Skip to content
This repository has been archived by the owner on Oct 23, 2020. It is now read-only.

Agenda Notes Tuesday January 10th 2017 11:00 AM

Jump to bottom
Michael Duda edited this page Jan 11, 2017 · 6 revisions

MPAS-Developer Telecon:

Date: 10 January 2017
Time: 11:00 MST / 19:00 CET

Call-in number: 1-866-740-1260
Access Code: 4978161

Agenda Topics:

  1. Framework modifications:
  2. Design Document Discussions
  3. Timekeeping issue affecting ACME runs starting in year 1850 has been narrowed down to mpas_calibrate_alarms routine
    • How to fix this?
  4. PIO performance issues with forcing
  5. Proposal for directory structure of test case infrastructure:
  6. Proposal to create a logging module to replace write statements to stdOutUnit, stdErrUnit
    • current approach has caused a lot of portability and performance issues in ACME
    • proposed approach would use logic to only write when necessary rather than writing unwanted messages to /dev/null or scratch files
  7. Perpetual to-do list:
  8. News/updates:
    • MPAS v5.0 was released on 7 January 2017

Notes:

Participants: Mark, Matt, Jon, Dom, Michael

Framework PRS:

  • PR #1192 can be expanded to introduce a general function for expanding a template based on the MPI comm size and MPI comm rank of the caller; this can be leveraged for log..abort as well as log..err/out files
  • Some merge testing still needed for rename of testing suite directory
  • PR #1124 - check to see if there is a performance impact (either positive or negative)
  • PR #1119 - we decided to open an issue and create a new PR to split the task of opening log files from the initialization of MPI; once this new PR has been created, we can close #1119
  • PR #1086 - discussed whether this can be expanded to allow arbitrary time instants to be specified for a stream's input or output "interval"

Several issues related to 8-byte integers in the ESMF time manager should be addressed through PRs to CIME. Progress on this will be discussed at next week's telecon.

For PIO performance, we should push forward on an algorithmic solution in the MPAS_seekStream routine with the expectation that there may not be an easy fix in PIO or the underlying netCDF library for the performance issue.

The idea of a logging module was well received. The introduction of such a module would be invasive, in that all cores would need to replace many existing write statements to stderrUnit or stdoutUnit, but the potential benefits are significant.

Clone this wiki locally