[test] New flexible CUDA memtest check #346
Conversation
vkarak
changed the title
| self.valid_prog_environs = ['PrgEnv-cray', 'PrgEnv-gnu', | ||
| 'PrgEnv-intel', 'PrgEnv-pgi'] | ||
| self.descr = 'Flexible Cuda Memtest' | ||
| self.executable = 'osu_alltoall' |
There was a problem hiding this comment.
Is this the name of the test, really?
| self.descr = 'Flexible Cuda Memtest' | ||
| self.executable = 'osu_alltoall' | ||
| self.maintainers = ['TM', 'VK'] | ||
| self.tags = {'maintenance'} |
There was a problem hiding this comment.
Don't put a tag here. Otherwise it will run on every maintenance session on all the nodes!
| 'cuda_memtest-1.2.3.tar.gz') | ||
| self.prebuild_cmd = ['wget %s' % src_url, | ||
| 'tar -xzf cuda_memtest-1.2.3.tar.gz ' | ||
| '--strip-components=1'] |
There was a problem hiding this comment.
If you format the list as follows, the second command might fit in a single line:
self.prebuild_cmd = [
'wget ...'
'tar ...'
]| self.executable = 'cuda_memtest_sm20' | ||
| self.executable_opts = ['--disable_test', '6', '--num_passes', '1'] | ||
|
|
||
| valid_test_ids = [i for i in range(11) if i not in {6, 9}] |
| valid_test_ids = [i for i in range(11) if i not in {6, 9}] | ||
| assert_finished_tests = [sn.assert_eq(sn.count(sn.findall( | ||
| 'Test%s finished' % test_id, self.stdout)), | ||
| self.num_tasks_assigned) for test_id in valid_test_ids] |
There was a problem hiding this comment.
This expression is a bit hard to follow. Can you format a bit better, so as to be more readable?
| super().__init__() | ||
| self.valid_systems = ['daint:gpu', 'dom:gpu'] | ||
| self.valid_prog_environs = ['PrgEnv-cray', 'PrgEnv-gnu', | ||
| 'PrgEnv-intel', 'PrgEnv-pgi'] |
There was a problem hiding this comment.
I don't think it make sense to run this test will all the programming environments. Just pick one.
|
|
||
| def compile(self): | ||
| # Here we set the target executable since by default the Makefile | ||
| # builds both cuda_memtest_sm13 and cuda_memtest_sm20 |
There was a problem hiding this comment.
Perhaps a comment here that sm20 is the maximum supported GPU architecture version from this test would help.
|
@jenkins-cscs retry daint kesch monch |
Closes #181