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
169
170
171
172
173
174
175
176
177
178
179
180
181
182
183
184
185
186
187
188
189
190
191
192
193
194
195
196
197
198
199
200
201
202
203
204
205
206
207
208
209
210
211
212
213
214
215
216
217
218
219
220
221
222
223
224
225
226
227
228
229
230
231
232
233
234
235
236
237
238
239
240
241
242
243
244
245
246
247
248
249
|
# Copyright 2013-2024 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)
import os
from llnl.util import tty
from spack.package import *
from spack.pkg.builtin.fftw import FftwBase
class Amdfftw(FftwBase):
"""FFTW (AMD Optimized version) is a comprehensive collection of
fast C routines for computing the Discrete Fourier Transform (DFT)
and various special cases thereof.
It is an open-source implementation of the Fast Fourier transform
algorithm. It can compute transforms of real and complex-values
arrays of arbitrary size and dimension.
AMD Optimized FFTW is the optimized FFTW implementation targeted
for AMD CPUs.
For single precision build, please use precision value as float.
Example : spack install amdfftw precision=float
LICENSING INFORMATION: By downloading, installing and using this software,
you agree to the terms and conditions of the AMD AOCL-FFTW license
agreement. You may obtain a copy of this license agreement from
https://www.amd.com/en/developer/aocl/fftw/eula/fftw-libraries-4-1-eula.html
https://www.amd.com/en/developer/aocl/fftw/eula/fftw-libraries-eula.html
"""
_name = "amdfftw"
homepage = "https://www.amd.com/en/developer/aocl/fftw.html"
url = "https://github.com/amd/amd-fftw/archive/3.0.tar.gz"
git = "https://github.com/amd/amd-fftw.git"
maintainers("amd-toolchain-support")
license("GPL-2.0-only")
version("4.1", sha256="f1cfecfcc0729f96a5bd61c6b26f3fa43bb0662d3fff370d4f73490c60cf4e59")
version("4.0", sha256="5f02cb05f224bd86bd88ec6272b294c26dba3b1d22c7fb298745fd7b9d2271c0")
version("3.2", sha256="31cab17a93e03b5b606e88dd6116a1055b8f49542d7d0890dbfcca057087b8d0")
version("3.1", sha256="3e777f3acef13fa1910db097e818b1d0d03a6a36ef41186247c6ab1ab0afc132")
version("3.0.1", sha256="87030c6bbb9c710f0a64f4f306ba6aa91dc4b182bb804c9022b35aef274d1a4c")
version("3.0", sha256="a69deaf45478a59a69f77c4f7e9872967f1cfe996592dd12beb6318f18ea0bcd")
version("2.2", sha256="de9d777236fb290c335860b458131678f75aa0799c641490c644c843f0e246f8")
variant("shared", default=True, description="Builds a shared version of the library")
variant("openmp", default=True, description="Enable OpenMP support")
variant("threads", default=False, description="Enable SMP threads support")
variant("debug", default=False, description="Builds a debug version of the library")
variant(
"amd-fast-planner",
default=False,
when="@3.0:",
description="Option to reduce the planning time without much "
"tradeoff in the performance. It is supported for "
"float and double precisions only.",
)
variant(
"amd-top-n-planner",
default=False,
when="@3.0.1: ~amd-fast-planner ~mpi ~openmp ~threads",
description="Build with amd-top-n-planner support",
)
variant(
"amd-mpi-vader-limit",
default=False,
when="@3.0.1:",
description="Build with amd-mpi-vader-limit support",
)
variant("static", default=False, description="Build with static suppport")
variant(
"amd-trans",
default=False,
when="~mpi ~openmp ~threads",
description="Build with amd-trans suppport",
)
variant(
"amd-app-opt",
default=False,
when="@3.1: ~mpi",
description="Build with amd-app-opt suppport",
)
variant(
"amd-dynamic-dispatcher",
default=True,
when="@4.1: %aocc@4.1.0:",
description="Single portable optimized library"
" to execute on different x86 CPU architectures",
)
variant(
"amd-dynamic-dispatcher",
default=True,
when="@3.2: %gcc",
description="Single portable optimized library"
" to execute on different x86 CPU architectures",
)
depends_on("texinfo")
provides("fftw-api@3", when="@2:")
conflicts(
"precision=quad",
when="@2.2 %aocc",
msg="Quad precision is not supported by AOCC clang version 2.2",
)
conflicts(
"+debug", when="@2.2 %aocc", msg="debug mode is not supported by AOCC clang version 2.2"
)
conflicts("%gcc@:7.2", when="@2.2:", msg="GCC version above 7.2 is required for AMDFFTW")
with when("+amd-fast-planner"):
conflicts("precision=quad", msg="Quad precision is not supported with amd-fast-planner")
conflicts(
"precision=long_double",
msg="long_double precision is not supported with amd-fast-planner",
)
with when("+amd-top-n-planner"):
conflicts("precision=quad", msg="Quad precision is not supported with amd-top-n-planner")
conflicts(
"precision=long_double",
msg="long_double precision is not supported with amd-top-n-planner",
)
conflicts(
"+amd-mpi-vader-limit",
when="precision=quad",
msg="Quad precision is not supported with amd-mpi-vader-limit",
)
with when("+amd-trans"):
conflicts(
"precision=long_double", msg="long_double precision is not supported with amd-trans"
)
conflicts("precision=quad", msg="Quad precision is not supported with amd-trans")
with when("+amd-app-opt"):
conflicts(
"precision=long_double", msg="long_double precision is not supported with amd-app-opt"
)
conflicts("precision=quad", msg="Quad precision is not supported with amd-app-opt")
requires("target=x86_64:", msg="AMD FFTW available only on x86_64")
def configure(self, spec, prefix):
"""Configure function"""
# Base options
options = ["--prefix={0}".format(prefix), "--enable-amd-opt"]
# Dynamic dispatcher builds a single portable optimized library
# that can execute on different x86 CPU architectures.
# It is supported for GCC compiler and Linux based systems only.
if "+amd-dynamic-dispatcher" in spec:
options.append("--enable-dynamic-dispatcher")
# Check if compiler is AOCC
if "%aocc" in spec:
options.append("CC={0}".format(os.path.basename(spack_cc)))
options.append("FC={0}".format(os.path.basename(spack_fc)))
options.append("F77={0}".format(os.path.basename(spack_fc)))
if not (
spec.satisfies(r"%aocc@3.2:4.1")
or spec.satisfies(r"%gcc@12.2:13.1")
or spec.satisfies(r"%clang@15:16")
):
tty.warn(
"AOCL has been tested to work with the following compilers\
versions - gcc@12.2:13.1, aocc@3.2:4.1, and clang@15:16\
see the following aocl userguide for details: \
https://www.amd.com/content/dam/amd/en/documents/developer/version-4-1-documents/aocl/aocl-4-1-user-guide.pdf"
)
if "+debug" in spec:
options.append("--enable-debug")
if "+mpi" in spec:
options.append("--enable-mpi")
options.append("--enable-amd-mpifft")
else:
options.append("--disable-mpi")
options.append("--disable-amd-mpifft")
options.extend(self.enable_or_disable("shared"))
options.extend(self.enable_or_disable("openmp"))
options.extend(self.enable_or_disable("threads"))
options.extend(self.enable_or_disable("amd-fast-planner"))
options.extend(self.enable_or_disable("amd-top-n-planner"))
options.extend(self.enable_or_disable("amd-mpi-vader-limit"))
options.extend(self.enable_or_disable("static"))
options.extend(self.enable_or_disable("amd-trans"))
options.extend(self.enable_or_disable("amd-app-opt"))
if not self.compiler.f77 or not self.compiler.fc:
options.append("--disable-fortran")
# Cross compilation is supported in amd-fftw by making use of target
# variable to set AMD_ARCH configure option.
# Spack user can not directly use AMD_ARCH for this purpose but should
# use target variable to set appropriate -march option in AMD_ARCH.
arch = spec.architecture
options.append(
"AMD_ARCH={0}".format(arch.target.optimization_flags(spec.compiler).split("=")[-1])
)
# Specific SIMD support.
# float and double precisions are supported
simd_features = ["sse2", "avx", "avx2", "avx512"]
# "avx512" is supported from amdfftw 4.0 version onwards
if "@2.2:3.2" in self.spec:
simd_features.remove("avx512")
simd_options = []
for feature in simd_features:
msg = "--enable-{0}" if feature in spec.target else "--disable-{0}"
simd_options.append(msg.format(feature))
# When enabling configure option "--enable-amd-opt", do not use the
# configure option "--enable-generic-simd128" or
# "--enable-generic-simd256"
# Double is the default precision, for all the others we need
# to enable the corresponding option.
enable_precision = {
"float": ["--enable-float"],
"double": None,
"long_double": ["--enable-long-double"],
"quad": ["--enable-quad-precision"],
}
# Different precisions must be configured and compiled one at a time
configure = Executable("../configure")
for precision in self.selected_precisions:
opts = (enable_precision[precision] or []) + options[:]
# SIMD optimizations are available only for float and double
if precision in ("float", "double"):
opts += simd_options
with working_dir(precision, create=True):
configure(*opts)
|
