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
94
95
96
97
98
99
100
101
102
103
104
105
106
107
108
109
110
111
112
113
114
115
116
117
118
119
120
121
122
123
124
125
126
127
128
129
130
131
132
133
134
135
136
137
138
139
140
141
142
143
144
145
146
147
148
149
150
151
152
153
154
155
156
157
158
159
160
161
162
163
164
165
166
167
168
|
# 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 SuiteSparse(Package):
"""
SuiteSparse is a suite of sparse matrix algorithms
"""
homepage = 'http://faculty.cse.tamu.edu/davis/suitesparse.html'
url = 'https://github.com/DrTimothyAldenDavis/SuiteSparse/archive/v4.5.3.tar.gz'
git = 'https://github.com/DrTimothyAldenDavis/SuiteSparse.git'
version('5.7.2', sha256='fe3bc7c3bd1efdfa5cffffb5cebf021ff024c83b5daf0ab445429d3d741bd3ad')
version('5.7.1', sha256='5ba5add1663d51a1b6fb128b50fe869b497f3096765ff7f8212f0ede044b9557')
version('5.6.0', sha256='76d34d9f6dafc592b69af14f58c1dc59e24853dcd7c2e8f4c98ffa223f6a1adb')
version('5.5.0', sha256='63c73451734e2bab19d1915796c6776565ea6aea5da4063a9797ecec60da2e3d')
version('5.4.0', sha256='d9d62d539410d66550d0b795503a556830831f50087723cb191a030525eda770')
version('5.3.0', sha256='d8ef4bee4394d2f07299d4688b83bbd98e9d3a2ebbe1c1632144b6f7095ce165')
version('5.2.0', sha256='68c431aef3d9a0b02e97803eb61671c5ecb9d36fd292a807db87067dadb36e53')
version('5.1.2', sha256='97dc5fdc7f78ff5018e6a1fcc841e17a9af4e5a35cebd62df6922349bf12959e')
version('5.1.0', sha256='0b0e03c63e67b04529bb6248808d2a8c82259d40b30fc5a7599f4b6f7bdd4dc6')
version('5.0.0', sha256='2f8694d9978033659f10ceb8bdb19147d3c519a0251b8de84be6ba8824d30517')
version('4.5.6', sha256='1c7b7a265a1d6c606095eb8aa3cb8e27821f1b7f5bc04f28df6d62906e02f4e4')
version('4.5.5', sha256='80d1d9960a6ec70031fecfe9adfe5b1ccd8001a7420efb50d6fa7326ef14af91')
version('4.5.3', sha256='b6965f9198446a502cde48fb0e02236e75fa5700b94c7306fc36599d57b563f4')
variant('tbb', default=False, description='Build with Intel TBB')
variant('pic', default=True, description='Build position independent code (required to link with shared libraries)')
variant('cuda', default=False, description='Build with CUDA')
variant('openmp', default=False, description='Build with OpenMP')
depends_on('blas')
depends_on('lapack')
depends_on('m4', type='build', when='@5.0.0:')
depends_on('cmake', when='@5.2.0:', type='build')
depends_on('metis@5.1.0', when='@4.5.1:')
# in @4.5.1. TBB support in SPQR seems to be broken as TBB-related linkng
# flags does not seem to be used, which leads to linking errors on Linux.
depends_on('tbb', when='@4.5.3:+tbb')
depends_on('cuda', when='+cuda')
patch('tbb_453.patch', when='@4.5.3:4.5.5+tbb')
# This patch removes unsupported flags for pgi compiler
patch('pgi.patch', when='%pgi')
patch('pgi.patch', when='%nvhpc')
# This patch adds '-lm' when linking libgraphblas and when using clang.
# Fixes 'libgraphblas.so.2.0.1: undefined reference to `__fpclassify''
patch('graphblas_libm_dep.patch', when='@5.2.0:5.2.99%clang')
conflicts('%gcc@:4.8', when='@5.2.0:', msg='gcc version must be at least 4.9 for suite-sparse@5.2.0:')
def install(self, spec, prefix):
# The build system of SuiteSparse is quite old-fashioned.
# It's basically a plain Makefile which include an header
# (SuiteSparse_config/SuiteSparse_config.mk)with a lot of convoluted
# logic in it. Any kind of customization will need to go through
# filtering of that file
cc_pic_flag = self.compiler.cc_pic_flag if '+pic' in spec else ''
f77_pic_flag = self.compiler.f77_pic_flag if '+pic' in spec else ''
make_args = [
# By default, the Makefile uses the Intel compilers if
# they are found. The AUTOCC flag disables this behavior,
# forcing it to use Spack's compiler wrappers.
'AUTOCC=no',
# CUDA=no does NOT disable cuda, it only disables internal search
# for CUDA_PATH. If in addition the latter is empty, then CUDA is
# completely disabled. See
# [SuiteSparse/SuiteSparse_config/SuiteSparse_config.mk] for more.
'CUDA=no',
'CUDA_PATH=%s' % (spec['cuda'].prefix if '+cuda' in spec else ''),
'CFOPENMP=%s' % (self.compiler.openmp_flag
if '+openmp' in spec else ''),
'CFLAGS=-O3 %s' % cc_pic_flag,
# Both FFLAGS and F77FLAGS are used in SuiteSparse makefiles;
# FFLAGS is used in CHOLMOD, F77FLAGS is used in AMD and UMFPACK.
'FFLAGS=%s' % f77_pic_flag,
'F77FLAGS=%s' % f77_pic_flag,
# use Spack's metis in CHOLMOD/Partition module,
# otherwise internal Metis will be compiled
'MY_METIS_LIB=%s' % spec['metis'].libs.ld_flags,
'MY_METIS_INC=%s' % spec['metis'].prefix.include,
# Make sure Spack's Blas/Lapack is used. Otherwise System's
# Blas/Lapack might be picked up. Need to add -lstdc++, following
# with the TCOV path of SparseSuite 4.5.1's Suitesparse_config.mk,
# even though this fix is ugly
'BLAS=%s' % (spec['blas'].libs.ld_flags + (
' -lstdc++' if '@4.5.1' in spec else '')),
'LAPACK=%s' % spec['lapack'].libs.ld_flags,
]
# Recent versions require c11 but some demos do not get the c11 from
# GraphBLAS/CMakeLists.txt, for example the file
# GraphBLAS/Demo/Program/wildtype_demo.c. For many compilers this is
# not an issue because c11 or newer is their default. However, for some
# compilers (e.g. xlc) the c11 flag is necessary.
if spec.satisfies('@5.4:5.7.1') and ('%xl' in spec or '%xl_r' in spec):
make_args += ['CFLAGS+=%s' % self.compiler.c11_flag]
# 64bit blas in UMFPACK:
if (spec.satisfies('^openblas+ilp64') or
spec.satisfies('^intel-mkl+ilp64') or
spec.satisfies('^intel-parallel-studio+mkl+ilp64')):
make_args.append('UMFPACK_CONFIG=-DLONGBLAS="long long"')
# SuiteSparse defaults to using '-fno-common -fexceptions' in
# CFLAGS, but not all compilers use the same flags for these
# optimizations
if any([x in spec
for x in ('%apple-clang', '%clang', '%gcc', '%intel')]):
make_args += ['CFLAGS+=-fno-common -fexceptions']
elif '%pgi' in spec:
make_args += ['CFLAGS+=--exceptions']
if spack_f77.endswith('xlf') or spack_f77.endswith('xlf_r'):
make_args += ['CFLAGS+=-DBLAS_NO_UNDERSCORE']
# Intel TBB in SuiteSparseQR
if 'tbb' in spec:
make_args += [
'SPQR_CONFIG=-DHAVE_TBB',
'TBB=%s' % spec['tbb'].libs.ld_flags,
]
if '@5.3:' in spec:
# Without CMAKE_LIBRARY_PATH defined, the CMake file in the
# Mongoose directory finds libsuitesparseconfig.so in system
# directories like /usr/lib.
make_args += [
'CMAKE_OPTIONS=-DCMAKE_INSTALL_PREFIX=%s' % prefix +
' -DCMAKE_LIBRARY_PATH=%s' % prefix.lib]
# In those SuiteSparse versions calling "make install" in one go is
# not possible, mainly because of GraphBLAS. Thus compile first and
# install in a second run.
if '@5.4.0:' in self.spec:
make('library', *make_args)
make_args.append('INSTALL=%s' % prefix)
make('install', *make_args)
@property
def libs(self):
"""Export the libraries of SuiteSparse.
Sample usage: spec['suite-sparse'].libs.ld_flags
spec['suite-sparse:klu,btf'].libs.ld_flags
"""
# Component libraries, ordered by dependency. Any missing components?
all_comps = ['klu', 'btf', 'umfpack', 'cholmod', 'colamd', 'amd',
'camd', 'ccolamd', 'cxsparse', 'ldl', 'rbio', 'spqr',
'suitesparseconfig']
query_parameters = self.spec.last_query.extra_parameters
comps = all_comps if not query_parameters else query_parameters
libs = find_libraries(['lib' + c for c in comps], root=self.prefix.lib,
shared=True, recursive=False)
if not libs:
return None
libs += find_system_libraries('librt')
return libs
|