nvpl-blas and nvpl-lapack: new packages for NVidia performance libraries (#41775)
Co-authored-by: Raffaele Solcà <rasolca@cscs.ch>
This commit is contained in:
parent
85b6c344bd
commit
9aed13adb9
2 changed files with 138 additions and 0 deletions
69
var/spack/repos/builtin/packages/nvpl-blas/package.py
Normal file
69
var/spack/repos/builtin/packages/nvpl-blas/package.py
Normal file
|
@ -0,0 +1,69 @@
|
|||
# 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 *
|
||||
|
||||
|
||||
class NvplBlas(Package):
|
||||
"""
|
||||
NVPL BLAS (NVIDIA Performance Libraries BLAS) is part of NVIDIA Performance Libraries
|
||||
that provides standard Fortran 77 BLAS APIs as well as C (CBLAS).
|
||||
"""
|
||||
|
||||
homepage = "https://docs.nvidia.com/nvpl/_static/blas/index.html"
|
||||
url = (
|
||||
"https://developer.download.nvidia.com/compute/nvpl/redist"
|
||||
"/nvpl_blas/linux-sbsa/nvpl_blas-linux-sbsa-0.1.0-archive.tar.xz"
|
||||
)
|
||||
|
||||
maintainers("albestro", "rasolca")
|
||||
|
||||
license("UNKNOWN")
|
||||
|
||||
version("0.1.0", sha256="4ccc894593cbcbfaa1a4f3c54505982691971667acf191c9ab0f4252a37c8063")
|
||||
|
||||
provides("blas")
|
||||
|
||||
variant("ilp64", default=False, description="Force 64-bit Fortran native integers")
|
||||
variant(
|
||||
"threads",
|
||||
default="none",
|
||||
description="Multithreading support",
|
||||
values=("openmp", "none"),
|
||||
multi=False,
|
||||
)
|
||||
|
||||
requires("target=armv8.2a:", msg="Any CPU with Arm-v8.2a+ microarch")
|
||||
|
||||
conflicts("%gcc@:7")
|
||||
conflicts("%clang@:13")
|
||||
|
||||
conflicts("threads=openmp", when="%clang")
|
||||
|
||||
@property
|
||||
def blas_headers(self):
|
||||
return find_all_headers(self.spec.prefix.include)
|
||||
|
||||
@property
|
||||
def blas_libs(self):
|
||||
spec = self.spec
|
||||
|
||||
if "+ilp64" in spec:
|
||||
int_type = "ilp64"
|
||||
else:
|
||||
int_type = "lp64"
|
||||
|
||||
if spec.satisfies("threads=openmp"):
|
||||
threading_type = "gomp"
|
||||
else:
|
||||
# threads=none
|
||||
threading_type = "seq"
|
||||
|
||||
name = ["libnvpl_blas_core", f"libnvpl_blas_{int_type}_{threading_type}"]
|
||||
|
||||
return find_libraries(name, spec.prefix.lib, shared=True, recursive=True)
|
||||
|
||||
def install(self, spec, prefix):
|
||||
install_tree(".", prefix)
|
69
var/spack/repos/builtin/packages/nvpl-lapack/package.py
Normal file
69
var/spack/repos/builtin/packages/nvpl-lapack/package.py
Normal file
|
@ -0,0 +1,69 @@
|
|||
# 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 *
|
||||
|
||||
|
||||
class NvplLapack(Package):
|
||||
"""
|
||||
NVPL LAPACK (NVIDIA Performance Libraries LAPACK) is part of NVIDIA Performance Libraries
|
||||
that provides standard Fortran 90 LAPACK APIs.
|
||||
"""
|
||||
|
||||
homepage = "https://docs.nvidia.com/nvpl/_static/lapack/index.html"
|
||||
url = (
|
||||
"https://developer.download.nvidia.com/compute/nvpl/redist"
|
||||
"/nvpl_lapack/linux-sbsa/nvpl_lapack-linux-sbsa-0.2.0.1-archive.tar.xz"
|
||||
)
|
||||
|
||||
maintainers("albestro", "rasolca")
|
||||
|
||||
license("UNKNOWN")
|
||||
|
||||
version("0.1.0", sha256="7054f775b18916ee662c94ad7682ace53debbe8ee36fa926000fe412961edb0b")
|
||||
|
||||
provides("lapack")
|
||||
|
||||
variant("ilp64", default=False, description="Force 64-bit Fortran native integers")
|
||||
variant(
|
||||
"threads",
|
||||
default="none",
|
||||
description="Multithreading support",
|
||||
values=("openmp", "none"),
|
||||
multi=False,
|
||||
)
|
||||
|
||||
requires("target=armv8.2a:", msg="Any CPU with Arm-v8.2a+ microarch")
|
||||
|
||||
conflicts("%gcc@:7")
|
||||
conflicts("%clang@:13")
|
||||
|
||||
conflicts("threads=openmp", when="%clang")
|
||||
|
||||
@property
|
||||
def lapack_headers(self):
|
||||
return find_all_headers(self.spec.prefix.include)
|
||||
|
||||
@property
|
||||
def lapack_libs(self):
|
||||
spec = self.spec
|
||||
|
||||
if "+ilp64" in spec:
|
||||
int_type = "ilp64"
|
||||
else:
|
||||
int_type = "lp64"
|
||||
|
||||
if spec.satisfies("threads=openmp"):
|
||||
threading_type = "gomp"
|
||||
else:
|
||||
# threads=none
|
||||
threading_type = "seq"
|
||||
|
||||
name = ["libnvpl_lapack_core", f"libnvpl_lapack_{int_type}_{threading_type}"]
|
||||
|
||||
return find_libraries(name, spec.prefix.lib, shared=True, recursive=True)
|
||||
|
||||
def install(self, spec, prefix):
|
||||
install_tree(".", prefix)
|
Loading…
Reference in a new issue