New package: py-torch-scatter (#18263)

* [py-torch-scatter] created template

* [py-torch-scatter] listed specific version of python

* [py-torch-scatter] extends py-torch

* [py-torch-scatter] setting TORCH_CUDA_ARCH_LIST

* [py-torch-scatter] setting more environemnt variables and added variant cuda

* [py-torch-scatter] added homepage and description. removed fixmes

* [py-torch-scatter] flake8

* [py-torch-scatter] Added variant description

* [py-torch-scatter] extends -> depends_on

* [py-torch-scatter] added dependencies of setup tools and pytest-runner

1 files changed, 39 insertions, 0 deletions

diff --git a/var/spack/repos/builtin/packages/py-torch-scatter/package.py b/var/spack/repos/builtin/packages/py-torch-scatter/package.py
new file mode 100644
index 0000000000..7950abb7b9
--- /dev/null
+++ b/var/spack/repos/builtin/packages/py-torch-scatter/package.py
@@ -0,0 +1,39 @@
+# Copyright 2013-2020 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 import *
+
+
+class PyTorchScatter(PythonPackage):
+    """This package consists of a small extension library of
+    highly optimized sparse update (scatter and segment)
+    operations for the use in PyTorch, which are missing in the
+    main package."""
+
+    homepage = "https://github.com/rusty1s/pytorch_scatter"
+    url      = "https://github.com/rusty1s/pytorch_scatter/archive/2.0.5.tar.gz"
+
+    version('2.0.5', sha256='e29b364beaa9c84a99e0e236be89ed19d4452d89010ff736184ddcce488b47f6')
+
+    variant('cuda', default=False, description="Enable CUDA support")
+
+    depends_on('python@3.6:', type=('build', 'run'))
+    depends_on('py-setuptools', type='build')
+    depends_on('py-pytest-runner', type='build')
+    depends_on('py-torch+cuda', when='+cuda')
+    depends_on('py-torch~cuda', when='~cuda')
+
+    def setup_build_environment(self, env):
+        if '+cuda' in self.spec:
+            cuda_arches = list(
+                self.spec['py-torch'].variants['cuda_arch'].value)
+            for i, x in enumerate(cuda_arches):
+                cuda_arches[i] = '{0}.{1}'.format(x[0:-1], x[-1])
+            env.set('TORCH_CUDA_ARCH_LIST', str.join(' ', cuda_arches))
+
+            env.set('FORCE_CUDA', '1')
+            env.set('CUDA_HOME', self.spec['cuda'].prefix)
+        else:
+            env.set('FORCE_CUDA', '0')