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
|
# 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)
from spack.package import *
target_arch_list = (
"armv7a",
"armv7a-hf",
"arm64-v8a",
"arm64-v8.2-a",
"x86_32",
"x86_64",
"armv8a",
"armv8.2-a",
"armv8.6-a",
"armv8r64",
"x86",
)
class Armcomputelibrary(SConsPackage):
"""The Arm Compute Library is a collection of low-level machine learning functions optimized
for Arm® Cortex®-A, Arm® Neoverse® and Arm® Mali™ GPUs architectures.
The library provides superior performance to other open source alternatives and
immediate support for new Arm® technologies e.g. SVE2."""
homepage = "https://arm-software.github.io/ComputeLibrary/latest/"
url = "https://github.com/ARM-software/ComputeLibrary/archive/refs/tags/v23.02.zip"
git = "https://github.com/ARM-software/ComputeLibrary.git"
maintainers("annop-w")
license("MIT")
version("23.02", sha256="bed1b24047ce00155e552204bc3983e86f46775414c554a34a7ece931d67ec62")
version("22.11", sha256="2f70f54d84390625222503ea38650c00c49d4b70bc86a6b9aeeebee9d243865f")
version("22.08", sha256="5d76d07406b105f0bdf74ef80263236cb03baf0ade882f2bf8446bbc239e0079")
version("22.05", sha256="8ff308448874c6b72c1ce8d9f28af41d8b47c8e5c43b8ccc069da744e3c0a421")
version("22.02", sha256="0c1fe30b24e78bf5ca313ee8a33ad95e0d2aaddf64d4518ecec6a95e4bfba6e5")
depends_on("scons@2.3:")
phases = ["build"]
variant("build_type", default="release", values=("release", "debug"), description="Build type")
variant(
"threads",
default="cppthreads",
values=("cppthreads", "openmp"),
description="Enable C++11 threads/OpenMP backend. OpenMP backend only "
"works when building with g++ and not clang++.",
)
variant(
"multi_isa",
default=False,
description="Build Multi ISA binary version of library." " Note works only for armv8.2-a.",
)
variant(
"target_arch",
default="armv7a",
values=target_arch_list,
description="Target Architecture. The x86_32 and x86_64 targets can only be"
" used with neon=0 and opencl=1.",
)
variant("sve", default=False, description="Build for SVE.")
variant("sve2", default=False, description="Build for SVE2.")
variant("neon", default=True, description="Enable Arm® Neon™ support")
variant(
"experimental_dynamic_fusion",
default=False,
description="Build the experimental dynamic fusion files.",
)
variant(
"experimental_fixed_format_kernels",
default=False,
description="Enable fixed format kernels for GEMM.",
)
variant("benchmark_examples", default=False, description="Build benchmark examples programs.")
variant("validate_examples", default=False, description="Build validate examples programs.")
variant("validation_tests", default=False, description="Build validation test programs.")
variant("benchmark_tests", default=False, description="Build benchmark test programs.")
def build_args(self, spec, prefix):
args = ["-j{0}".format(make_jobs)]
if spec.satisfies("build_type=debug"):
args.append("debug=1")
elif spec.satisfies("build_type=release"):
args.append("debug=0")
if spec.satisfies("threads=openmp"):
args.append("cppthreads=0")
args.append("openmp=1")
elif spec.satisfies("threads=cppthreads"):
args.append("cppthreads=1")
args.append("openmp=0")
arch_value = spec.variants["target_arch"].value
arch_value += "-sve" if spec.variants["sve"].value else ""
arch_value += "-sve2" if spec.variants["sve2"].value else ""
args.append("arch={}".format(arch_value))
args.append("multi_isa={}".format(int(spec.variants["multi_isa"].value)))
args.append("neon={}".format(int(spec.variants["neon"].value)))
args.append(
"experimental_dynamic_fusion={}".format(
int(spec.variants["experimental_dynamic_fusion"].value)
)
)
args.append(
"experimental_fixed_format_kernels={}".format(
int(spec.variants["experimental_fixed_format_kernels"].value)
)
)
args.append("benchmark_examples={}".format(int(spec.variants["benchmark_examples"].value)))
args.append("validate_examples={}".format(int(spec.variants["validate_examples"].value)))
args.append("validation_tests={}".format(int(spec.variants["validation_tests"].value)))
args.append("benchmark_tests={}".format(int(spec.variants["benchmark_tests"].value)))
args.append("build=native")
args.append("install_dir={0}".format(prefix))
return args
def setup_build_environment(self, env):
# Spack compiler wrapper inject -mcpu flag for some targets.
# This can conflict with -march set in scons script, so override it here.
env.set("SPACK_TARGET_ARGS", "")
@property
def libs(self):
acl_libs = find_libraries(
["libarm_compute", "libarm_compute_core", "libarm_compute_graph"],
root=self.spec.prefix,
shared=True,
recursive=True,
)
return acl_libs
@property
def headers(self):
incdir = join_path(self.spec.prefix, "include")
hlist = find_all_headers(incdir)
hlist.directories = [incdir]
return hlist
@run_after("build")
def make_sym_link(self):
# oneDNN expects header and lib files for ACL in arm_compute/ and build/ directory
prefix = self.spec.prefix
symlink(join_path(prefix, "include/arm_compute"), join_path(prefix, "arm_compute"))
symlink(join_path(prefix, "lib"), join_path(prefix, "build"))
@run_after("build")
def copy_additional_header(self):
# copy some additional header files needed by oneDNN
cpuinfo_dir = "src/common/cpuinfo"
mkdirp(join_path(self.spec.prefix, cpuinfo_dir))
for f in ["CpuInfo.h", "CpuIsaInfo.h", "CpuModel.h"]:
copy(
join_path(self.stage.source_path, cpuinfo_dir, f),
join_path(self.spec.prefix, cpuinfo_dir, f),
)
|
