zhangyiss/spack
1
0
Fork 0
Code Issues Pull Requests Actions Packages Projects Releases Wiki Activity

Magma: add ROCm support and v2.6.0 (#24663)

Browse Source
This commit is contained in:
G-Ragghianti 2021-07-06 15:04:37 -04:00 committed by GitHub
parent ea2d4b05bc
commit c12dc1a5de
No known key found for this signature in database
GPG Key ID: 4AEE18F83AFDEB23
1 changed files with 32 additions and 4 deletions
Show Stats Download Patch File Download Diff File Expand all files Collapse all files

36
var/spack/repos/builtin/packages/magma/package.py
View File

@ -7,18 +7,21 @@
from spack import *
class Magma(CMakePackage, CudaPackage):
class Magma(CMakePackage, CudaPackage, ROCmPackage):
"""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.
"""
homepage = "http://icl.cs.utk.edu/magma/"
git = 'https://bitbucket.org/icl/magma'
url = "http://icl.cs.utk.edu/projectsfiles/magma/downloads/magma-2.2.0.tar.gz"
maintainers = ['stomov', 'luszczek', 'G-Ragghianti']
test_requires_compiler = True
version('master', branch='master')
version('2.6.0', sha256='50cdd384f44f06a34469e7125f8b2ffae13c1975d373c3f1510d91be2b7638ec')
version('2.5.4', sha256='7734fb417ae0c367b418dea15096aef2e278a423e527c615aab47f0683683b67')
version('2.5.3', sha256='c602d269a9f9a3df28f6a4f593be819abb12ed3fa413bba1ff8183de721c5ef6')
version('2.5.2', sha256='065feb85558f9dd6f4cc4db36ac633a3f787827fc832d0b578a049a43a195620')
@ -38,10 +41,14 @@ class Magma(CMakePackage, CudaPackage):
depends_on('blas')
depends_on('lapack')
depends_on('cuda@8:', when='@2.5.1:') # See PR #14471
depends_on('cuda@8:', when='@2.5.1: +cuda') # See PR #14471
depends_on('hipblas', when='+rocm')
depends_on('hipsparse', when='+rocm')
conflicts('~cuda', msg='Magma requires cuda')
conflicts('cuda_arch=none',
conflicts('~cuda', when='~rocm', msg='Either CUDA or HIP support must be enabled')
conflicts('+rocm', when='+cuda', msg='CUDA must be disabled to support HIP (ROCm)')
conflicts('+rocm', when='@:2.5.4', msg='HIP support starts in version 2.6.0')
conflicts('cuda_arch=none', when='+cuda',
msg='Please indicate a CUDA arch value or values')
# currently not compatible with CUDA-11
@ -57,6 +64,21 @@ class Magma(CMakePackage, CudaPackage):
patch('cmake-W.patch', when='@2.5.0:%nvhpc')
patch('sm_37.patch', when='@2.5.4 cuda_arch=37')
@run_before('cmake')
def generate_cuda(self):
if '@master' in self.spec:
backend = 'cuda'
cuda_arch = self.spec.variants['cuda_arch'].value
gpu_target = ' '.join('sm_{0}'.format(i) for i in cuda_arch)
if '+rocm' in self.spec:
backend = 'hip'
gpu_target = self.spec.variants['amdgpu_target'].value
with open('make.inc', 'w') as inc:
inc.write('FORT = true\n')
inc.write('GPU_TARGET = %s\n' % gpu_target)
inc.write('BACKEND = %s\n' % backend)
make('generate')
def cmake_args(self):
spec = self.spec
options = []
@ -104,6 +126,12 @@ def cmake_args(self):
if spec.compiler.name in ['xl', 'xl_r']:
options.extend(['-DCMAKE_DISABLE_FIND_PACKAGE_OpenMP=TRUE'])
if '+rocm' in spec:
options.extend(['-DMAGMA_ENABLE_HIP=ON'])
options.extend(['-DCMAKE_CXX_COMPILER=hipcc'])
else:
options.extend(['-DMAGMA_ENABLE_CUDA=ON'])
return options
@run_after('install')