE]
sources = ['v%(version)s-1.tar.gz']
patches = [
    'NCCL-2.16.2_fix-cpuid.patch',
    'NCCL-2.18.3_fix-cudaMemcpyAsync.patch',
]
checksums = [
    ('6477d83c9edbb34a0ebce6d751a1b32962bc6415d75d04972b676c6894ceaef9',
     'b4f5d7d9eea2c12e32e7a06fe138b2cfc75969c6d5c473aa6f819a792db2fc96'),
    {'NCCL-2.16.2_fix-cpuid.patch': '0459ecadcd32b2a7a000a2ce4f675afba908b2c0afabafde585330ff4f83e277'},
    {'NCCL-2.18.3_fix-cudaMemcpyAsync.patch': '7dc8d0d1b78e4f8acefbc400860f47432ef67c225b50d73c732999c23483de90'},
]

builddependencies = [('binutils', '2.40')]

dependencies = [
    ('CUDA', '12.1.1', '', SYSTEM),
    ('UCX-CUDA', '1.14.1', 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'