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
|
# Copyright 2013-2023 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 *
# Wrapper around depends_on to propagate dependency variants
def dav_sdk_depends_on(spec, when=None, propagate=None):
# Do the basic depends_on
depends_on(spec, when=when)
# Strip spec string to just the base spec name
# ie. A +c ~b -> A
spec = Spec(spec).name
# If the package is in the spec tree then it must be enabled in the SDK.
if "+" in when:
_when_variants = when.strip("+").split("+")
if any(tok in when for tok in ["~", "="]):
tty.error("Bad token in when clause, only positive boolean tokens allowed")
for variant in _when_variants:
conflicts("~" + variant, when="^" + spec)
# Skip if there is nothing to propagate
if not propagate:
return
# Map the propagated variants to the dependency variant. Some packages may need
# overrides to propagate a dependency as something else, e.g., {"visit": "libsim"}.
# Most call-sites will just use a list.
if not type(propagate) is dict:
propagate = dict([(v, v) for v in propagate])
# Determine the base variant
base_variant = ""
if when:
base_variant = when
def is_boolean(variant):
return "=" not in variant
# Propagate variants to dependecy
for v_when, v_then in propagate.items():
if is_boolean(v_when):
depends_on(
"{0} +{1}".format(spec, v_then), when="{0} +{1}".format(base_variant, v_when)
)
depends_on(
"{0} ~{1}".format(spec, v_then), when="{0} ~{1}".format(base_variant, v_when)
)
else:
depends_on("{0} {1}".format(spec, v_then), when="{0} {1}".format(base_variant, v_when))
def exclude_variants(variants, exclude):
return [variant for variant in variants if variant not in exclude]
class EcpDataVisSdk(BundlePackage, CudaPackage, ROCmPackage):
"""ECP Data & Vis SDK"""
homepage = "https://ecp-data-vis-sdk.github.io/"
tags = ["ecp"]
maintainers("kwryankrattiger", "svenevs")
version("1.0")
############################################################
# Variants
############################################################
# I/O
variant("adios2", default=False, description="Enable ADIOS2")
variant("darshan", default=False, description="Enable Darshan")
variant("faodel", default=False, description="Enable FAODEL")
variant("hdf5", default=False, description="Enable HDF5")
variant("pnetcdf", default=False, description="Enable PNetCDF")
variant("unifyfs", default=False, description="Enable UnifyFS")
variant("veloc", default=False, description="Enable VeloC")
# Vis
variant("ascent", default=False, description="Enable Ascent")
variant("cinema", default=False, description="Enable Cinema")
variant("paraview", default=False, description="Enable ParaView")
variant("sensei", default=False, description="Enable Sensei")
variant("sz", default=False, description="Enable SZ")
variant("visit", default=False, description="Enable VisIt")
variant("vtkm", default=False, description="Enable VTK-m")
variant("zfp", default=False, description="Enable ZFP")
# Language Options
variant("fortran", default=True, sticky=True, description="Enable fortran language features.")
############################################################
# Dependencies
############################################################
cuda_arch_variants = ["cuda_arch={0}".format(x) for x in CudaPackage.cuda_arch_values]
amdgpu_target_variants = ["amdgpu_target={0}".format(x) for x in ROCmPackage.amdgpu_targets]
dav_sdk_depends_on(
"adios2+shared+mpi+python+sst+dataman",
when="+adios2",
propagate=["cuda", "hdf5", "sz", "zfp", "fortran"] + cuda_arch_variants,
)
dav_sdk_depends_on("darshan-runtime+mpi", when="+darshan")
dav_sdk_depends_on("darshan-util", when="+darshan")
dav_sdk_depends_on("faodel+shared+mpi network=libfabric", when="+faodel", propagate=["hdf5"])
# HDF5 1.14 is a soft requirement for the ECP Data and Vis SDK.
# When building with VisIt and CinemaSci in the same SDK environment there is a conflict
# in the build dependency of py-setuptools which prevents building a py-h5py that is
# compatible with 'hdf5@1.14:'. Until there is a version of VisIt with an updated VTK or Spack
# allows the concretization of multiple versions of the same build only dependency
# concretization with VisIt and Cinema variants will not allow building VOLs.
dav_sdk_depends_on("hdf5@1.12: +shared+mpi", when="+hdf5", propagate=["fortran"])
# HDF5 VOL Adapters require hdf5@1.14:
# hdf5-vfd-gds needs cuda@11.7.1 or later, only enable when 11.7.1+ available.
depends_on("hdf5-vfd-gds@1.0.2:", when="+cuda+hdf5 ^cuda@11.7.1: ^hdf5@1.14:")
for cuda_arch in cuda_arch_variants:
depends_on(
"hdf5-vfd-gds@1.0.2: {0}".format(cuda_arch),
when="+cuda+hdf5 {0} ^cuda@11.7.1: ^hdf5@1.14:".format(cuda_arch),
)
conflicts("~cuda", when="^hdf5-vfd-gds@1.0.2:")
conflicts("~hdf5", when="^hdf5-vfd-gds@1.0.2:")
conflicts("~hdf5", when="^hdf5-vol-async")
conflicts("~hdf5", when="^hdf5-vol-cache")
conflicts("~hdf5", when="^hdf5-vol-log")
depends_on("hdf5-vol-async", when="+hdf5 ^hdf5@1.14:")
depends_on("hdf5-vol-cache", when="+hdf5 ^hdf5@1.14:")
depends_on("hdf5-vol-log", when="+hdf5 ^hdf5@1.14:")
dav_sdk_depends_on("parallel-netcdf+shared", when="+pnetcdf", propagate=["fortran"])
dav_sdk_depends_on("unifyfs", when="+unifyfs ")
conflicts("^unifyfs@develop")
dav_sdk_depends_on("veloc", when="+veloc")
# Skipping propagating ascent, catalyst(paraview), and libsim(visit) to sensei
# due to incomaptiblity between these variants in sensei.
dav_sdk_depends_on("sensei@4: ~vtkio +python", when="+sensei", propagate=["adios2", "hdf5"])
# Fortran support with ascent is problematic on some Cray platforms so the
# SDK is explicitly disabling it until the issues are resolved.
dav_sdk_depends_on(
"ascent+mpi~fortran+python+shared+vtkh+dray~test",
when="+ascent",
propagate=["adios2", "cuda"] + cuda_arch_variants,
)
depends_on("ascent+openmp", when="~rocm+ascent")
depends_on("ascent~openmp", when="+rocm+ascent")
# Need to explicitly turn off conduit hdf5_compat in order to build
# hdf5@1.12 which is required for SDK
depends_on("conduit ~hdf5_compat", when="+ascent +hdf5")
# Disable configuring with @develop. This should be removed after ascent
# releases 0.8 and ascent can build with conduit@0.8: and vtk-m@1.7:
conflicts("^ascent@develop", when="+ascent")
depends_on("py-cinemasci", when="+cinema")
# ParaView needs @5.11: in order to use CUDA/ROCM, therefore it is the minimum
# required version since GPU capability is desired for ECP
dav_sdk_depends_on(
"paraview@5.11:+mpi+openpmd+python+kits+shared+catalyst+libcatalyst" " use_vtkm=on",
when="+paraview",
propagate=["adios2", "cuda", "hdf5", "rocm"] + amdgpu_target_variants + cuda_arch_variants,
)
dav_sdk_depends_on("libcatalyst@2:+mpi", when="+paraview")
conflicts("^paraview@master", when="+paraview")
dav_sdk_depends_on("visit+mpi+python+silo", when="+visit", propagate=["hdf5", "adios2"])
dav_sdk_depends_on(
"vtk-m@1.7:+shared+mpi+rendering",
when="+vtkm",
propagate=["cuda", "rocm"] + cuda_arch_variants + amdgpu_target_variants,
)
# TODO: When Ascent is updated to use VTK-m >= 1.8 move examples to
# the main spec.
depends_on("vtk-m+examples", when="+vtkm ^vtk-m@1.8:")
depends_on("vtk-m+openmp", when="~rocm+vtkm")
depends_on("vtk-m~openmp", when="+rocm+vtkm")
# +python is currently broken in sz
# dav_sdk_depends_on('sz+shared+python+random_access',
dav_sdk_depends_on("sz+shared+random_access", when="+sz", propagate=["hdf5", "fortran"])
dav_sdk_depends_on("zfp", when="+zfp", propagate=["cuda"] + cuda_arch_variants)
|
