f501def | Michael Kelleher | 27 October 2023, 21:10:59 UTC | Add size 'S' pes for chrysalis ne4 tests | 27 October 2023, 21:10:59 UTC |
b2ea1c3 | James Foucar | 27 October 2023, 16:06:16 UTC | Merge branch 'jgfouca/csm_share_no_makefile' into master (PR #6025) CSM_share: no cime/Tools/Makefile The CSM_share build is very similar to how we build our main components, except it was being driven by the CIME/Tools/Makefile. This PR changes it to use our CMake build_model function. A few minor tweaks were needed, the main problem was with the Depends files. The CSM_share build cannot use our standard Depends system due to the likelihood that the Depends file will modify flags for files that CSM doesn't know about, which is currently an error. Instead, the CSM depends files are moved to E3SM/share, which is a bit unfortunate since it is not consistent and also makes them non-customizable within a case. On the plus side, this allows us to get rid of our last remaining Makefile-style Depends files. [BFB] | 27 October 2023, 16:06:16 UTC |
6e70c30 | James Foucar | 26 October 2023, 21:42:19 UTC | Don't change submodule | 26 October 2023, 21:42:19 UTC |
b64b5d6 | James Foucar | 26 October 2023, 19:20:18 UTC | Reorg csm_share Depends | 26 October 2023, 19:20:18 UTC |
0b367fb | Jon Wolfe | 26 October 2023, 19:19:47 UTC | Merge branch 'vanroekel/ocean/change-redi-slopes-and-limiter' (PR #5947) Improve slope calculation and fix slope limiting for redi mixing In current MPAS code, slopes are altered directly, e.g. when S > Scrit it is set to zero. Griffies et al 1998 (Appendix C) equates this to slope clipping. The correct approach is to taper the redi kappa coefficient for each individual flux instead. The cross terms ("term2" and "term 3") end up being identical as they are d/dz(kappa S grad phi) so tapering kappa or S are identical. However for the vertical it is d/dz(kappa |S|^2 d phi / dz) so tapering S is different from tapering kappa. Also term 1 (laplacian diffusion) was not being tapered with slopes creating a cross isopycnal flux in steeply sloping regions. These issues are fixed here. In addition a new slope calculation from @dengwirda is introduced to prevent the slopes from getting very large when grad phi or d/dz(phi) get small. This function limits the slope to be between -1 and 1. [NML] [CC] | 26 October 2023, 19:19:47 UTC |
510fca6 | Jon Wolfe | 26 October 2023, 19:11:43 UTC | Merge branch 'vanroekel/ocean/fix-richardson-number-calculation' (PR #5946) Fixes MPAS-Ocean Richardson number calculation This changes the calculation of the gradient and bulk richardson numbers to utilize the normal and tangential velocities to avoid the factor of two issue and also fixes a number of places where the loop was over maxLevelCell and should have been to maxLevelEdgeTop. [NML] [CC] | 26 October 2023, 19:11:43 UTC |
4d48d07 | Jon Wolfe | 26 October 2023, 19:06:26 UTC | Merge branch 'mark-petersen/ocean/redi-flux-limiter' (PR #5945) A quasi-monotone flux limiter for isopycnal diffusion (Redi) This PR addresses a range of issues concerning the isopycnal diffusion operator (Redi): * Updating the use of the triad slopes in one of the Redi fluxes. * Correcting the implementation of the Redi fluxes. * Adding a new new quasi-monotone flux limiting strategy that attempts to prevent the development of non-monotone tracer values. [non-BFB] [CC] | 26 October 2023, 19:06:26 UTC |
c98976b | James Foucar | 26 October 2023, 18:32:08 UTC | Clean up csm_share depends files a bit | 26 October 2023, 18:32:08 UTC |
e8e45c2 | James Foucar | 26 October 2023, 18:05:22 UTC | Merge branch 'master' into jgfouca/csm_share_no_makefile * master: Update CIME submodule Fix a few things I missed | 26 October 2023, 18:05:22 UTC |
9d100e4 | Jon Wolfe | 26 October 2023, 18:04:23 UTC | Merge branch 'xylar/mesh/add-ecwisc30to60e3r2' (PR #5927) Add new ECwISC30to60E3r2 ocean and sea-ice mesh Long name: ECwISC30to60L64E3SMv3r2 This mesh is a new eddy closure (EC) mesh with ice-shelf cavities (wISC) that has: * 30 km resolution at the equator * 60 km resolution at mid latitudes * 35 km resolution near the poles This mesh is a candidate for the E3SM v3 (E3) low res mesh. It is revision 2 (r2) of the mesh. (Revision 1 had incorrect bathymetry and coastline around Greenland.) All new supporting files have been staged on the inputdata repo [BFB] | 26 October 2023, 18:04:23 UTC |
6c88ed4 | James Foucar | 26 October 2023, 17:15:25 UTC | Merge branch 'jgfouca/mct_no_makefile' into master (PR #6009) MCT/Mpi-serial: Remove dependence on CIME/Tools/Makefile This is a continuation of the effort to remove all use of the CIME/Tools/Makefile from the E3SM build system so we can be free to change how we express our build settings in our cmake macros. Changes: 1) Move the python build wrapper scripts for MCT and mpi-serial from CIME to E3SM. 2) Refactor and clean up these scripts to use a similar approach as was used for the scorpio build wrapper: remove the reliance on CIME/Tools/Makefile and instead pull the needed items from the Macros directly and call the native build system (autoconf in this case) directly. 3) Clean up the macro extraction a bit 4) Fix pnetcdf setting for spio Testing: * Mappy_gnu A case (Mpilib=serial, openmpi) * anlgce_gnu A case (Mpilib=serial, mpich) * pm-cpu_gnu A case ((Mpilib=mpich, mpi-serial does not work on pm-cpu on master) * summit_ibm A case (Mpilib=spectrum-mpi, mpi-serial does not work on summit_ibm on master) * summit_gnu A case (Mpilib=spectrum-mpi, mpi-serial) [BFB] | 26 October 2023, 17:15:25 UTC |
e573ab5 | James Foucar | 26 October 2023, 15:55:42 UTC | Merge branch 'jgfouca/cime_update_2023_10_25' into master (PR #6020) Update CIME submodule ... to 7b579110a1c8c8e1843bb61dbd7ad9dad531087a Changes: 1) case.submit: reduce the number of times env_batch is rewriten Fixes: 1) bless_test_results: A number of fixes related to recent perf updates 2) Fix docker [BFB] | 26 October 2023, 15:55:42 UTC |
cfb1208 | Jon Wolfe | 25 October 2023, 19:04:44 UTC | Add missing variable to omp directive | 25 October 2023, 19:04:44 UTC |
ec67ee7 | Jon Wolfe | 25 October 2023, 18:57:10 UTC | Add missing variable to omp directive | 25 October 2023, 18:57:10 UTC |
4015538 | James Foucar | 25 October 2023, 15:40:48 UTC | Update CIME submodule ... to 7b579110a1c8c8e1843bb61dbd7ad9dad531087a Changes: 1) case.submit: reduce the number of times env_batch is rewriten Fixes: 1) bless_test_results: A number of fixes related to recent perf updates 2) Fix docker [BFB] | 25 October 2023, 15:40:48 UTC |
4e47fce | James Foucar | 24 October 2023, 20:14:34 UTC | Fix a few things I missed | 24 October 2023, 20:14:34 UTC |
9b027e2 | James Foucar | 20 October 2023, 21:56:20 UTC | Get rid of last Makefile Depends | 20 October 2023, 21:56:20 UTC |
7f1b1f9 | James Foucar | 20 October 2023, 21:26:29 UTC | Fixes. LIBROOT is not really used anymore | 20 October 2023, 21:26:55 UTC |
dbfcc2b | James Foucar | 20 October 2023, 21:00:11 UTC | csm_share: Build via cmake | 20 October 2023, 21:00:11 UTC |
e595a69 | James Foucar | 20 October 2023, 18:05:23 UTC | Fix pnetcdf config for spio | 20 October 2023, 18:05:23 UTC |
7c27446 | James Foucar | 20 October 2023, 17:11:46 UTC | Fix mistake in mct build script | 20 October 2023, 17:11:46 UTC |
5bec758 | James Foucar | 20 October 2023, 16:54:58 UTC | Cleanup macro extraction | 20 October 2023, 16:54:58 UTC |
3f5df2b | James Foucar | 20 October 2023, 16:06:50 UTC | Merge branch 'master' into jgfouca/mct_no_makefile * master: (22 commits) Fix macros extraction and hdf5 bool cmake setting Revert "Turn off hdf5 on summit for scorpio" Add WITH_HDF5_SCORPIO support Turn off hdf5 on summit for scorpio Be less aggressive about turning hdf5 on for scorpio Add the hdf5 dep to PIO Fix zlib/libz szip/zlib Fix for adios/gptl More help for scorpio to find HDF5 Fix NDEBUG cppdef Fixed SSP585 surfdata filename Update lnd surfdata to use TOP for ne30pg2 and r05 with workaround for wcprod tests Add jenkins save timing to cime update Update CIME submodule Updates ELM surface datasets for few resolutions Revert "Revert "Merge branch 'jonbob/seaice/fix-ir-zero-check' (PR #5252)"" Update p3_nc_autocon_expon tuning for v3alpha04 config Turn on TOP solar radation scheme in elm for all EAM%CMIP6 compsets TOP solar rad parameterization set from ELM compset name ... | 20 October 2023, 16:06:50 UTC |
15fa673 | James Foucar | 20 October 2023, 16:00:03 UTC | Merge branch 'jgfouca/sharedlib_no_makefile' into next (PR #5997) PIO: Remove dependence on CIME/Tools/Makefile We need to remove all dependencies on CIME/Tools/Makefile so we can change our macros to use cmake-style names for things. PIO/Scorpio was the first thing I picked. The situation with spio was strange. It was using CIME/Tools/Makefile to call CMake. It translated the Makefile-style settings to Cmake using a lot of extra code in the Makefile to set up CMAKE_OPTS. We change the python wrapper build script, buildlib.spio, to instead just set up the cmake options in python, allowing us to skip the CIME/Tools/Makefile. On top of removing the dependency, I think this makes the overall spio build a lot easier to understand, although it was a bit tedious pulling things out of the macros. Once the macros are in proper cmake-stype, I think we could just pass them directly to scorpio's CMake build via -C and things will be even simpler. [BFB] | 20 October 2023, 16:00:03 UTC |
464321f | James Foucar | 20 October 2023, 15:12:39 UTC | Fix macros extraction and hdf5 bool cmake setting | 20 October 2023, 15:12:39 UTC |
213d5e5 | James Foucar | 19 October 2023, 21:21:05 UTC | Revert "Turn off hdf5 on summit for scorpio" This reverts commit 9d1bd9d40ade12531697b9c687b75c6855d50db7. | 19 October 2023, 21:21:05 UTC |
dc2b58b | James Foucar | 19 October 2023, 21:11:59 UTC | Add WITH_HDF5_SCORPIO support | 19 October 2023, 21:11:59 UTC |
9d1bd9d | James Foucar | 19 October 2023, 19:59:35 UTC | Turn off hdf5 on summit for scorpio | 19 October 2023, 19:59:35 UTC |
02d87e0 | James Foucar | 19 October 2023, 19:47:38 UTC | Be less aggressive about turning hdf5 on for scorpio | 19 October 2023, 19:47:38 UTC |
2ac88ca | James Foucar | 19 October 2023, 19:36:42 UTC | Add the hdf5 dep to PIO | 19 October 2023, 19:36:42 UTC |
ff15482 | James Foucar | 19 October 2023, 18:53:52 UTC | Fix zlib/libz | 19 October 2023, 18:53:52 UTC |
d729e2e | James Foucar | 19 October 2023, 18:19:38 UTC | szip/zlib | 19 October 2023, 18:19:38 UTC |
9b15d1f | James Foucar | 19 October 2023, 17:57:01 UTC | Fix for adios/gptl | 19 October 2023, 17:57:01 UTC |
2871f56 | James Foucar | 19 October 2023, 17:40:05 UTC | More help for scorpio to find HDF5 | 19 October 2023, 17:40:05 UTC |
ff07d40 | James Foucar | 19 October 2023, 16:18:32 UTC | Merge branch 'jgfouca/cime_update_2023_10_16' into master (PR #6000) Update CIME submodule ... to 88255d5a5a1b160ec523a95fd26d6be079c54935 Changes: 1) Copy only necessary cmake macros to case 2) Update ESMCI docker, no impact to E3SM 3) update instructions for building cprnc with cmake 4) Updates how performance baselines are created and compared! 5) Add --save-timing to Jenkins_generic_job args. Fixes: 1) Fixes workflow job dependencies 2) Fixes ww3 archiving [BFB] | 19 October 2023, 16:18:32 UTC |
596103a | Wuyin Lin | 19 October 2023, 15:52:26 UTC | Merge branch 'wlin/restore_PR5252' (PR #6004) Restore the fix in PR 5252 The fix was reverted via #5970 for creating v3alpha03 tag. [NBFB] relative to the state before this revert of revert. | 19 October 2023, 15:52:26 UTC |
b592407 | Wuyin Lin | 19 October 2023, 15:50:32 UTC | Merge branch 'wlin/wrapup_v3alpha04' (PR #5973) Wrap up v3alpha04 configuration This is in preparation for a code base for v3alpha04 tag. It is on top of the v3alpha03, with a retuned p3_nc_autocon_expon. [non-BFB] for all tests involving EAM [NML] p3_nc_autocon_expon setting changed | 19 October 2023, 15:50:32 UTC |
bfaadd3 | Wuyin Lin | 19 October 2023, 15:48:13 UTC | Merge branch 'wlin/wrapup_v3alpha03' (PR #5970) Set parameters to wrap up v3alpha03 configuration To set default the extra parameters for atm and lnd used by v3alpha03 configuration. The merge of this PR should create a state ready for tagging v3alpha03. Land surface data for WCYCL config are updated to enable subgrid topographic radiative parameterization (TOP). The surfdata files for simyr 1850 and 2010 are fully ready for ne30pg2 and r05 grids. Workaround are applied for F2010 ne4, ne4pg2, ne11 grids, 1850 NARRM, and SSP370 and SSP585 ne30pg2 grid, with placeholder for TOP-related variables. The TOP developers will update these files. Land initial files for several wcprod F2010 and F20R tests with older trigrid are disabled. The trigrid for the tests will be updated at a later time along with new initial files. [non-BFB] for all F and B-cases. [NML] extra default namelist variables in atm_in and lnd_in along with some update on input files | 19 October 2023, 15:48:13 UTC |
d76f093 | James Foucar | 18 October 2023, 22:00:46 UTC | MCT/Mpi-serial: Sever dependence on CIME/Tools/Makefile [BFB] | 18 October 2023, 22:00:46 UTC |
67abc10 | Jon Wolfe | 18 October 2023, 21:19:14 UTC | Update new mapping files to use updated algotihms and naming convention | 18 October 2023, 21:19:14 UTC |
572be65 | James Foucar | 18 October 2023, 21:09:38 UTC | Merge branch 'jgfouca/fix_dndebug' into master (PR #6005) Fix NDEBUG cppdef The logic for adding it accidentally got inverted by an earlier PR. [BFB] | 18 October 2023, 21:09:38 UTC |
5811e29 | James Foucar | 18 October 2023, 21:03:33 UTC | Fix NDEBUG cppdef The logic for adding it accidently got inverted by an earlier PR. [BFB] | 18 October 2023, 21:03:33 UTC |
a699733 | James Foucar | 18 October 2023, 19:20:42 UTC | Prefer ROOT over PATH for env for cmake | 18 October 2023, 19:20:42 UTC |
5cc953b | James Foucar | 18 October 2023, 19:16:21 UTC | Merge remote-tracking branch 'origin/master' into jgfouca/sharedlib_no_makefile * origin/master: (24 commits) Update Depends file Update and remove ADIOS2_ROOT for some E3SM machines pm-cpu/gnu use Generic blas/lapack Fix blas/lapack on ascent/summit Fix MKL linking on pm-cpu_intel Add netcdf libs to LD_LIBRARY_PATH after recent CMake updates Update Sunspot's OneAPI compiler to 2023.05.15.007 Update Anvil e3sm_prod PEs for in test-PEs Upgrading SCORPIO to 1.4.2 Update Anvil e3sm_prod PEs to avoid OOM errors. Update Anvil cmake 3.20.3->3.26.3 Fix csm_share link errors Update intel cmake macros for chicoma Update modules for chicoma-cpu Fix trailing whitespace Clean up remaining cmake deps for pm-cpu/alvarez, remove work-around to use CPL_PSTRID=8 after pm-cpu maintenance Adding CICE-QC testing script and documentation Use SSP specific pattern match and fix use of eqv operator with or/and Update ncol/pcols in EAM array refs to fix SIGFPEs in GNU runs ... | 18 October 2023, 19:16:21 UTC |
17e6feb | James Foucar | 18 October 2023, 19:11:49 UTC | Merge branch 'dqwu/machinefiles/adios-upgrade-cleanup' into master (PR #5988) Cleanup ADIOS2_ROOT for some E3SM machines So far, we will only test the new ADIOS IO type on the following machines: Perlmutter, Frontier, Summit, and ANL GCE workstations. Additionally, the ADIOS2 library has been upgraded from version 2.8.3 to 2.9.1 on these machines. Existing ADIOS2_ROOT configurations have been removed on specific machines, including Chrysalis, Anvil, Bebop, and Compy. [BFB] | 18 October 2023, 19:11:49 UTC |
6817fd9 | James Foucar | 18 October 2023, 19:10:52 UTC | Merge branch 'jgfouca/more_cmake_cleanup' into master (PR #5990) Changes: * Remove hdf5 stuff from the macros. FindNETCDF should take care of this stuff * Remove ESMF stuff that we don't use * Clean up some MPI stuff that was unused or had no effect on the build * gptl is now being handled properly in the new style With this PR, I think all of our main dependencies are in a good enough state as far as cmake is concerned. [BFB] * jgfouca/more_cmake_cleanup: Clean up remaining cmake deps | 18 October 2023, 19:10:52 UTC |
b25e37f | Jon Wolfe | 18 October 2023, 19:04:38 UTC | Merge branch 'xylar/machinefiles/update-chicoma-cpu' (PR #5986) Update modules for chicoma-cpu Several modules are no longer available after a recent machine upgrade. [BFB] | 18 October 2023, 19:04:38 UTC |
f703513 | Jon Wolfe | 18 October 2023, 18:59:30 UTC | Merge branch 'darincomeau/ice/add-cice-qc-testing' (PR #5969) Add CICE-QC testing script Adds a script to perform "CICE-QC" test, which determines if non-BFB changes in the sea ice model are climate-changing. This has been adapted from the CICE implementation for MPAS-Seaice. Documentation on usage is provided in a README.md alongside test script. [BFB] | 18 October 2023, 18:59:30 UTC |
1cab48c | Wuyin Lin | 18 October 2023, 16:41:35 UTC | Fixed SSP585 surfdata filename | 18 October 2023, 16:41:35 UTC |
9ed66d9 | Wuyin Lin | 17 October 2023, 22:01:17 UTC | Update lnd surfdata to use TOP for ne30pg2 and r05 with workaround for wcprod tests | 17 October 2023, 22:01:17 UTC |
9a5d441 | James Foucar | 17 October 2023, 17:00:00 UTC | Add jenkins save timing to cime update | 17 October 2023, 17:00:00 UTC |
f963238 | noel | 17 October 2023, 16:11:09 UTC | Merge branch E3SM-Project/ndk/machinefiles/pm-cpu-remove-pelayout-workaround (PR #5971) For pm-cpu/alvarez, adjust pelayouts to remove performance work-around with CPL_PSTRID after maintenance | 17 October 2023, 16:11:09 UTC |
94822ce | James Foucar | 16 October 2023, 19:45:54 UTC | Try setting hdf5 path explicitly | 16 October 2023, 19:45:54 UTC |
6b75a3a | James Foucar | 16 October 2023, 19:28:48 UTC | Fix cmake_opts mistake | 16 October 2023, 19:28:48 UTC |
2236937 | Azamat Mametjanov | 16 October 2023, 19:04:55 UTC | Merge branch 'azamat/sunspot/update-oneapi-20230515007' (PR #5881) Update Sunspot's OneAPI compiler to 2023.05.15.007 . Also, - add netcdf libs to LD_LIBRARY_PATH after recent CMake updates - rearrange modules by loading defaults after module purge - clean up cmake Fixes E3SM-Project/E3SM#5707 [BFB] | 16 October 2023, 19:04:55 UTC |
b22c83b | James Foucar | 16 October 2023, 17:06:28 UTC | Update CIME submodule ... to 88255d5a5a1b160ec523a95fd26d6be079c54935 Changes: 1) Copy only necessary cmake macros to case 2) Update ESMCI docker, no impact to E3SM 3) update instructions for building cprnc with cmake 4) Updates how performance baselines are created and compared! Fixes: 1) Fixes workflow job dependencies 2) Fixes ww3 archiving [BFB] | 16 October 2023, 17:06:28 UTC |
d9c80c3 | James Foucar | 16 October 2023, 16:56:52 UTC | Fix circleCI to refer to correct PIO build name | 16 October 2023, 16:56:52 UTC |
db2cac0 | Azamat Mametjanov | 14 October 2023, 01:38:46 UTC | Update Depends file Update Depends file - don't need `-fc=ifort` any more, newer ifx can build 50K+ line sources - move OpenMP-offload to GPU-enabled CIME compiler | 14 October 2023, 01:38:46 UTC |
d61fb8e | dqwu | 11 October 2023, 22:27:20 UTC | Update and remove ADIOS2_ROOT for some E3SM machines To facilitate testing of the ADIOS IO type, ADIOS2_ROOT is now configured on the following machines: Perlmutter, Frontier, Summit, and ANL GCE workstations. Additionally, the ADIOS2 library has been upgraded from version 2.8.3 to 2.9.1 on these machines. Existing ADIOS2_ROOT configurations have been removed on specific machines, including Chrysalis, Anvil, Bebop, and Compy. | 14 October 2023, 00:07:00 UTC |
0fd8844 | jayeshkrishna | 13 October 2023, 23:15:45 UTC | Merge branch 'jayeshkrishna/externals/scorpio_1_4_2' (PR #5991) Upgrading SCORPIO from 1.4.1 to 1.4.2 SCORPIO v1.4.2 release includes the following fixes over v1.4.1, * Performance tuning (file striping settings) for Frontier * ADIOS performance improvements and fixes [BFB] | 13 October 2023, 23:15:45 UTC |
f6f7b13 | Azamat Mametjanov | 13 October 2023, 23:07:04 UTC | Merge branch 'azamat/machines/add-gnu-invalid-check' (PR #5808) Add check for invalid fp-ops for GNU in debug mode. Also, fix ncol/pcol indexes in 2 eam-chemistry subroutines (flagged by `-ffpe-trap=invalid`). [non-BFB] - in fully-coupled tests | 13 October 2023, 23:07:04 UTC |
cff4226 | James Foucar | 13 October 2023, 22:36:47 UTC | Merge branch 'jgfouca/pm_cpu_gnu_lapack_blas' into master (PR #5998) pm-cpu/gnu use Generic blas/lapack Not the 'implicitly linked' ones. This fixes DIFFs. | 13 October 2023, 22:36:47 UTC |
396bcb4 | James Foucar | 13 October 2023, 22:34:47 UTC | pm-cpu/gnu use Generic blas/lapack Not the 'implicitly linked' ones. This fixes DIFFs. | 13 October 2023, 22:34:47 UTC |
95a13d9 | James Foucar | 13 October 2023, 22:07:11 UTC | Merge branch 'jgfouca/ascent_summit_blas' into master (PR #5995) Fix blas/lapack on ascent/summit Tell cmake to use Generic so it doesn't think ESSL is sufficient. [BFB] | 13 October 2023, 22:07:11 UTC |
67e2180 | James Foucar | 13 October 2023, 22:04:07 UTC | PIO: Remove dependence on CIME/Tools/Makefile We need to remove all dependencies on CIME/Tools/Makefile so we can change our macros to use cmake-style names for things. PIO/Scorpio was the first thing I picked. The situation with spio was strange. It was using CIME/Tools/Makefile to call CMake. It translated the Makefile-style settings to Cmake using a lot of extra code in the Makefile to set up CMAKE_OPTS. We change the python wrapper build script, buildlib.spio, to instead just set up the cmake options in python, allowing us to skip the CIME/Tools/Makefile. On top of removing the dependency, I think this makes the overall spio build a lot easier to understand, although it was a bit tedious pulling things out of the macros. Once the macros are in proper cmake-stype, I think we could just pass them directly to scorpio's CMake build via -C and things will be even simpler. [BFB] | 13 October 2023, 22:04:07 UTC |
9e6d115 | James Foucar | 13 October 2023, 21:16:17 UTC | Fix blas/lapack on ascent/summit Tell cmake to use Generic so it doesn't think ESSL is sufficient. [BFB] | 13 October 2023, 21:16:17 UTC |
97e527f | Azamat Mametjanov | 13 October 2023, 20:50:36 UTC | Merge branch 'azamat/anvil/update-e3sm-prod-PEs' (PR #5987) Update Anvil config to fix OOM and link errors - Fix csm_share link errors - Update Anvil cmake 3.20.3->3.26.3 - Update Anvil e3sm_prod PEs to avoid OOM errors [BFB] | 13 October 2023, 20:50:36 UTC |
235a851 | Wuyin Lin | 13 October 2023, 20:44:54 UTC | Merge branch 'yunpengshan2014/E3SM/P3SedimenttaionFluxChange' (PR #5891) Fix the calculations of p3 sedimentation fluxes Original calculation in the P3 code: the precipitation (rain and snow) flux (sflx for snow ) is summed over the flux at each sub-timestep during one microphysics timestep: sflx=flux_qit_1+ flux_qit_2+ flux_qit_3+… flux_qit_n, where the flux_qit_i (i=1,2,3,…n) represents the snow flux at each sub-timestep for sedimentation. This should not be correct. The correct flux should be the mean value for each microphysics timestep. Changed calculation in the current version: In E3SM-P3, since the sub-timestep is not a fixed value, the mean flux over the microphysics timestep should be calculated with: sflx=(flux_qit_1dt1+ flux_qit_2dt2+ flux_qit_3dt3+… flux_qit_ndtn)/( dt1+ dt2+ dt3+… dtn), where dti represents sub-timestep length. This is similar to the equation used to calculate the surface precipitation rate. This PR does not effect the model simulation results, but only change the diagnostic precipitation flux output. [BFB] for simulation results, COSP output affected. | 13 October 2023, 20:44:54 UTC |
d98c4fc | Wuyin Lin | 13 October 2023, 20:42:58 UTC | Merge branch 'wlin/atm/ssp_tentativefix' (PR #5965) Implemented a tentative fix for SSP tests Future scenario forcing files for new species introduced by v3atm are not ready. The cime SSP tests using default v3 configuration therefore failed to be created. The tentative fix uses config that mixes v3 physics with v2 chemistry. Also fixed a couple misuses of eqv operator in linoz-related code that do not properly consider the precedence relative to AND/OR operators. The misuses have no impact on default v3 configuration, but would cause fatal runtime error when using the hybrid configuration as implemented here. [BFB] for existing tests. | 13 October 2023, 20:42:58 UTC |
495ba1d | James Foucar | 13 October 2023, 18:22:39 UTC | Merge branch 'jgfouca/fix_pm_intel_blas' into master (PR #5993) Fix MKL linking on pm-cpu_intel Need to tell cmake to use Intel vendor for blas. [BFB] | 13 October 2023, 18:22:39 UTC |
a063972 | James Foucar | 13 October 2023, 16:17:38 UTC | Fix MKL linking on pm-cpu_intel Need to tell cmake to use Intel vendor for blas. [BFB] | 13 October 2023, 16:19:59 UTC |
3f78dfa | Azamat Mametjanov | 13 October 2023, 00:31:18 UTC | Add netcdf libs to LD_LIBRARY_PATH after recent CMake updates Also rearrange modules by loading defaults after module purge and clean up cmake. | 13 October 2023, 00:31:18 UTC |
2a93b58 | Azamat Mametjanov | 15 August 2023, 00:12:52 UTC | Update Sunspot's OneAPI compiler to 2023.05.15.007 | 12 October 2023, 22:28:04 UTC |
0dfc07f | Azamat Mametjanov | 12 October 2023, 21:40:59 UTC | Update Anvil e3sm_prod PEs for in test-PEs Also remove `-mkl` from link line after CMake BLAS updates | 12 October 2023, 21:40:59 UTC |
ed32fda | jayeshkrishna | 12 October 2023, 18:33:47 UTC | Upgrading SCORPIO to 1.4.2 Upgrading SCORPIO from 1.4.1 to 1.4.2 SCORPIO v1.4.2 release includes the following fixes over v1.4.1, * Performance tuning (file striping settings) for Frontier * ADIOS performance improvements and fixes | 12 October 2023, 18:33:47 UTC |
6223567 | James Foucar | 12 October 2023, 17:34:49 UTC | Merge remote-tracking branch 'origin/master' into jgfouca/more_cmake_cleanup * origin/master: (325 commits) Set mappy to use generic blas blas/lapack settings should not depend on mpilib Refactor blas and lapack stuff a bit more revert fix for gravity wave convective heating Add back in kokkos options Revert to master version of gnugpu_pm-gpu.cmake add Kokkos_ENABLE_OPENMP=Off for gnugpu on pm-gpu add Kokkos_ENABLE_OPENMP=Off for gnugpu on Frontier add Kokkos_ENABLE_OPENMP=Off for Summit remove configure support for MMF m2005 microphysics configure clean up and remove samomp support add new lines re-enable CRM variance and momentum transport for MMF1 add new line at end of pam_debug update default PAM time step update PAM submodule disable kokkos finalize in PAM driver add hyperdiffusion to PAM and refactor PAM driver update pam_statistics refactor pam_output ... | 12 October 2023, 17:34:49 UTC |
14d8fa1 | James Foucar | 12 October 2023, 17:13:48 UTC | Merge branch 'jgfouca/cmake_blas_lapack' into master (PR #5972) Convert blas/lapack/mkl to new cmake style This was a more challenging task than I thought it would. FindBLAS / FindLAPACK are built in to cmake and can also handle MKL stuff via the BLA_VENDOR setting (if you set it to Intel... you get various flavors of MKL). The find_package for blas/lapack is pretty smart and should do the right thing most of the time. Machine POCs should refine the settings on their machines using https://cmake.org/cmake/help/latest/module/FindBLAS.html#blas-lapack-vendors as as guide. I suspect many of these cache files were copy/pasted without a ton of consideration. As an example, pm-cpu_nvidia is one platform where you will get DIFFs if you don't use the right blas/lapack. To set this up, I added the following to the environment block for that machine/compiler combo: <environment_variables compiler="nvidia"> <env name="BLAS_ROOT">$SHELL{if [ -z "$BLAS_ROOT" ]; then echo /opt/nvidia/hpc_sdk/Linux_x86_64/22.7/compilers; else echo "$BLAS_ROOT"; fi}</env> <env name="LAPACK_ROOT">$SHELL{if [ -z "$LAPACK_ROOT" ]; then echo /opt/nvidia/hpc_sdk/Linux_x86_64/22.7/compilers; else echo "$LAPACK_ROOT"; fi}</env> <env name="BLA_VENDOR">NVHPC</env> </environment_variables> The documentation indicates that the find modules for blas may be smart enough to auto detect the bla_vendor: BLA_VENDOR Set to one of the [BLAS/LAPACK Vendors] to search for BLAS only from the specified vendor. If not set, all vendors are considered. Once again, I am impressed at how much clutter is removed when we do things t "right" (using find_package instead of packing stuff into SLIBS) way. [BFB] | 12 October 2023, 17:13:48 UTC |
bcd8798 | James Foucar | 12 October 2023, 17:04:16 UTC | Set mappy to use generic blas | 12 October 2023, 17:04:16 UTC |
c64b721 | Gautam Bisht | 12 October 2023, 04:18:54 UTC | Updates ELM surface datasets for few resolutions Surface datasets dataset for following resolutions have been updated to include new fields needed for TOP parameterization in ELM: - ne4_oQU240 - ne4pg2_oQU480 - northamericax4v1pg2_WC14to60E2r3 | 12 October 2023, 04:18:54 UTC |
d9e1400 | Wuyin Lin | 11 October 2023, 22:29:34 UTC | Revert "Revert "Merge branch 'jonbob/seaice/fix-ir-zero-check' (PR #5252)"" This reverts commit 40a7f522369aaf4c8d3d643d0f90d9219b71a633. | 11 October 2023, 22:29:34 UTC |
26170af | James Foucar | 11 October 2023, 21:35:43 UTC | blas/lapack settings should not depend on mpilib | 11 October 2023, 21:35:43 UTC |
23e9e30 | James Foucar | 11 October 2023, 21:30:20 UTC | Refactor blas and lapack stuff a bit more Get rid of all the BLA_VENDOR stuff in the macros. It had no impact because deps are loading before macros are processed. We should prefer using the environment to set this stuff like we are doing for the other libs. pm-cpu_nvidia seems to be the one platform sensitive to lapack/blas so can be used as an example of how to specify these libraries. | 11 October 2023, 21:30:20 UTC |
613312c | Azamat Mametjanov | 11 October 2023, 20:52:14 UTC | Update Anvil e3sm_prod PEs to avoid OOM errors. | 11 October 2023, 20:52:14 UTC |
048078f | Azamat Mametjanov | 11 October 2023, 20:46:46 UTC | Update Anvil cmake 3.20.3->3.26.3 Also disable test archiving on Bebop and Anvil. | 11 October 2023, 20:46:46 UTC |
9525073 | Azamat Mametjanov | 11 October 2023, 20:36:07 UTC | Fix csm_share link errors Link to MKL single dynamic library with `-mkl` to avoid undefined references to Intel's vector math routines vdsqrt, vdexp and others in csm_share. | 11 October 2023, 20:36:07 UTC |
99a3acc | Xylar Asay-Davis | 31 August 2023, 10:47:13 UTC | Update intel cmake macros for chicoma | 11 October 2023, 18:09:29 UTC |
c18e11b | Xylar Asay-Davis | 30 August 2023, 10:10:56 UTC | Update modules for chicoma-cpu | 11 October 2023, 18:01:35 UTC |
8a14ed5 | Xylar Asay-Davis | 30 August 2023, 10:03:26 UTC | Fix trailing whitespace | 11 October 2023, 18:01:35 UTC |
a093fa2 | James Foucar | 10 October 2023, 20:29:20 UTC | Clean up remaining cmake deps | 10 October 2023, 20:29:20 UTC |
732e699 | Benjamin Hillman | 10 October 2023, 20:00:05 UTC | Merge branch 'whannah/mmf/pam-impl' (#5906) ntegration of new CRM for MMF, referred to as the "portable atmosphere model" or PAM, that uses the P3 and SHOC physics implementation from SCREAM/EAMxx. The new CRM is used by the compsets with "MMF2" in their name. While the PAM code is isolated to a submodule, several large changes outside were required to facilitate this integration. The PAM interface code is a substantial part of this, but there was also new conflicts created by EAMv3 integration efforts which led to the creation of various "dummy" modules to avoid building the conflicting modules, which are unnecessary for MMF. A new integration is added to test the new CRM => SMS_Ln3.ne4pg2_ne4pg2.F2010-MMF2 A few things to note about the current MMF2 implementation - When running with PAM on the GPU the use of openMP threads is problematic for reasons that are unclear. This seems fixable, but more work is needed to look into this. For now, running on GPUs without OMP threads is sufficient, and CPU cases are not affected by this problem - Although some offline PAM tests have used longer time steps, when running within the MMF configuration at ne30 the PAM time step needs to be 4 seconds for stability. A lot of performance can be gained by increasing this time step, but more work is needed to understand the source of the instability. - PAM performance also seems highly machine dependent. For example, using 22 nodes on Perlmutter runs 6x faster than 64 nodes on Summit for the same case. [BFB] | 10 October 2023, 20:00:05 UTC |
9b40b2d | James Foucar | 10 October 2023, 16:48:35 UTC | Merge remote-tracking branch 'origin/master' into jgfouca/cmake_blas_lapack * origin/master: Fix nvidia_pm-cpu diffs for some tests Fix PIO dep on mpi-serial add ghp preview | 10 October 2023, 16:48:35 UTC |
e195c80 | James Foucar | 10 October 2023, 16:43:27 UTC | Merge branch 'jgfouca/fix_nvidia_pm_cpu_diffs' into master (PR #5976) Fix nvidia_pm-cpu diffs for some tests Be explicit about which hdf5 and blas/lapack libraries to link so that later -L options don't mess things up. Also, make some minor enhancements to the compare-flags tool in order to get more useful output when comparing unordered flags. [BFB] | 10 October 2023, 16:43:27 UTC |
85addcc | James Foucar | 10 October 2023, 16:40:57 UTC | Merge branch 'jgfouca/fix_mpi_serial_link' into master (PR #5975) Fix PIO dep on mpi-serial PIO needs to depend on mpi-serial in order for the link order to be correct. [BFB] | 10 October 2023, 16:40:57 UTC |
ddf0c37 | Walter Hannah | 09 October 2023, 22:12:30 UTC | revert fix for gravity wave convective heating | 09 October 2023, 22:12:30 UTC |
c4e81db | James Foucar | 09 October 2023, 21:20:18 UTC | Fix nvidia_pm-cpu diffs for some tests Be explicit about which hdf5 and blas/lapack libraries to link so that later -L options don't mess things up. Also, make some minor enhancements to the compare-flags tool in order to get more useful output when comparing unordered flags. [BFB] | 09 October 2023, 21:20:18 UTC |
05cbb2e | James Foucar | 09 October 2023, 21:13:54 UTC | Fix PIO dep on mpi-serial PIO needs to depend on mpi-serial in order for the link order to be correct. [BFB] | 09 October 2023, 21:13:54 UTC |
2f2639c | Benjamin Hillman | 09 October 2023, 18:50:54 UTC | Add back in kokkos options | 09 October 2023, 18:50:54 UTC |
b39578d | Benjamin Hillman | 09 October 2023, 18:49:53 UTC | Revert to master version of gnugpu_pm-gpu.cmake | 09 October 2023, 18:49:53 UTC |
4d82d99 | James Foucar | 09 October 2023, 18:31:40 UTC | Update minimum cmake | 09 October 2023, 18:31:40 UTC |
589f08c | Robert Jacob | 09 October 2023, 18:07:34 UTC | Merge branch 'mahf708/docs/improve-gh-pages-docs' (PR #5951) Adds docs preview to PR page. Preview appears in a comment. This workflow will update the same comment if there are changes (it won't pollute the PR page with more comments). Additionally, I spaced out the GitHub action so that it is easier to read and maintain (and shortened the names) [BFB] | 09 October 2023, 18:07:34 UTC |