Skip to content

MetalWalls

This module tests the binary 'mw' in available modules containing substring 'MetalWalls'. Test input files are defined in MetalWalls's repo under hackathonGPU/benchmark*, see https://github.com/reframe-hpc/reframe/blob/develop/hpctestlib/sciapps/qespresso/benchmarks.py

ReFrame terminology:

"pipeline stages": https://reframe-hpc.readthedocs.io/en/stable/regression_test_api.html#pipeline-hooks

"test parameter": a list of values, which will generate different test variants. https://reframe-hpc.readthedocs.io/en/stable/regression_test_api.html#reframe.core.builtins.parameter

"test variant": a version of a test with a specific value for each test parameter https://reframe-hpc.readthedocs.io/en/stable/regression_test_api.html#test-variants

"concrete test cases": all test combinations that will actually run: - test variants - valid system:partition+programming environment combinations https://reframe-hpc.readthedocs.io/en/stable/tutorial_deps.html#listing-dependencies

Tests can be filtered by name, tag, programming environment, system, partition, or maintainer, see https://reframe-hpc.readthedocs.io/en/stable/manpage.html#test-filtering

Hooks acting on all possible test combinations (before filtering) are called after the 'init' stage. Hooks acting on concrete test cases (after filtering) are called after the 'setup' stage.

See also https://reframe-hpc.readthedocs.io/en/stable/pipeline.html

EESSI_MetalWalls_MW

Bases: MetalWallsCheck

MetalWalls benchmark tests.

MetalWalls <https://gitlab.com/ampere2/metalwalls>__

request_mem()

Request memory per node based on the benchmark.

run_after_init()

Hooks to run after the init phase

run_after_setup()

Hooks to run after the setup phase

set_binding()

Set binding to compact to improve performance reproducibility.

set_increased_walltime()

Increase the amount of time for the largest benchmark, when running with few cores.

set_omp_num_threads()

Set number of OpenMP threads via OMP_NUM_THREADS. Set default number of OpenMP threads equal to number of CPUs per task.

set_tag_ci()

Set tag CI on smallest benchmark, so it can be selected on the cmd line via --tag CI

skip_max_corecnt()

Skip tests if number of tasks per node exceeds maximum core count.