-
Notifications
You must be signed in to change notification settings - Fork 0
Agenda Notes Monday December 5th 2016 11:00 AM
Date: 5 December 2016
Time: 11:00 MST / 19:00 CET
Call-in number: 1-866-740-1260
Access Code: 4978161
- Framework modifications:
- Design Document Discussions
- Timekeeping issue affecting ACME runs starting in year 1850 has been narrowed down to
mpas_calibrate_alarms
routine- How to fix this?
- PIO performance issues with forcing
- Proposal for directory structure of test case infrastructure:
- see https://github.com/MPAS-Dev/MPAS/pull/1160
- renames: test_cases/ocean -> testing_and_setup/compass
- Perpetual to-do list:
- Integer overflow in ESMF time interval code
- Fortran 2008 CONTIGUOUS keyword
- Exploring cmake build system
- News/updates:
- MPAS v5.0 release branch has been created (release-v5.0) with a planned release on 9 December. PRs for cleanup and bugfixes for MPAS v5.0 should target the
release-v5.0
branch.
- MPAS v5.0 release branch has been created (release-v5.0) with a planned release on 9 December. PRs for cleanup and bugfixes for MPAS v5.0 should target the
-
No objections to the new testing directory structure or the renaming of infrastructure to COMPASS.
-
For PR #1199, we may run into problems with block streams, which make calls to
mpas_dmpar_init
. It seems that we're overloading thempas_dmpar_init
to initialize the message-passing layer as well as to redirect standard error and output to log files. One possible solution may be to split the logic for redirecting log files into a new routine. See discussion for PR #1199. -
We'll hold off on code changes to address timekeeping issues in ACME until after ACME has been updated to the latest CIME.
-
Slow forcing I/O performance seems to be a result of terrible performance in PIO when reading
xtime
variables inMPAS_seekStream
(see details in Issue #1156). Using more clever search logic inMPAS_seekStream
may help to reduce the number of times that thextime
variable needs to be read; but, we should also follow-up with PIO developers, since even a small number of reads causes a measurable performance hit.