Tag: technical

Minutes

Date: 13th June 2017
Attendees:

• Marshall Ward (NCI, Chair)
• Aidan Heerdegen (ARCCSS ANU)
• Scott Wales (ARCCSS Melbourne Uni)
• Nicholas Hannah (ARCCSS/Double Precision)
• Russ Fiedler and Matt Chamberlain (CSIRO Hobart)
• Justin Freeman (BoM)
• Peter Dobrohotoff and Roger Bodman (CSIRO Aspendale)

COSIMA Workshop

•  Marshall: Good meeting. A couple of main messages:
  1. Must have convergence of source code
  2. System for efficient sharing of model configurations
• First item we are well on the way to achieve. Second has never been addressed.

Configuration Sharing

• Marshall: Should focus our energies on configuration sharing.
• Marshall: having configurations in OceansAus is close to what we want?
• Currently when using payu configs are saved into local git repos, free floating. This could work for everyone. Should gather config files for runs.
• Aidan: can have a GitHub organisation and push configurations there, or push to own GitHub accounts and tag repositories to allow search and discovery. Use OceansAus or dedicated GitHub organisation for more canonical configurations.
• Peter agreed it is ok to have many thousands of configurations. Scott agreed as long as there is a canonical version that is easily identifiable for others to fork from.
• Nic felt that unless there was a good pay off most individuals would not make the effort to push their configurations up to shared spaces, or make the visible to others, but there is a pay-off for everyone if configurations are shared.
• Nic: How do create a searchable index or database for others to find the stuff? May be bigger problem than storing data.
• Marshall: Payu users can be forced to do it. Tracking history not seem such a big deal but a pay off for users down the track
• Pete: if someone wants to use my config I give them my config and people branch off that. rosie-graph will show relationships between them.
• Marshall: rose solves a lot of these problems already. We don’t benefit from that. Do we need to just use rose or emulate what it does best?
• Scott: rose setup: metadata file in each repo branch, hook within repo which reads meta data file and adds to database. Currently there is not a lot of meta data in most branches.
• Justin: does rose use git? Scott: should be back-end agnostic. Nic: possible solution? Use GitHub for configs, but use rose DB to share and index configurations?
• Marshall: should we draft a plan?
• Nic: Just try with rose? Gives us DB which is searchable and viewable. Doesn’t stop anyone from working in the way we want?
• Marshall: Do we need accessdev to do this? Scott: I think it is a just a sqlite DB, so can just be on filesystem. Hooks in GitHub repo won’t work the same way to update it however.
• Marshall: just start with something searchable? Nic: yeah, see what people are running.
• Scott: Just start with a simple list. Nic: would want a GitHub repo and commit hash in meta data as a minimum.
• Justin: make notes about what we think is possible. Common set of functionality will guide us. Rose sounds good, but may be too large a solution? Want something that won’t get in my way and lightweight. Don’t know too much about rose. Everyone agrees this is a good idea.
• Scott: can just have a list of experiments. Don’t need the editor. Will send around link so Justin and others can take a look.
• Marshall: really need a bunch of git repos and a way to organise them. Justin: metadata would be a great, to find things without bothering others.
• Marshall: how do we share input data?
• Justin: share through RDS? Jingbo working on this.
• Nic: Justin said all our data sources should be pointing to URLs. Great idea! Never have to care about where data is coming from. Input through THREDDS URIs. Marshall: too ambitious? If you’re using netCDF interface, the library takes care of network. Should all happen under netCDF. Model thinks it is a file.
• Aidan: slower? Nic: might not be slower?

COSIMA Models

• Marshall: talking with Peter and Roger at optimisation meetings, seem to be bit-reproducibility issues with CM2. Has Nic ever looked at bit repro in OM2?
• Nic: have checksums, and make sure they don’t change with code changes under same layout etc. haven’t noticed things aren’t reproducible in a typical short run. As soon as processor layout changes you’ll run into issues.
• Peter: if run model 3 months and then 1 month at a time, get different answers. A few degrees over a few months. Restarts are not reproducible. CMIP5 worked ok. Now doesn’t work. Maybe need to check that the ice dumping ok.
• Marshall: have faith in MOM5 core. Issues with flux exchange reproducibility. Anyone got issues with that?
• Russ: Fabio Dias and Russ have been working on an energy leakage issue. MOM-SIS conserves, MOM-CICE has small leakage. Is a computational issue. ASCII diagnostics don’t close. Should close 10^-12 W/m2, only 10^-6 W/m2.
• Marshall: Has CM2 been subjected to same scrutiny for flux exchange. How far has Dave Bi got?
• Peter: does UM have this issue? Scott: bit repro is heavily tested. Processor decomp and run length shouldn’t be an issue.
• Peter: maybe CICE is a bit more vulnerable? Nic: if someone hasn’t made an effort to make it reproducible, odds on that is isn’t. As far as Nic knows no checking has been done on the ACCESS-CM/OM specific boundary code. Suspicious this might be the issue.
• Roger: reproducibility tends to be over the higher latitudes and higher altitudes. A couple of degrees over a few months.
• Marshall: MOM5 is not reproducible over processor layout changes north of tripole. There is an expensive operation in the MOM5 flux exchange that isn’t turned on by default. Scott: MPI reduce is not generally reproducible unless special steps taken.
• Nic: repro needs to be tested with correct compiler flags turned on. Normally optimised code will not reproduce. Roger did this, but will double check this for CICE. Should have fpmodel precise across all 3 models.
• Russ: fixed ice salinity issue that Nic had heard of before from Dave Bi. Was a default namelist option that people don’t often change. Nic: will forward the conversation to Roger.
• Marshall: if we routinely shared configs this would not be an issue.

Actions

New:

• Create document outlining options for configuration sharing (?)

Existing:

• Nic to present MATM code re-write proposal to TWG for feedback before sign-off. Will then be presented to Andy Hogg for approval.
• Nic create a discussion document (on COSIMA?) to document current approaches and strategies for future
• Move FMS to submodule of MOM5 github repo (Marshall). Liase with Nic on implementation?
• Test Nic’s access-om model config on OceansAus (All)
• Work up test cases to cover the nudging code (Justin, Mirko) and supply them to Nic.
• Add new test cases to Jenkins test suite (Nic).
• Start a new google doc about coupler issues and MATM (Marshall)
• Ask Dale Roberts about effects of OpenMP for Roger (Marshall)
• Make a proper plan for model release — discuss at COSIMA meeting. Ask students/researchers what they need to get started with a model (Marshall and TWG)
• Blog post around issues with high core count jobs and mxm mtl (Nic)
• Do longer runs with Nic’s 1 deg and 0.25 deg ACCESS-OM2-JRA55 configs (Andy and Aidan)
• Try repeat year forcing with Nic’s configurations (Nic and Andy)

Minutes

Date: 11th April 2017
Attendees:

• Marshall Ward (NCI, Chair)
• Aidan Heerdegen (ARCCSS)
• Nicholas Hannah (ARCCSS/Double Precision)
• Russ Fiedler and Matt Chamberlain (CSIRO Hobart)
• Justin Freeman and Mirko Velic (BoM)

Updates on previous actions

COSIMA Workshop

• Agreement that we want to present some summary of the work of the TWG to date. Important to show what we’ve done: Model sharing, Grid sharing, Important infrastructure activities.
• Discussion around what to put cover in the presentation: What are the most important projects? What is most interesting? What do we want to work on?
• Some ideas for important topics:
  1. Coherent sharing of experiments with others, as is possible with Rose/cylc
  2. Performance is always important, can motivate others to support our work
  3. Coherence around OASIS.
• Nic will present a 1/10th coupled model update and results. Should have something decent by then
• Need to raise issue of the coupler, present some options and get some feedback from the stakeholders. Justin is concerned about wave coupling in future. Inform wider group about what we’re thinking. Judge appetite for changing tools.
• OASIS is not well liked, is difficult to use and performs poorly.
• If someone has concerns about the coupler/MATM then we need to back that up with numbers. Marshall will start a new google discussion doc for this
• Marshall has CM2 numbers that are difficult to understand. Probably not worth presenting to COSIMA
• Aidan not presenting at present. Russ might talk about grid stuff. Matt might talk about some of the technical aspects of 0.25 deg run. Maybe parameterisation of bulk formulas.
• There was a general feeling the TWG was functioning well and no significant change required. Matt suggested a face to face meet-up might be worthwhile. Mirko pointed out we need some metrics to show to others that TWG is a worthwhile venture and why we’re so pleased with ourselves

Updates

• Nic has had 3 week break. Visited GFDL. Presented his coupler at a coupling workshop. Talked to OASIS, MCT, ESMF and YAC (Yet Another Coupler). ESMF are building 2nd order conservative support into their remapping weight software which is good for us.
• Matt: has been working with GFDL coupled model for Terry O’Kane. Using ACCESS 1 deg grid in CM2 with AM2 atmosphere. Several 100 years. Decadal forecasting. Atmosphere is same as CM2.1. Maybe 2.5 deg. Would be nice to have higher resolution, but for forecasting need ensembles. 10-40 ensembles. UM is too slow for this sort of work. There is some innterest in benchmarking for ocean structure, maybe against CM3. Maybe the go to 0.25 ocean, AM3 atmosphere model. Then put through data assimilation system. Then put in forecast mode. Look for predictive skill.

COSIMA Models

• Not made much progress on ACCESS-OM2-01. Technically coupled. Now trying to speed it up for multi-year runs. Currently running 600s timestep. Still in January.
• Not currently JRA55 forced. Can’t talk about tenth ACCESS-OM2 performance yet.
• CICE has some shocks on spin-up. Was aiming for 600s, but the current operational time step is 450s. Pulling back to 450s will be a lot easier. Goal to make 600s the prroduction time step. Russ can sometimes run 720s or even 900s for this tenth simulation.
• Matt pointed out it is difficult to start a simulation with CICE when initial conditions have no sea ice. Better to create initial conditions from World Ocean Atlas.
• Matt is now running COREII IAF. Sometimes blows up, just drops time step for a month, and if no issue, doesn’t try to diagnose problem, just goes back to normal time step. Matt is running 600s with a tenth MOM/SIS.
• Aidan talked about issues with the current ACCESS-OM2-025+JRA55, discovered the scripts for MOM5 are not using the same OASIS build as the other components. Trying to work out why the model is running so slowly.

Actions

New:

• Start a new google doc about coupler issues and MATM (Marshall)

Existing:

• Nic to present MATM code re-write proposal to TWG for feedback before sign-off. Will then be presented to Andy Hogg for approval.
• Add Peter’s CICE5.1 config to OceansAus github repo (Nic and Peter)
• Port MOM5 build system to cmake (Aidan)
• Push updated MATM code with JRA-55 support to OceansAus github (Aidan)
• Nic create a discussion document (on COSIMA?) to document current approaches and strategies for future
• Move FMS to submodule of MOM5 github repo (Marshall). Liase with Nic on implementation?
• Test Nic’s access-om model config on OceansAus (All)
• Work up test cases to cover the nudging code (Justin, Mirko) and supply them to Nic.
• Move to CICE5 on OceansAus repo (Nic).
• Add new test cases to Jenkins test suite (Nic).

Minutes

Date: 14th March 2017
Attendees:

• Aidan Heerdegen, Scott Wales (ARCCSS, Acting Chair)
• Nicholas Hannah (ARCCSS/Double Precision)
• Russ Fiedler and Matt Chamberlain (CSIRO Hobart)
• Peter Dobrohotoff (CSIRO Aspendale)

Updates on previous actions

• MATM update: Nic will do a code review and document probably beginning of April.
• CICE5.1 on repo: This will be part of the ACCESS-OM-0.1 development when this code is incorporated into the model
• Matt has 1/10th data, and diagnostics were ok. Needed them to explore how heat is going into the ocean, and how that is affected by bulk formula and forcing fields. Could be useful analyses for changing to JRA55.

COSIMA Models

• Nic time-stepping ACCESS-OM-0.1 with core forcing. Will not be working on this again for about a month. Is time stepping with CORE forcing CICE 4.1. Will add in cice5 and JRA55. Took about as long as he thought, but there were a number of issues that required fixing.
• Nic wants to cleanup coupling code, remove cruft and commented out code. Also clean up date handling. There are twenty places where you have to change time step for example. Similar for run time. Need to set up regression tests to allow code changes without breaking stuff. In MOM and CICE namelists there are duplicate fields. Need code changes. Maybe C-style macros are a simple approach?
• Nic Has developed some tools for generating model inputs required for ACCESS-OM. For example, creating the cice grid out of ocean grid. Using ESMF for creating grids and restarts. Started a wiki to document steps to develop config.

Updates

• Russ is working on some mixing models, more complex than KPP. GOTM was patched on to MOM. Have to advect around a couple of extra tracers but doesn’t seem to impact speed much. Supposed to be a lot better way of mixing.
• Russ has been having issues with number of retries time outs with PBS, occurs in first few lines of code. Aidan suggested noting error message and PBS ids and passing on to NCI Help, as this is often the side effect of a bad node.
• Nic also having issues with tenth. Crashes, with lots of MPI output and no error code. A lot of the processors were sitting in an MPP_global call. ACCESS-OM coupling code was making it fall over. Raijin doesn’t like MPP_global operations as implemented by FMS. Due to implementation issues. To globally distribute an array in FMS you use MPP_global, which uses multiple MPI_Sends, rather than MPI_broadcast.
• Peter: status is similar to last month’s update.
• Russ has found issues with MLD diagnostic in MOM. If only have 3 levels returns zero. Strange issues near coast, getting zeroes at those locations.
• Russ will be working on bathymetry for COSIMA 1/10th configuration. Maybe need different bathymetry for climate and BRANS type runs? Aidan was adamant the goal was to have the same bathymetry for all COSIMA models
• Russ is running his 1/10th degree simulation with a 900s time step, whereas Aidan’s configuration is 450s due to instability in Arctic
• JRA-55 is now being housed centrally. All those who wish to use it can they please make their requirements (versions, update frequency) known to the ARCCSS CMS team, and Paola Petrelli in particular

MOM5 Repo Move

• Aidan has had some preliminary discussion with Stephen Griffies and Alistair Adcroft about having a new “official” home for the MOM5 source code repository. They favour a separate repo with the community supported model, and infrequent but regular updates of the “Official GFDL” repo version, which could be used by those needing a badge of officialness.
• Nic felt that there was little point in having an “official” repo, especially if it just creates more work for little actual gain.
• Nic was in favour of a standalone MOM github organisation, as it is bigger than any one of the groups that use it. It could also host all the older versions of MOM also. This was considered a good approach. If all partner organisations (COSIMA, ARCCSS, BoM, CSIRO, GFDL?) then gave this github organisation their support, similar to the CICE development model, this could tick the boxes for those that required an endorsed software product.

Actions

New:

• Nic to organise transferring MOM code repos to a new organisation MOM-ocean

Existing:

• Nic to present MATM code re-write proposal to TWG for feedback before sign-off. Will then be presented to Andy Hogg for approval.
• Add Peter’s CICE5.1 config to OceansAus github repo (Nic and Peter)
• Port MOM5 build system to cmake (Aidan)
• Push updated MATM code with JRA-55 support to OceansAus github (Aidan)
• Get licensing for MOM5 input files (Marshall)
• Work on hosting MOM5 input files on NCI THREDDS server (Marshall, Aidan)
• Nic create a discussion document (on COSIMA?) to document current approaches and strategies for future
• Move FMS to submodule of MOM5 github repo (Marshall). Liase with Nic on implementation?
• Test Nic’s access-om model config on OceansAus (All)
• Work up test cases to cover the nudging code (Justin, Mirko) and supply them to Nic.
• Move to CICE5 on OceansAus repo (Nic).
• Add new test cases to Jenkins test suite (Nic).

Minutes

Date: 14th February 2017
Attendees:

• Marshall Ward (NCI, Chair)
• Aidan Heerdegen, Scott Wales (ARCCSS)
• Nicholas Hannah (ARCCSS/Double Precision)
• Justin Freeman and Mirko Velic (BoM)
• Russ Fiedler and Matt Chamberlain (CSIRO Hobart)
• Peter Dobrohotoff (CSIRO Aspendale)

COSIMA Models

• Nic is working on a COSIMA configuration with 1/10th degree ocean. Currently setting up all the files for a coupled ice/ocean. Using the existing ARCCSS 1/10th degree MOM5 ocean, but there is no 1/10th degree CICE. Have not previously coupled 1/10th MOM5 with OASIS. Plan is to use JRA-55 forcing.
• OASIS won’t work as it has in the past, creating remapping weights and regridding forcings on to a 1/10th grid. Too slow. Currently takes 2-3 hours for the 0.25 deg grid. They use SCRIP for regridding, which has limitations: uniprocessor, conservative remapping not accurate in tripolar, Not quantified. Not too bad for 1/0.25 deg.
• New method is to use ESMF. Doesn’t do 2nd order conservative remapping. First order conservative ok for now. Maybe revisit. Nic has set up test and evaluation cases to compare SCRIP and ESMF.
• OASIS isn’t that great at regridding/remapping. If we’re using another tool for remapping, what is OASIS doing for us? Nic has written an OASIS replacement: TANGO. Simple and basic as possible. Uses ESMF. Two step process: generate regridding weights, and then distribute fields. Might be faster than OASIS, but performance isn’t a big issue for this.
• Nic is writing a model configuration development toolchain? Supports different models. Documented here.
• Justin agreed that OASIS is poorly designed, difficult to work with and not well liked. Suggestions to replace it have met with resistance from BoM. Might be able to push something under the TWG umbrella.
• Aidan unhappy with MATM. Code has been changed ad-hoc to support new forcing datasets. Nic is happy to rewrite MATM it if Andy Hogg wants to. TANGO supports python/fortran/C++. Maybe re-write MATM in python? Not performance critical. Aidan not a fan of introducing a new dependency on python.
• Aidan: working on ACCESS-OM2-025+JRA55. Problems with generating iced restart files.
• Marshall: Working on CM2 with Peter’s help. ROSE suite needs some work, but in a reasonable state. Just profiling. Slight bug in OASIS restart file generation. Model would hang occasionally. Just a random hang. Due to a bug in MOM that Marshall had fixed. Source code had been reverted to a version that was not patched. Dave Bi had earlier described a bug that might be the same one?
• Marshall has Profiler working in all 3 sub models. Some prelim numbers.
• Need to update MOM source inside CM2. Use main trunk version and see what blows up. Marshall and Peter to work offline on getting their version up to date.
• Do we need ACCESS branch on MOM repo? Definitely want everyone on the same repo as there are issues with some people using out of data code. Can we have more rigorous tagging of bug fixes for example? Which would allow bug-fixes to be incorporated in other versions. Nic thought that now we have better communication this may not such a problem?

Updates

• Peter is working on ACCESS-CM2 for CMIP6. Currently focussing on global atmosphere. MetOffice are atmosphere only, and incorporate into their coupled model. GA7.1 is next target for ACCESS-CM2. Not currently with CABLE. Prelim testing with GA7.1. PI will wait for UM10.6 for correct aerosol coding. Martin is focussing on 10.6. Won’t have a version of CM2 with 10.6 for a while. CABLE group working with 10.5, and always difficulty with incorporating CABLE in UM versions. CSIRO still doing present day in 10.3.
• Russ found a resolution error when packing data that ended up destroying a 20 month assimilation run. Will be a complete spin up when restarted
• Justin has been doing a lot of interpolation. Puddles natural neighbour library. Been using it a lot. Very flexible and nice library
• Matt has provided WOMBAT code to Paul Spence so he can recompile with new libraries and with flags appropriate for new broadwell hardware on raijin.
• Since shutdown Matt has also recompiled. Asks for 850 cpus, 6-7 hours / year. When processor count > 1.5K crashes too often.
• Marshall: NCI trouble shooted 1.8.4 problems. MPI reduction operations no longer work 1.8.4.
• Marshall: openmpi 2.0 is slower than 1.10.2.
• Aidan reiterated: don’t use -03 with intel MPI.
• Justin: BoM internal meeting regarding COSIMA project. All signed off. Developments over 1-2 year timeframe. Wavewatch 3 coupled into COSIMA code. Stefan and Mirko will be looking at that. Would like some feedback from TWG about assessing it, and getting back to the community. MOM+wavewatch coupled initially as a technical demo. Wavewatch 3 will be part of COSIMA. GFDL is interested in Wavewatch. Stephen Griffies was curious. People have coupled wave watch with MOM in past, so know it is possible. Maybe OASIS-MCT, or OASIS3. These were climate scale runs. Mark Hannah’s postdoc did this work.
• Marshall: about github source for MOM under Breakaway Labs. Only custodian is Nic. Need some shared ownership of these codes? Move to GFDL? Justin: as an organisation BoM will look at the code more favourably if it is sitting under GFDL.

Actions

New:

• Marshall will be away, Aidan to organise next meeting.
• Update MOM source inside CM2 (Marshall).
• Nic to present MATM code re-write proposal to TWG for feedback before sign-off. Will then be presented to Andy Hogg for approval.

Existing:

• Add Peter’s CICE5.1 config to OceansAus github repo (Nic and Peter)
• Port MOM5 build system to cmake (Aidan)
• Push updated MATM code with JRA-55 support to OceansAus github (Aidan)
• Get licensing for MOM5 input files (Marshall)
• Work on hosting MOM5 input files on NCI THREDDS server (Marshall, Aidan)
• Nic create a discussion document (on COSIMA?) to document current approaches and strategies for future
• Move FMS to submodule of MOM5 github repo (Marshall). Liase with Nic on implementation?
• Test Nic’s access-om model config on OceansAus (All)
• Work up test cases to cover the nudging code (Justin, Mirko) and supply them to Nic.
• Move to CICE5 on OceansAus repo (Nic).
• Add new test cases to Jenkins test suite (Nic).
• Aidan to provide Matt with location of tenth model test data. Check if capturing all the diagnostics Matt might be interested in.
• Matt to provide Marshall with some test cases for the Xeon Phi test cases, maybe 1 deg configurations.