zhangyiss/spack
1
0
Fork 0
Code Issues Pull Requests Actions Packages Projects Releases Wiki Activity

Package rename: intel-mkl-dnn -> dnnl (#15852)

Browse Source
This commit is contained in:
Adam J. Stewart 2020-04-04 11:35:23 -05:00 committed by GitHub
parent 4ee9fd46bc
commit 9b9b46c0a9
No known key found for this signature in database
GPG Key ID: 4AEE18F83AFDEB23
2 changed files with 39 additions and 10 deletions
Show Stats Download Patch File Download Diff File Expand all files Collapse all files

47
var/spack/repos/builtin/packages/intel-mkl-dnn/package.py → var/spack/repos/builtin/packages/dnnl/package.py
View File

@ -3,14 +3,15 @@
#
# SPDX-License-Identifier: (Apache-2.0 OR MIT)
from spack import *
import sys
class IntelMklDnn(CMakePackage):
"""Intel(R) Math Kernel Library for Deep Neural Networks
(Intel(R) MKL-DNN)."""
class Dnnl(CMakePackage):
"""Deep Neural Network Library (DNNL).
homepage = "https://intel.github.io/mkl-dnn/"
Formerly known as Intel MKL-DNN."""
homepage = "https://01.org/dnnl"
url = "https://github.com/intel/mkl-dnn/archive/v1.3.tar.gz"
maintainers = ['adamjstewart']
@ -46,15 +47,38 @@ class IntelMklDnn(CMakePackage):
version('0.10', sha256='59828764ae43f1151f77b8997012c52e0e757bc50af1196b86fce8934178c570')
version('0.9', sha256='8606a80851c45b0076f7d4047fbf774ce13d6b6d857cb2edf95c7e1fd4bca1c7')
default_cpu_runtime = 'omp'
if sys.platform == 'darwin':
default_cpu_runtime = 'tbb'
variant('cpu_runtime', default=default_cpu_runtime,
description='CPU threading runtime to use',
values=('omp', 'tbb', 'seq'), multi=False)
variant('gpu_runtime', default='none',
description='Runtime to use for GPU engines',
values=('ocl', 'none'), multi=False)
# https://github.com/intel/mkl-dnn#requirements-for-building-from-source
depends_on('cmake@2.8.11:', type='build')
depends_on('intel-mkl')
depends_on('llvm-openmp', when='%clang platform=darwin')
depends_on('tbb@2017:', when='cpu_runtime=tbb')
depends_on('llvm-openmp', when='%clang platform=darwin cpu_runtime=omp')
depends_on('opencl@1.2:', when='gpu_runtime=ocl')
def cmake_args(self):
args = []
args = [
'-DDNNL_CPU_RUNTIME={0}'.format(
self.spec.variants['cpu_runtime'].value.upper()),
'-DDNNL_GPU_RUNTIME={0}'.format(
self.spec.variants['gpu_runtime'].value.upper()),
]
if self.run_tests:
args.append('-DDNNL_BUILD_TESTS=ON')
else:
args.append('-DDNNL_BUILD_TESTS=OFF')
# https://github.com/intel/mkl-dnn/issues/591
if self.spec.satisfies('%clang platform=darwin'):
if self.spec.satisfies('%clang platform=darwin cpu_runtime=omp'):
args.extend([
'-DOpenMP_CXX_FLAGS={0}'.format(self.compiler.openmp_flag),
'-DOpenMP_C_FLAGS={0}'.format(self.compiler.openmp_flag),
@ -67,5 +91,10 @@ def cmake_args(self):
self.spec['llvm-openmp'].libs.ld_flags
),
])
elif self.spec.satisfies('cpu_runtime=tbb'):
args.append('-DTBBROOT=' + self.spec['tbb'].prefix)
if self.spec.satisfies('gpu_runtime=ocl'):
args.append('-DOPENCLROOT=' + self.spec['opencl'].prefix)
return args

2
var/spack/repos/builtin/packages/py-torch/package.py
View File

@ -154,7 +154,7 @@ class PyTorch(PythonPackage, CudaPackage):
# TODO: See if there is a way to use an external mkldnn installation.
# Currently, only older versions of py-torch use an external mkldnn
# library.
depends_on('intel-mkl-dnn', when='@0.4:0.4.1+mkldnn')
depends_on('dnnl', when='@0.4:0.4.1+mkldnn')
# TODO: add dependency: https://github.com/Maratyszcza/NNPACK
# depends_on('nnpack', when='+nnpack')
depends_on('qnnpack', when='+qnnpack')