Skip to end of metadata
Go to start of metadata

You are viewing an old version of this page. View the current version.

Compare with Current View Page History

« Previous Version 7 Next »

Every ACME developer should be able to run the acme_developer test suite on at least one machine to which they have convenient access so that they can test their changes before submitting a pull request.

You have to first follow the instructions in Installing the ACME Model on a cluster or HPC platform

Adding more restart files.

The tests needs restart files not normally used in model runs.

  1. Populate restart files in to the inputdata directory (these do not get downloaded automatically by the tools as of beta10)  
    1. cd $ACMEHOME/inputdata * svn co https://svn-ccsm-inputdata.cgd.ucar.edu/trunk/inputdata/ccsm4_init

Installing individual tests

  1. Once the model is installed, you can immediately run individual tests.
    1. cd $ACMEHOME/scripts
    2. ./create_test -testid t01 -testname ERS_D.f10_f10.ICLM45BGC.edison_intel

Installing the acme test suites on a new machine

A testlist needs to be created for your new machine.  Start with the acme_developer tests.  This step can be skipped if you are running on a supported machine.  This is code development and should be done on a branch following the Development Quick Guide

  1. Add <ACME>/scripts/acme to your PATH
  2. Cd to <ACME>/scripts/ccsm_utils/Testlistxml
  3. Run: update_acme_tests acme_developer ./testlist.xml -p <machine>,<compiler>

Installing test baselines with first run of a test suite.

Baselines are similar to the input data directory in that there should ideally be one of them per machine that all the ACME developers on that machine can read from.  There only needs to be, for example, one baseline made from ACME v0.2 and the pgi compiler on Titan.   The data sizes  can get large because history files are kept for all components for dozens of (short) runs.  All ACME developers on Titan can then compare against those baselines if they expect their changes to be BFB.

Baselines should be updated when an answer-changing or namelist-changing modification has been made to master.  That should be obvious to anyone following github comments but we may need another way to broadcast it.  Local developers then have to know the new path for the baseline to compare against.  How this will be managed is all still being worked out.

Note: Baselines will accumulate over time so there needs to be some rule for how many to keep.  CSEG has settled on the 5 most recent (beta) tags.

The first run of acme_developer can generate baselines, all subsequent runs should compare against these baselines

  1. cd <ACME>/scripts
  2. ./create_test -xml_category acme_developer -xml_mach <machine> -xml_compiler <compiler> -testroot <scratch> -project <batch project> -generate <some-name-for-these-baselines> -baselineroot <path to baselineroot>
    1. <scratch> is where the tests case roots will be created.
  3. cd <testroot>
  4. <ACME>/scripts/acme/wait_for_tests */TestStatus
    1. Note: The wait_for_tests script requires a minimum python version number of 2.7.
  5. All tests should pass

Filter by label

There are no items with the selected labels at this time.

  • No labels