Files
os-traits/os_traits/tests/test_cuda.py
Jay Pipes 189fa9f027 clean up CUDA traits
There was a misunderstanding of what the CUDA traits really stood for.
This wasn't helped by the fact that the NVIDIA CUDA documentation and
versioning system is so awkward.

Basically, there's the CUDA Compute Capability version, which represents
the hardware GPU instruction set extensions supported by various NVIDIA
GPUs. And there is the CUDA SDK version which represents a window of
CUDA Compute Capability versions that the library is able to work with.

This patch adds a new os_traits.hw.gpu.cuda module with these new
traits. It removes the old os_traits.hw.gpu.api.CUDA* traits since those
were misleading and inaccurate.

It also adds a new utility function that maps CUDA SDK version to
supported CUDA Compute Capability versions that the SDK supports. This
support matrix information was taken from the CUDA wikipedia article
which lists the supported version information.

Change-Id: Id4adb6874615ad5f9f8aff780a83119e10a38c35
2018-09-19 10:03:54 -04:00

43 lines
1.5 KiB
Python

# -*- coding: utf-8 -*-
# Licensed under the Apache License, Version 2.0 (the "License"); you may
# not use this file except in compliance with the License. You may obtain
# a copy of the License at
#
# http://www.apache.org/licenses/LICENSE-2.0
#
# Unless required by applicable law or agreed to in writing, software
# distributed under the License is distributed on an "AS IS" BASIS, WITHOUT
# WARRANTIES OR CONDITIONS OF ANY KIND, either express or implied. See the
# License for the specific language governing permissions and limitations
# under the License.
from os_traits.hw.gpu import cuda
from os_traits.tests import base
class TestCUDA(base.TestCase):
def test_unknown_sdk_support(self):
self.assertIsNone(cuda.compute_capabilities_supported('UNKNOWN'))
def test_sdk6_5_support(self):
expected = set([
cuda.COMPUTE_CAPABILITY_V1_0,
cuda.COMPUTE_CAPABILITY_V1_1,
cuda.COMPUTE_CAPABILITY_V1_2,
cuda.COMPUTE_CAPABILITY_V1_3,
cuda.COMPUTE_CAPABILITY_V2_0,
cuda.COMPUTE_CAPABILITY_V2_1,
cuda.COMPUTE_CAPABILITY_V3_0,
cuda.COMPUTE_CAPABILITY_V3_2,
cuda.COMPUTE_CAPABILITY_V3_5,
cuda.COMPUTE_CAPABILITY_V3_7,
cuda.COMPUTE_CAPABILITY_V5_0,
cuda.COMPUTE_CAPABILITY_V5_2,
cuda.COMPUTE_CAPABILITY_V5_3,
])
actual = cuda.compute_capabilities_supported(cuda.SDK_V6_5)
self.assertEqual(expected, actual)