summaryrefslogtreecommitdiff
path: root/var/spack/repos/builtin/packages/blaspp/package.py
blob: ddacfa22aa3af0afcec6001c2c5ca3769c83609f (plain) (blame)
1
2
3
4
5
6
7
8
9
10
11
12
13
14
15
16
17
18
19
20
21
22
23
24
25
26
27
28
29
30
31
32
33
34
35
36
37
38
39
40
41
42
43
44
45
46
47
48
49
50
51
52
53
54
55
56
57
58
59
60
61
62
63
64
65
66
67
68
69
70
71
72
73
74
75
76
77
78
79
80
81
82
83
84
85
86
87
88
89
90
91
92
93
# Copyright 2013-2020 Lawrence Livermore National Security, LLC and other
# Spack Project Developers. See the top-level COPYRIGHT file for details.
#
# SPDX-License-Identifier: (Apache-2.0 OR MIT)

from spack import *


class Blaspp(CMakePackage, CudaPackage):
    """C++ API for the Basic Linear Algebra Subroutines. Developed by the
       Innovative Computing Laboratory at the University of Tennessee,
       Knoxville."""

    homepage = "https://bitbucket.org/icl/blaspp"
    git      = "https://bitbucket.org/icl/blaspp"
    maintainers = ['teonnik', 'Sely85', 'G-Ragghianti', 'mgates3']

    version('develop', commit='6293d96')

    variant('gfort',
            default=False,
            description=('Use GNU Fortran interface. '
                         'Default is Intel interface. (MKL)'))
    variant('ilp64',
            default=False,
            description=('Use 64bit integer interface. '
                         'Default is 32bit. (MKL & ESSL)'))
    variant('openmp',
            default=False,
            description=('Use OpenMP threaded backend. '
                         'Default is sequential. (MKL & ESSL)'))

    depends_on('blas')

    # 1) The CMake options exposed by `blaspp` allow for a value called `auto`.
    #    The value is not needed here as the choice of dependency in the spec
    #    determines the appropriate flags.
    #
    # 2) BLASFinder.cmake handles most options. For `auto`, it searches all
    #    blas libraries listed in `def_lib_list`.
    #
    # 3) ?? Custom blas library can be supplied via `BLAS_LIBRARIES`.
    #
    def cmake_args(self):
        spec = self.spec
        args = ['-DBLASPP_BUILD_TESTS:BOOL={0}'.format(
            'ON' if self.run_tests else 'OFF')]

        if '+gfort' in spec:
            args.append('-DBLAS_LIBRARY_MKL="GNU gfortran conventions"')
        else:
            args.append('-DBLAS_LIBRARY_MKL="Intel ifort conventions"')

        if '+ilp64' in spec:
            args.append('-DBLAS_LIBRARY_INTEGER="int64_t (ILP64)"')
        else:
            args.append('-DBLAS_LIBRARY_INTEGER="int (LP64)"')

        if '+openmp' in spec:
            args.append(['-DUSE_OPENMP=ON',
                         '-DBLAS_LIBRARY_THREADING="threaded"'])
        else:
            args.append('-DBLAS_LIBRARY_THREADING="sequential"')

        # `blaspp` has an implicit CUDA detection mechanism. This disables it
        # in cases where it may backfire. One such case is when `cuda` is
        # external and marked with `buildable=false`. `blaspp`'s CMake CUDA
        # detection mechanism finds CUDA but doesn't set certain paths properly
        # which leads to a build issues [1].
        #
        # [1]: https://bitbucket.org/icl/blaspp/issues/6/compile-error-due-to-implicit-cuda
        if '~cuda' in spec:
            args.append('-DCMAKE_CUDA_COMPILER=')

        # Missing:
        #
        # - acml  : BLAS_LIBRARY="AMD ACML"
        #           BLAS_LIBRARY_THREADING= threaded/sequential
        #
        # - apple : BLAS_LIBRARY="Apple Accelerate" (veclibfort ???)
        #
        if '^mkl' in spec:
            args.append('-DBLAS_LIBRARY="Intel MKL"')
        elif '^essl' in spec:
            args.append('-DBLAS_LIBRARY="IBM ESSL"')
        elif '^openblas' in spec:
            args.append('-DBLAS_LIBRARY="OpenBLAS"')
        elif '^cray-libsci' in spec:
            args.append('-DBLAS_LIBRARY="Cray LibSci"')
        else:  # e.g. netlib-lapack
            args.append('-DBLAS_LIBRARY="generic"')

        return args