diff options
Diffstat (limited to 'var')
| -rw-r--r-- | var/spack/repos/builtin/packages/tioga/package.py | 53 |
1 files changed, 36 insertions, 17 deletions
diff --git a/var/spack/repos/builtin/packages/tioga/package.py b/var/spack/repos/builtin/packages/tioga/package.py index e436bfbc6b..7957bf64e9 100644 --- a/var/spack/repos/builtin/packages/tioga/package.py +++ b/var/spack/repos/builtin/packages/tioga/package.py @@ -2,24 +2,37 @@ # Spack Project Developers. See the top-level COPYRIGHT file for details. # # SPDX-License-Identifier: (Apache-2.0 OR MIT) - +import sys from spack import * -class Tioga(CMakePackage): +class Tioga(CMakePackage, CudaPackage): """Topology Independent Overset Grid Assembly (TIOGA)""" homepage = "https://github.com/jsitaraman/tioga" git = "https://github.com/jsitaraman/tioga.git" + maintainers = ['jsitaraman', 'sayerhs'] + + version('develop', branch='exawind') version('master', branch='master') - variant('shared', default=True, - description="Enable building shared libraries") + variant('shared', default=sys.platform != 'darwin', + description="Build shared libraries") variant('pic', default=True, description="Position independent code") + variant('nodegid', default=True, + description="Enable support for global Node IDs") + variant('timers', default=False, + description="Enable timers") + variant('stats', default=False, + description="Enable output of holecut stats") + variant('cxxstd', default='11', + values=('11', '14'), multi=False, + description="C++ standard to use") depends_on('mpi') + depends_on('cuda@9.0.0:', when='+cuda') # Tioga has the fortran module file problem with parallel builds parallel = False @@ -27,20 +40,26 @@ class Tioga(CMakePackage): def cmake_args(self): spec = self.spec - options = [ - '-DBUILD_SHARED_LIBS:BOOL=%s' % ( - 'ON' if '+shared' in spec else 'OFF'), - '-DCMAKE_POSITION_INDEPENDENT_CODE:BOOL=%s' % ( - 'ON' if '+pic' in spec else 'OFF'), - '-DCMAKE_C_COMPILER=%s' % spec['mpi'].mpicc, - '-DCMAKE_CXX_COMPILER=%s' % spec['mpi'].mpicxx, - '-DCMAKE_Fortran_COMPILER=%s' % spec['mpi'].mpifc, - '-DMPI_CXX_COMPILER:PATH=%s' % spec['mpi'].mpicxx, - '-DMPI_C_COMPILER:PATH=%s' % spec['mpi'].mpicc, - '-DMPI_Fortran_COMPILER:PATH=%s' % spec['mpi'].mpifc + args = [ + self.define_from_variant('BUILD_SHARED_LIBS', 'shared'), + self.define_from_variant('CMAKE_POSITION_INDEPENDENT_CODE', 'pic'), + self.define_from_variant('CMAKE_CXX_STANDARD', 'cxxstd'), + self.define_from_variant('TIOGA_HAS_NODEGID', 'nodegid'), + self.define_from_variant('TIOGA_ENABLE_TIMERS', 'timers'), + self.define_from_variant('TIOGA_OUTPUT_STATS', 'stats'), + self.define_from_variant('TIOGA_ENABLE_CUDA', 'cuda'), ] + if '+cuda' in self.spec: + args.append(self.define('CMAKE_CUDA_SEPARABLE_COMPILATION', True)) + + # Currently TIOGA only supports one device arch during specialization + cuda_arch = self.spec.variants['cuda_arch'].value + if cuda_arch: + arch_sorted = list(sorted(cuda_arch, reverse=True)) + args.append(self.define('TIOGA_CUDA_SM', arch_sorted[0])) + if 'darwin' in spec.architecture: - options.append('-DCMAKE_MACOSX_RPATH:BOOL=ON') + args.append(self.define('CMAKE_MACOSX_RPATH', True)) - return options + return args |