19c20563cc
This patch adds license information for about 5,300 packages from automated sources. The license information was obtained from Alpine Linux and PyPI and processed using tooling available in https://github.com/boomanaiden154/spack-license-utils. The license field was added in after all other directives in an automated fashion. Note that while this license information is probably fairly accurate, it is not guaranteed to be accurate. In addition some of the license strings from Alpine Linux might not be valid SPDX license strings. Invalid SPDX identifiers can be picked up and fixed once we have validation/parsing infrastructure in place for the solver, and issues can be fixed as they come up.
111 lines
4.1 KiB
Python
111 lines
4.1 KiB
Python
# 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 *
|
|
|
|
|
|
class Alpaka(CMakePackage, CudaPackage):
|
|
"""Abstraction Library for Parallel Kernel Acceleration."""
|
|
|
|
homepage = "https://alpaka.readthedocs.io"
|
|
url = "https://github.com/alpaka-group/alpaka/archive/refs/tags/0.6.0.tar.gz"
|
|
git = "https://github.com/alpaka-group/alpaka.git"
|
|
|
|
maintainers("vvolkl")
|
|
|
|
license("MPL-2.0-no-copyleft-exception")
|
|
|
|
version("develop", branch="develop")
|
|
version("0.8.0", sha256="e01bc377a7657d9a3e0c5f8d3f83dffbd7d0b830283c59efcbc1fb98cf88de43")
|
|
version("0.7.0", sha256="4b61119a7b3b073f281ba15b63430db98b77dbd9420bc290a114f80121fbdd97")
|
|
version("0.6.0", sha256="7424ecaee3af15e587b327e983998410fa379c61d987bfe923c7e95d65db11a3")
|
|
version("0.5.0", sha256="0ba08ea19961dd986160219ba00d6162fe7758980d88a606eff6494d7b3a6cd1")
|
|
version("0.4.0", sha256="ad7905b13c22abcee4344ba225a65078e3f452ad45a9eda907e7d27c08315e46")
|
|
|
|
variant(
|
|
"backend",
|
|
multi=True,
|
|
values=(
|
|
"serial",
|
|
"threads",
|
|
"fiber",
|
|
"tbb",
|
|
"omp2_gridblock",
|
|
"omp2_blockthread",
|
|
"omp5",
|
|
"oacc",
|
|
"cuda",
|
|
"cuda_only",
|
|
"hip",
|
|
"hip_only",
|
|
),
|
|
description="Backends to enable",
|
|
default="serial",
|
|
)
|
|
|
|
variant("examples", default=False, description="Build alpaka examples")
|
|
|
|
depends_on("boost")
|
|
depends_on("boost+fiber", when="backend=fiber")
|
|
depends_on("cmake@3.18:", when="@0.7.0:")
|
|
|
|
# make sure no other backend is enabled if using cuda_only or hip_only
|
|
for v in (
|
|
"serial",
|
|
"threads",
|
|
"fiber",
|
|
"tbb",
|
|
"oacc",
|
|
"omp2_gridblock",
|
|
"omp2_blockthread",
|
|
"omp5",
|
|
"cuda",
|
|
"hip",
|
|
):
|
|
conflicts("backend=cuda_only,%s" % v)
|
|
conflicts("backend=hip_only,%s" % v)
|
|
conflicts("backend=cuda_only,hip_only")
|
|
for v in ("omp2_blockthread", "omp2_blockthread", "omp5"):
|
|
conflicts("backend=oacc,%s" % v)
|
|
|
|
# todo: add conflict between cuda 11.3 and gcc 10.3.0
|
|
# see https://github.com/alpaka-group/alpaka/issues/1297
|
|
|
|
def cmake_args(self):
|
|
spec = self.spec
|
|
args = []
|
|
if "backend=serial" in spec:
|
|
args.append(self.define("ALPAKA_ACC_CPU_B_SEQ_T_SEQ_ENABLE", True))
|
|
if "backend=threads" in self.spec:
|
|
args.append(self.define("ALPAKA_ACC_CPU_B_SEQ_T_THREADS_ENABLE", True))
|
|
if "backend=fiber" in spec:
|
|
args.append(self.define("ALPAKA_ACC_CPU_B_SEQ_T_FIBERS_ENABLE", True))
|
|
if "backend=tbb" in spec:
|
|
args.append(self.define("ALPAKA_ACC_CPU_B_TBB_T_SEQ_ENABLE", True))
|
|
if "backend=omp2_gridblock" in spec:
|
|
args.append(self.define("ALPAKA_ACC_CPU_B_OMP2_T_SEQ_ENABLE", True))
|
|
if "backend=omp2_blockthread" in spec:
|
|
args.append(self.define("ALPAKA_ACC_CPU_B_SEQ_T_OMP2_ENABLE", True))
|
|
if "backend=omp5" in spec:
|
|
args.append(self.define("ALPAKA_ACC_ANY_BT_OMP5_ENABLE", True))
|
|
if "backend=oacc" in spec:
|
|
args.append(self.define("ALPAKA_ACC_ANY_BT_OACC_ENABLE", True))
|
|
if "backend=cuda" in spec:
|
|
args.append(self.define("ALPAKA_ACC_GPU_CUDA_ENABLE", True))
|
|
if "backend=cuda_only" in spec:
|
|
args.append(self.define("ALPAKA_ACC_GPU_CUDA_ENABLE", True))
|
|
args.append(self.define("ALPAKA_ACC_GPU_CUDA_ONLY_MODE", True))
|
|
if "backend=hip" in spec:
|
|
args.append(self.define("ALPAKA_ACC_GPU_HIP_ENABLE", True))
|
|
if "backend=hip_only" in spec:
|
|
args.append(self.define("ALPAKA_ACC_GPU_HIP_ENABLE", True))
|
|
args.append(self.define("ALPAKA_ACC_GPU_HIP_ONLY_MODE", True))
|
|
|
|
args.append(self.define_from_variant("alpaka_BUILD_EXAMPLES", "examples"))
|
|
# need to define, as it is explicitly declared as an option by alpaka:
|
|
args.append(self.define("BUILD_TESTING", self.run_tests))
|
|
return args
|