Skip to content

WIP: Improve 'mutrino' test run-times some (ATDV-131, TRIL-219)

James Willenbring requested to merge bartlettroscoe:atdv-131-atdm-fix-knl-runtime-settings into develop

Created by: bartlettroscoe

CC: @fryeguy52

These improvements are based on the work by Brad King at Kitware in:

Also see ATDV-131

How was this tested?

I ran this on 'mutrino' in some extensive testing in summarized in ATDV-131. The final commit to use ctest -j4 for both the builds was verified on 'mutrino' with:

$  env \
      Trilinos_PACKAGES=Panzer \
      CTEST_START_WITH_EMPTY_BINARY_DIRECTORY=FALSE \
      CTEST_DO_CONFIGURE=OFF \
      CTEST_DO_BUILD=FALSE \
    ./ctest-s-local-test-driver.sh intel-opt-openmp-HSW intel-opt-openmp-KNL
***
*** ./ctest-s-local-test-driver.sh  intel-opt-openmp-HSW intel-opt-openmp-KNL
***

ATDM_TRILINOS_DIR = '/home/rabartl/Trilinos.base/Trilinos'

Load some env to get python, cmake, etc ...

Hostname 'mutrino' matches known ATDM host 'mutrino' and system 'mutrino'
Setting compiler and build options for buld name 'default'
Using mutrino compiler stack INTEL to build DEBUG code with Kokkos node type SERIAL and KOKKOS_ARCH=HSW

Running builds:
    intel-opt-openmp-HSW
    intel-opt-openmp-KNL


Running Jenkins driver Trilinos-atdm-mutrino-intel-opt-openmp-HSW.sh ...


real    6m15.261s
user    0m2.926s
sys     0m0.814s

Running Jenkins driver Trilinos-atdm-mutrino-intel-opt-openmp-KNL.sh ...


real    14m15.213s
user    0m2.558s
sys     0m0.617s

which posted to CDash builds:

Merge request reports