Update cgridDEV to main #8d7314f (CICE6.3.1 release) #64
Conversation
This file contains bidirectional Unicode text that may be interpreted or compiled differently than what appears below. To review, open the file in an editor that reveals hidden Unicode characters.
Learn more about bidirectional Unicode characters
…m#665) * Fix some raketests for izumi * fix some rake tests
…ICE-Consortium#667) When 'make' is invoked on the CICE Makefile, the first thing it does is to try to make the included dependency files (*.d) (which are in fact Makefiles themselves) [1], in alphabetical order. The rule to make the dep files have the dependency generator, 'makdep', as a prerequisite, so when processing the first dep file, make notices 'makdep' does not exist and proceeds to build it. If for whatever reason this compilation fails, make will then proceed to the second dep file, notice that it recently tried and failed to build its dependency 'makdep', give up on the second dep file, proceed to the third, and so on. In the end, no dep file is produced. Make then restarts itself and proceeds to build the code, which of course fails catastrophically because the Fortran source files are not compiled in the right order because the dependency files are missing. To avoid that, add a dependency on the dep file to the rules that make the object file out of the Fortran source files. Since old-fashioned suffix rules cannot have their own prerequisites [2], migrate the rules for the Fortran source files to use pattern rules [3] instead. While at it, also migrate the rule for the C source files. With this new dependency, the builds abort early, before trying to compile the Fortran sources, making it easier to understand what has gone wrong. Since we do not use suffix rules anymore, remove the '.SUFFIXES' line that indicates which extension to use suffix rules for (but keep the line that eliminates all default suffix rules). [1] https://www.gnu.org/software/make/manual/html_node/Remaking-Makefiles.html [2] https://www.gnu.org/software/make/manual/html_node/Suffix-Rules.html [3] https://www.gnu.org/software/make/manual/html_node/Pattern-Rules.html#Pattern-Rules
* update parsing scripts to improve robustness, fix multi-pe advection=none * Update cice script to improve performance including minor refactoring of parse_namelist and parse_settings to reduce cost and ability to use already setup ice_in file from a prior case in the suite. Added commented out timing ability in cice.setup. Change test default to PEND from FAIL. * fix cice.setup for case * add sedbak implementation to support Mac sed * s/spend/spent
* add debug_model feature * add required variables and calls for tr_snow
* Adding method to write erroneous namelist options * Remove erroneous comma in abort_ice for namelist check * Added check for zbgc_nml. I missed that namelist in this file. * Added space and colons for namelist error output * Added space and colons for namelist error output Co-authored-by: David A. Hebert <[email protected]>
* updated orbital calculations needed for cesm * fixed problems in updated orbital calculations needed for cesm * update CICE6 to support coupling with UFS * put in changes so that both ufsatm and cesm requirements for potential temperature and density are satisfied * update icepack submodule * Revert "update icepack submodule" This reverts commit e70d1ab. * update comp_ice.backend with temporary ice_timers fix * Fix threading problem in init_bgc * Fix additional OMP problems * changes for coldstart running * Move the forapps directory * remove cesmcoupled ifdefs * Fix logging issues for NUOPC * removal of many cpp-ifdefs * fix compile errors * fixes to get cesm working * fixed white space issue * Add restart_coszen namelist option * Update NUOPC cap to work with latest CICE6 master * nuopc,cmeps or s2s build updates * fixes for dbug_flag * Update nuopc2 to latest CICE master * Fix some merge problems * Fix dbug variable * Manual merge of UFS changes * fixes to get CESM B1850 compset working * refactored ice_prescribed_mod.F90 to work with cdeps rather than the mct data models * Fix use_restart_time * changes for creating masks at runtime * added ice_mesh_mod * implemented area correction factors as option * more cleanup * Fix dragio * Fix mushy bug * updates to nuopc cap to resolve inconsistency between driver inputs and cice namelists * changed error message * added icepack_warnings_flush * updates to get ice categories working * updates to have F compset almost working with cice6 - still problems in polar regions - need to resolve 253K/cice6 versus 273K/cice5 differences * changed tolerance of mesh/grid comparison * added issues raised in PR * Update CESM-CICE sync with new time manager * Add back in latlongrid * Add new advanced snow physics to driver * Fix restart issue with land blocks * Update mesh check in cap * fix scam problems * reintroduced imesh_eps check * Put dragio in the namelist instead * Remove redundant code * Fix some indents Co-authored-by: Mariana Vertenstein <[email protected]> Co-authored-by: apcraig <[email protected]> Co-authored-by: Denise Worthen <[email protected]>
* Add CESM1_PIO for fill value check * Revert PIO_FILL_DOUBLE change for now
…nsortium#677) - Remove recent update to namelist read that traps bad lines, it conflicts with flexibility to read groups in random order picked up by NAG. - change print* statements to write(nu_diag,*)
* Add narwhal intel, gnu, cray, aocc Add perf_suite.ts * update narwhal_cray and perf_suite
* Add narwhal intel, gnu, cray, aocc Add perf_suite.ts * update narwhal_cray and perf_suite * Review and update OMP implementation - Fix call to timers in block loops - Fix some OMP Private variables - Test OMP Scheduling, add SCHEDULE(runtime) to some OMP loops - Review column and advection OMP implementation - ADD OMP_TIMERS CPP option (temporary) to time threaded sections - Add timer_tmp timers (temporary) - Add omp_suite.ts test suite - Add ability to set OMP_SCHEDULE via options (ompscheds, ompscheds1, ompschedd1) * - Review diagnostics OMP implementation - Add timer_stats namelist to turn on extra timer output information - Add ICE_BFBTYPE and update bit-for-bit comparison logic in scripts - Update qc and logbfb testing - Remove logbfb and qchkf tests, add cmplog, cmplogrest, cmprest set_env files to set ICE_BFBTYPE - Update documentation * Update EVP OMP implementation * - Refactor puny/pi scalars in eap dynamics to improve performance - Update OMP in evp and eap * Clean up * Comment out temporary timers * Update OMP env variables on Narwhal * Update gaffney OMP_STACKSIZE * update OMP_STACKSIZE on cori * Update Onyx OMP_STACKSIZE Update documentation * Update OMP_STACKSIZE on mustang * - Update Tsfc values on land in various places in the code, was affecting testing. Specifically fix upwind advection. - Comment out OMP in ice_dyn_evp_1d_kernel, was producing non bit-for-bit results with different thread counts
* refactor seabed_stress. Bit for bit * Removed if statement from stepu. Results are binary identical, however taubx and tauby is updated on all iterations instead of just the last one. Not used within iteration * changed capping from logical to numeric in order to remove if statement. Moved call to deformation out of loop * clean dyn_finish, correct intent(inout) to intent(in) for Cw, resse Cb in stepu, remove if from seabed_stress_LKD * Reolve conflicts after updating main * modified environment for Freya to accomodate for additional OMP commands * Requested changes after review. Only changed in seabed stress and not bit for bit if cor=0.0 added legacy comment in ice_dyn_finish * move deformation to subcycling
- Remove gordon and conrad machines. - Add setenv OMP_STACKSIZE commented out in env files - Update Icepack to fc4b809
Update tauxbx, tauxby calculations on C/CD to be consistent with main B changes
Extend omp_suite to include C/CD tests
apcraig
requested review from
eclare108213,
dabail10,
JFLemieux73 and
phil-blain
They should be reproducible over long runs! Definitely need to look into this. |
|
I've been debugging some of these issues the last couple days. It's been challenging because some issues arise due to multiple bugs. Those are always the most difficult to sort out. Have found a few issues, one an uninitialized array associated with seabed_stress. That could be the issue with reproducibility of the prod tests, but I still need to check that. |
|
This is ready to merge. It includes just the merge from main. All the other problems existed before on cgridDEV. I will create a new PR to fix some of the issues I've found. Anyone think we shouldn't merge this? |
|
I'm reviewing it now. Have a few questions but probably nothing to stop it from being merged. |
eclare108213
approved these changes
Mar 7, 2022
| @@ -328,7 +328,7 @@ subroutine implicit_solver (dt) | |||
|
|
|||
| ! tcraig, tcx, threading here leads to some non-reproducbile results and failures in icepack_ice_strength | |||
| ! need to do more debugging | |||
| !$TCXOMP PARALLEL DO PRIVATE(iblk,ilo,ihi,jlo,jhi,this_block) | |||
| !$TCXOMP PARALLEL DO PRIVATE(iblk,ilo,ihi,jlo,jhi,this_block,ij,i,j) | |||
There was a problem hiding this comment.
Is this loop still not working? If it is, then remove TCX?
There was a problem hiding this comment.
The VP still has a few known problems. One with this OMP loop, another with restarts, and possibly another with bit-for-bit with different decomps and/or OMP, CICE-Consortium#518.
| trcrn(i,j,nt_Tsfc,n) = Tf(i,j) ! surface temperature | ||
| else | ||
| trcrn(i,j,nt_Tsfc,n) = c0 ! on land gridcells | ||
| endif |
There was a problem hiding this comment.
Would it be more efficient (or maybe just more elegant) to use the merge function instead of all the if conditionals for setting Tsfc, here and elsewhere?
There was a problem hiding this comment.
I'm not a huge fan of merge syntax, but willing to start using it as it is a bit more compact. Will keep that in mind moving forward, update the current implementation, and try to use merge more.
| @@ -745,7 +745,8 @@ subroutine read_restart_field(nu,nrec,work,atype,vname,ndim3,diag, & | |||
| status = pio_inq_varid(File,trim(vname),vardesc) | |||
|
|
|||
| if (status /= PIO_noerr) then | |||
| call abort_ice(subname//"ERROR: CICE restart? Missing variable: "//trim(vname)) | |||
| call abort_ice(subname// & | |||
| "ERROR: CICE restart? Missing variable: "//trim(vname)) | |||
There was a problem hiding this comment.
why is there a question mark in this error statement?
There was a problem hiding this comment.
Not sure, I'll try to review that.
apcraig
added a commit
that referenced
this pull request
Mar 8, 2022
address the problems with the boxrestore test errors. - Update box2001 so it's bit-for-bit with different blocks/decomps/pe counts. - Initialize bathymetry values at all gridcells when bathymetry_format='default' and use_bathymetry=.false.
apcraig
added a commit
that referenced
this pull request
Mar 10, 2022
…, omp_suite (#65) * Fix several problems noted in PR #64. This should address the problems with the boxrestore test errors. - Update box2001 so it's bit-for-bit with different blocks/decomps/pe counts. - Initialize bathymetry values at all gridcells when bathymetry_format='default' and use_bathymetry=.false. * update comparelog to exclude extraneous icepack output, help prod comparisons * - Zero out certain fields on land on restart files. Some fields have non-zero values over land by default and this causes problems with land block elimination and comparisons of different decompositions. Does not affect science. - Update omp_suite to use different block sizes in comparisons * use c0 instead of 0. * fix OpenMP private variables on new code * update pio restart diagnostics format
Sign up for free
to join this conversation on GitHub.
Already have an account?
Sign in to comment
Add this suggestion to a batch that can be applied as a single commit.
This suggestion is invalid because no changes were made to the code.
Suggestions cannot be applied while the pull request is closed.
Suggestions cannot be applied while viewing a subset of changes.
Only one suggestion per line can be applied in a batch.
Add this suggestion to a batch that can be applied as a single commit.
Applying suggestions on deleted lines is not supported.
You must change the existing code in this line in order to create a valid suggestion.
Outdated suggestions cannot be applied.
This suggestion has been applied or marked resolved.
Suggestions cannot be applied from pending reviews.
Suggestions cannot be applied on multi-line comments.
Suggestions cannot be applied while the pull request is queued to merge.
Suggestion cannot be applied right now. Please check back later.
PR checklist
Update to main #8d7314f and fix C/CD deformations, taubx/tauby, and OpenMP
apcraig
full test suites run on 3 compilers on cheyenne,
https://github.com/CICE-Consortium/Test-Results/wiki/cice_by_hash_forks#71500b6021c002daba522d75f239ed26bfe37e1a
Differences vs main are documented below
Full test suites run on cheyenne intel, gnu, pgi compared to the main #8d7314f with the following results,