This repository was archived by the owner on Oct 23, 2020. It is now read-only.
-
Notifications
You must be signed in to change notification settings - Fork 0
Agenda Notes Monday October 24th 2016 3:00 PM
Mark Petersen edited this page Oct 25, 2016
·
14 revisions
Date: 24 October 2016
Time: 3:00pm MDT
Call-in number: 1-866-740-1260
Access Code: 4978161
- Framework modifications:
- Addition of new build targets: PR #1087
- New stream output_interval "final_only": PR #1086
- Critical error log files: "regularize" filenames, and add ability to print multiline messages
- See proposed changes in this commit
- Integer overflow in ESMF time interval code
- Fortran 2008 CONTIGUOUS keyword
- Alternate I/O layers (See, e.g., framework/nopio-clean
- potential PR: a new PR in ACME's MPAS repo wants to remove the if-construct from around this bit of code in dmpar: https://github.com/MPAS-Dev/MPAS/blob/develop/src/framework/mpas_dmpar.F#L260-L265 because mpi is initialized within ACME. Confirm this doesn't affect standalone MPAS functionality.
- Design Document Discussions
- Interest in exploring cmake build system
- Transfer/maintenance of MPAS build notes that Doug had created: https://github.com/douglasjacobsen/mpas_machine_instructions Maybe move to MPAS-Dev/MPAS wiki?
- On a related note, discuss places to put 'random' documentation. E.g., https://docs.google.com/document/d/1aURr4UdEm0AMeYlc478yL_46biD19naVp3nCMYgrYLI/edit?usp=sharing; how to run the test case infrastructure on LANL's Wolf machine; how to set up an optimal git environment on a new machine
- Consider moving telecons to earlier time to allow easier participation by Xylar in Germany?
- Matt found another bug in the interval division routine that had caused the problem before (the version that was in ACME before fixed it). It's fixed on develop, but maybe we want to add a new test to the test core? Symptom: set a stream's file_interval to 200 years. Files got written at 200, 400, 800, 1600, 3200 years instead of 200, 400, 600, 800,... (But all the time levels written were correct.)
- test case infrastructure - the "temporary" location in test_cases/ocean isn't so temporary anymore. Do we want to move it?
Re: cmake
- Doug and Andy Salinger have both experimented with converting parts of MPAS to cmake. Registry was a problem (compile an executable, then run it, then compile other libraries and executable). Andy thinks it is a challenge but not insurmountable.
- Previously, having cmake as an additional dependency was considered a negative to pursuing it. It the required version of cmake was stable over multiple MPAS releases, that might alleviate that concern somewhat.
- Another potential concern is the familiarity and comfort of developers and users with cmake files vs. Makefiles.
- Andy introduced the ACME Software Modernization project which has resources to convert most or all of ACME to cmake to achieve consistency across the project and speed up compile times. Resources to extend this to MPAS are probably available. Project has started but availability to work on MPAS build system would probably not be for another 3-6 months.
- Andy noted that cmake build system can sit alongside Makefile system during transition, but maintaining both is not a good long term solution.
- Plan: MPAS developer community to consider cmake. ACME SM project to 'cmake-ify' an example directory or MPAS tool to demonstrate how cmake works.
Re: machine build instructions
- Mark P. to transfer the machine build instructions from Doug's repo to the MPAS wiki pages.
Re: telecon time
- New telecon time of 11am Mountain Time to be tried next week to facilitate participation of participants in Europe.
Re: #8, test_cases/ocean directory
- Mark P and others can think of a name that is more functionally descriptive, as 'ocean' was a temporary name, and other cores may put in other types of setup/testing in the test_cases directory. The name 'test_cases' could be changed as well, as this directory is used for set-up of full-scale initialization as well as idealized test cases.