espresso
This module tests ESPResSo in available modules containing substring 'ESPResSo' which is different from Quantum Espresso. Tests included: - P3M benchmark - Ionic crystals - Weak scaling - Strong scaling Weak and strong scaling are options that are needed to be provided to the script and the system is either scaled based on number of cores or kept constant. - LJ benchmark - LB benchmark
EESSI_ESPRESSO_LB
¶
Bases: EESSI_ESPRESSO_base, EESSI_Mixin
assert_completion()
¶
Check completion
assert_convergence()
¶
Check convergence
required_mem_per_node()
¶
LB requires 250 MB per core
set_executable_opts()
¶
Set executable opts based on device_type parameter
skip_tests_module_version_LB()
¶
The LB module versions need to be >= 5.0.0 or a expermental release version which includes walberla. The earlier lb method does not scale beyond 16 MPI tasks and is extremely slow in terms of case setup. Assumption: 1. Versions with commit hashes have walberla in them. If not then they will not be filtered here and will run.
EESSI_ESPRESSO_LJ_PARTICLES
¶
Bases: EESSI_ESPRESSO_base, EESSI_Mixin
EESSI_ESPRESSO_P3M_IONIC_CRYSTALS
¶
Bases: EESSI_ESPRESSO_base, EESSI_Mixin
EESSI_ESPRESSO_base
¶
filter_scales()
¶
Filtering function for filtering scales for P3M test and the LJ test. This is currently required because the 16 node test takes way too long and always fails due to time limit. Once a solution to mesh tuning algorithm is found, where we can specify the mesh sizes for a particular scale, this function can be removed.