bb7d5f1 | Gautam Bisht | 21 February 2019, 19:02:26 UTC | Merge branch 'brhillman/clm/fix-mksurfdatamap-build' (PR #2757) Fix build environment for the `mksurfdata_map` tool, which is used to build new land surface datasets. This only requires a simple change to the Makefile to not assume a path for `nf-config` relative to `LIB_NETCDF`. This was causing a problem because different machines would have a different location for `nf-config` relative to `LIB_NETCDF`. So calling `nf-config` without specifying a path (and assuming that `nf-config` exists in the `PATH` environment variable) seems to make the build more portable and not require hand editing the Makefile. [BFB] | 21 February 2019, 19:02:26 UTC |
34e9cee | Benjamin Hillman | 20 February 2019, 23:17:29 UTC | Do not assume relative path for nf-config Do not assume a path for `nf-config` relative to `LIB_NETCDF`. This was causing a problem because different machines would have a different location for `nf-config` relative to `LIB_NETCDF`. So calling `nf-config` without specifying a path (and assuming that `nf-config` exists in the `PATH` environment variable) seems to make the build more portable and not require hand editing the Makefile. | 20 February 2019, 23:17:29 UTC |
721ce2d | Jon Wolfe | 20 February 2019, 15:24:00 UTC | Merge branch 'jonbob/scripts/add-ocnice-extra-tests' (PR #2749) Add oceanice extra coverage tests including PEM and ERS low-res tests This PR adds e3sm_ocnice_extra_coverage definition to the tests.py file and includes them in the e3sm_extra_coverage suite. It also adds the first two tests to this sub-suite: * ERS_P480_Ld5.T62_oEC60to30v3wLI.GMPAS-DIB-IAF-ISMF * PEM_P480_Ld5.T62_oEC60to30v3wLI.GMPAS-DIB-IAF-ISMF to provide testing of the new data iceberg (DIB) and ice-sheet melt flux (ISMF) capabilities in the ocnice components. [BFB] | 20 February 2019, 15:24:00 UTC |
bb0fd8c | James Foucar | 19 February 2019, 23:05:26 UTC | Merge branch 'jgfouca/btr_auto_testid' into master (PR #2755) bless_test_results: when multiple sets of tests match test-id, use newest [BFB] * jgfouca/btr_auto_testid: bless_test_results: when multiple sets of tests match test-id, use newest | 19 February 2019, 23:06:02 UTC |
1ff0f3e | James Foucar | 19 February 2019, 22:51:51 UTC | bless_test_results: when multiple sets of tests match test-id, use newest | 19 February 2019, 22:51:51 UTC |
4531c4f | Mark Taylor | 15 February 2019, 23:36:03 UTC | Merge remote-tracking branch 'origin/imkg252a' (PR #2746) Added the imkg252a method (5 stages, 2 implicit solves). [BFB] | 15 February 2019, 23:36:03 UTC |
ee3338b | Jon Wolfe | 14 February 2019, 16:20:03 UTC | Add oceanice extra coverage tests including PEM and ERS low-res tests | 14 February 2019, 16:20:03 UTC |
4f22ec7 | Jon Wolfe | 14 February 2019, 15:46:05 UTC | Merge branch 'mark-petersen/ocean/fix_ERS_PEM_with_ISMF' (PR #2747) Fixes for bfb restart and partitions tests in E3SM This PR brings in a new mpas-source submodule with three changes: 1. Widen halo computation ring for kinetic energy. 2. Widen halo computation ring for melt fluxes. 3. Add halo exchange for surfaceFrictionVelocity When running E3SM with melt fluxes on (compsets with ISMF), the first is needed for bfb restarts and all three for bfb partition tests (#2739). Fixes #2702 Fixes #2739 [BFB] for all non-ISMF configurations | 14 February 2019, 15:46:05 UTC |
777fbcb | Min Xu | 14 February 2019, 04:17:41 UTC | Merge branch 'minxu74/machines/summit/add_netcdf_path_env' (PR #2723) Add the NETCDF_PATH env for Summit. The MPAS components use the NETCDF_PATH env to find the NetCDF package and the env has not been defined in the machine and compiler xml files. [BFB] - Bit-For-Bit | 14 February 2019, 04:17:41 UTC |
6ecdeda | Min Xu | 14 February 2019, 04:02:45 UTC | Merge branch 'worleyph/cime/project_updates_2019' (PR #2738) Add performance archiving support for cli133 at OLCF | 14 February 2019, 04:02:45 UTC |
65ab567 | mark-petersen | 13 February 2019, 22:12:04 UTC | Update mpas-source | 13 February 2019, 22:12:04 UTC |
f292cba | James Foucar | 13 February 2019, 21:04:18 UTC | Merge branch 'jgfouca/cime/create_test_single_exe' into master (PR #2694) create_test: Add new single-exe option Allows for all cases to share a single build. [BFB] * jgfouca/cime/create_test_single_exe: create_test: Add new single-exe option | 13 February 2019, 21:08:32 UTC |
a94a6fb | Jon Wolfe | 13 February 2019, 15:36:24 UTC | Merge branch 'milena/scripts/update-IC-machines' (PR #2641) Update machine files for LANL IC This PR updates machine files for LANL IC machines: * removes wolf, which has been deactivated * updates grizzly * introduces badger, which is not fully supported yet [BFB] | 13 February 2019, 15:36:24 UTC |
0400d9a | asteyer | 12 February 2019, 23:10:24 UTC | added the imkg252a method as tstep_type = 8 | 12 February 2019, 23:10:24 UTC |
f567a62 | Junqi Yin | 12 February 2019, 14:43:11 UTC | Merge branch 'bishtgautam/lnd/fix-WaterBudgetMod' (PR #2688) Minor fixes to WaterBudgetMod Incorporates minor fixes to WaterBudgetMod for IBM compiler as identified by Min Xu. Fixes #2687 [BFB] | 12 February 2019, 14:43:11 UTC |
ec95980 | Patrick Worley | 10 February 2019, 19:21:43 UTC | Add performance archiving support for cli133 at OLCF Support for a new project (cli133) needs to be added for Titan/Eos. [BFB] | 10 February 2019, 19:21:43 UTC |
65ccd97 | Jon Wolfe | 07 February 2019, 17:57:55 UTC | Merge branch 'jonbob/scripts/add-jra-forcing' (PR #2712) Initial scripts changes to add JRA support to mpas-ocean This PR brings in the scripts changes to support JRA forcing in mpas-ocean, including new compsets and grids. These new configurations will need to be tested and validated before they can be considered scientifically supported. Tested with: * SMS.TL319_oEC60to30v3.GMPAS-JRA.anvil_intel * SMS.T62_oEC60to30v3.GMPAS-IAF.anvil_intel [BFB] | 07 February 2019, 17:57:55 UTC |
b1ac533 | James Foucar | 31 January 2019, 18:07:21 UTC | Merge branch 'worleyph/cime/startstop_fix' into master (PR #2696) modify behavior of t_startstop_valsf The perf_mod command t_startstop_valsf (GPTL command GPTLstartstop_vals) is used to create a timer or add time to an existing timer when t_startf/t_stopf cannot be used to time the corresponding event. The motivating example is measuring the time spent in MPI_Init (before the timing library has been initialized). When the "call count" parameter callcount is set to 1, the max and min for the event are updated as expected. However, when callcount is 0 (adding time without incrementing the call count) or greater than 1 (adding time associated with multiple calls that are not recorded individually), max and min cannot be updated in the usual way. The current implementation is insufficient even given this basic ambiguity, and this PR modifies the min/max update logic to improve upon this. [BFB] * origin/worleyph/cime/startstop_fix: modify behavior of t_startstop_valsf | 06 February 2019, 20:15:20 UTC |
4e4f213 | Junqi Yin | 06 February 2019, 18:05:35 UTC | Merge branch 'bishtgautam/lnd/add-finidat-for-ne30' (PR #2681) Adds new land initial conditions at ne30np4 Land initial conditions at ne30np4 are added for 1850 and 2000. The new initial conditions are for satellite phenology mode (i.e. non-BGC) and were created by remapping the following files: * For 1850, clmi.I1850CRUCLM45SP.0521-01-01.0.9x1.25_g1v6_simyr1850_c140111.nc. * For 2000, clmi.ICRUCLM45SP.2000-01-01.0.9x1.25_g1v6_simyr2000_c140111.nc. Fixes #2523 [non-BFB] | 06 February 2019, 18:05:35 UTC |
73651d6 | Min Xu | 05 February 2019, 04:06:22 UTC | Add the NETCDF_PATH env for Summit *********1*********2*********3*********4*********5*********6*********7** The MPAS components use the NETCDF_PATH env to find the NetCDF package and the env has not been defined in the machine and compiler xml files. [BFB] - Bit-For-Bit | 05 February 2019, 04:06:22 UTC |
71f4e19 | Mark Taylor | 02 February 2019, 22:38:21 UTC | Merge branch 'oksanaguba/homme/bench-nl-theta' (PR #2693) NEw namelists are created for theta-l hydro and nonhydro versions for NGGPS and V1 benchmarks for both Euler and SL transport. There will also be namelists for ne1024 (3km). Note that all new namelists will use lim9. namelists only, skipping integration testing [BFB] | 02 February 2019, 22:38:21 UTC |
5fc2505 | Mark Taylor | 02 February 2019, 22:37:32 UTC | Merge branch 'mt5555/nggps-nh' into next (PR #2708) namelists only, skipping integration testing [BFB] | 02 February 2019, 22:37:32 UTC |
b72f231 | Mark Taylor | 01 February 2019, 23:44:38 UTC | nu_q should be 0 for SL transport | 02 February 2019, 22:28:21 UTC |
65cbbc5 | Mark Taylor | 01 February 2019, 22:21:12 UTC | updates to hydrostatic SL namelist so it will work with theta | 02 February 2019, 22:28:21 UTC |
f568312 | Mark Taylor | 01 February 2019, 16:04:43 UTC | hopefully final tweaks to NGGPS NH namelists | 02 February 2019, 22:28:21 UTC |
60f9341 | Mark Taylor | 25 January 2019, 22:37:31 UTC | tweaks to NE1024 namelist | 02 February 2019, 22:28:21 UTC |
55afe79 | Mark Taylor | 25 January 2019, 13:24:14 UTC | adding cori run script for NH benchmark | 02 February 2019, 22:28:21 UTC |
c85cf28 | Mark Taylor | 25 January 2019, 02:16:55 UTC | scripts and namelists for NH version of NGGPS benchmark | 02 February 2019, 22:28:21 UTC |
44d5be1 | Mark Taylor | 02 February 2019, 22:26:31 UTC | Merge branch 'ambrad/homme/vertical-remap-fix' (PR #2709) This commit focuses on the line kk = k ! Keep from an order n^2 search operation by assuming the old cell ! index is close. in remap_Q_ppm. It is not always the case that this value of kk gives pio(kk) <= pin(k+1). Then the subsequent search cannot succeed. To fix this, if pio(kk) > pin(k+1) perform a binary search to find the correct kk. [CC] This fixes a rarely triggered bug and may be BFB w.r.t. to our existing tests, but is potentially [CC] | 02 February 2019, 22:26:31 UTC |
b3e0d39 | Oksana Guba | 02 February 2019, 04:56:31 UTC | minor | 02 February 2019, 04:56:31 UTC |
b90f056 | Oksana Guba | 02 February 2019, 04:39:09 UTC | deleting theta h namelists, fixing hydro sl list | 02 February 2019, 04:39:09 UTC |
53f1834 | Jon Wolfe | 31 January 2019, 21:31:44 UTC | Merge branch 'jonbob/seaice/fix-oRRS30to10v3wLI-IC' (PR #2718) Fix seaice initial conditions for oRRS30to10v3wLI grid This PR changes the latitudes north and south that define the initial seaice disks as default initial conditions for the oRRS30to10v3wLI grid. The current initial conditions result in failures in the ocean model when the ice sheet melt fluxes are off, due to a negative layer thickness. Initializing with reduced seaice initial disks seems to allow the models to spinup together without issue. Tested with several one year runs on theta, with ice shelf melt fluxes on and off Fixes #2606 Fixes #2654 [non-BFB] only for G-compsets using this medium-res grid [NML] only for configurations using this medium-res grid | 31 January 2019, 21:31:44 UTC |
395b818 | Jon Wolfe | 31 January 2019, 20:10:16 UTC | Fix seaice initial conditions for oRRS30to10v3wLI grid | 31 January 2019, 20:10:16 UTC |
fa9300f | James Foucar | 31 January 2019, 18:45:14 UTC | Merge branch 'oksanaguba/homme/thetal-sha' into master (PR #2715) add sha to theta-l [BFB] * origin/oksanaguba/homme/thetal-sha: add sha to theta-l | 31 January 2019, 18:45:57 UTC |
18077e3 | Min Xu | 31 January 2019, 03:17:59 UTC | Merge pull request #2701 from E3SM-Project/machines/summit/update-modules Update modules and paths on Summit | 31 January 2019, 03:17:59 UTC |
ddac584 | Min Xu | 31 January 2019, 02:36:23 UTC | Update the software env and path on Summit - Update the software env for PGI and GNU compilers - Set E3SM pathes (run, bld, case, short-term archive directories) under project directory - Set E3SM input, baseline and tool directories under cli115 world shared directory | 31 January 2019, 02:36:23 UTC |
2aa6b9c | Jon Wolfe | 30 January 2019, 20:50:59 UTC | Combine COREFORCED and JRAFORCED flags into DATMFORCED | 30 January 2019, 20:50:59 UTC |
932f14c | Balwinder Singh | 30 January 2019, 17:50:38 UTC | Merge branch 'polunma/atm/eamv2' (PR #2679) Gustiness, tpert, wsubmin for EAMv2 This includes subgrid wind gustiness (ocean and land; ZM and clubb), subgrid temperature (tpert), lower wsubmin for activation, and theta-v formula fix. These code mods are controlled by namelist. EAMv1 is BFB. [BFB] [NML] | 30 January 2019, 17:50:38 UTC |
95ede2a | Jon Wolfe | 30 January 2019, 15:39:02 UTC | Merge branch 'jonbob/ocean/fix-salinity-restoring' (PR #2707) Changes to mpas-ocean and its driver to fix issues with salinity restoring This PR brings in a new mpas_source submodule which includes a fix to the salinity restoring forcing time interval (see MPAS PR #138). It also includes a reordering of the mpas-ocean driver so that the salinity restoring forcing passes ERS testing. Tested with: * ERS.T62_oEC60to30v3.GMPAS-IAF.anvil_intel * ERS.ne30_oECv3.BGCEXP_BCRC_CNPRDCTC_1850.anvil_intel [non-BFB] (only for G-cases) | 30 January 2019, 15:39:02 UTC |
288ee8c | Balwinder Singh | 29 January 2019, 17:53:03 UTC | Fixes a GNU compile time error The error was due to gust_fac being declared as an array and used as a scalar: /global/cscratch1/sd/bsingh/delete/tmp/gnu_compile/E3SM/components/cam /src/physics/cam/clubb_intr.F90:1253:30: real(r8) :: gust_fac(pcols) 1 Error: Argument 'gust_fac' of statement function at (1) must be scalar /global/cscratch1/sd/bsingh/delete/tmp/gnu_compile/E3SM/components/cam /src/physics/cam/clubb_intr.F90:1248:20: real(r8) :: ugust ! function: gustiness as a function of convective rainfall 1 Error: Incompatible ranks 0 and 1 in assignment at (1) /global/cscratch1/sd/bsingh/acme_scratch/cori-knl/SMS_Ln1.ne4_ne4. FC5AV1C-L.cori-knl_gnu.gnu_tst_del/Tools/Makefile:934: recipe for target 'clubb_intr.o' failed gmake: *** [clubb_intr.o] Error 1 [BFB] - Bit-For-Bit [NML] - Namelist Changing See confluence for a more detailed description about these tags. | 29 January 2019, 17:56:54 UTC |
fcbd574 | Mark Taylor | 29 January 2019, 14:30:22 UTC | Merge branch 'azamat/homme/add-kind-parameter' (PR #2700) Add kind parameter to centroids in mesh_mod. [BFB] except for 1 HOMME subtest | 29 January 2019, 14:30:22 UTC |
96bd7f6 | Robert Jacob | 28 January 2019, 18:51:00 UTC | Merge branch 'worleyph/components/taskmap_control2' (PR #2672) A recent commit added the env_run.xml parameter INFO_TASKMAP_COMP and the associated cime_pes namelist parameter info_taskmap_comp to enable (> 0) or disable (<= 0) the output of the task-to-node mapping for the component models. If enabled, it further specifies whether to use a compact (== 1) or a more verbose (> 1) output format. Here the necessary control logic is added to CAM, CLM, Mosart, mpas-albany-landice, mpas-ocean, and mpas-seaice. At the moment, the default for INFO_TASKMAP_COMP is 0, so this removes all component-specific task-to-node mapping information from the log files but allows us to experiment with this and supports writing out all task-to-node mapping information in the new more compact representation simply by changing the default for INFO_TASKMAP_COMP to 1 (or 2). [BFB] | 28 January 2019, 18:51:00 UTC |
6fb9b56 | Oksana Guba | 28 January 2019, 15:43:35 UTC | add sha to theta-l | 28 January 2019, 15:43:35 UTC |
c9f0178 | Mark Taylor | 26 January 2019, 16:57:26 UTC | Merge branch 'HStopo3' (PR #2699) Support of ultra high resolutions in standalone HOMME Add support for 'netcdf4p' and 'pnetcdf64' formats change "compdof" array to integer*8 (including intermediate calclulations) use -D_NO_MPI_RSEND for PIO builds, following E3SM practices. This prevents hangs during the PIO1 box rearraanger on KNL Non scalable temporary arrays in SetElemOffset() function: reduced from 3 arrays to 2 arrays. This will have to be fixed in the future if we want to run NE>=4096 for nodes with ~1GB per MPI task. bug fix (unitialized variable) in surfarea() added missing edgebuf deallocate Also has updates and tweaks to many of the scripts used to test and debug the above ( ultra-high resolution DCMIP 2016 test1 and SCRIP template generation tool). Fixes #2697 [BFB] | 26 January 2019, 16:57:26 UTC |
1cb8a84 | Balwinder Singh | 25 January 2019, 22:49:24 UTC | Adds some cosmetic changes and removes code which is not required [BFB] - Bit-For-Bit [NML] - Namelist Changing See confluence for a more detailed description about these tags. | 25 January 2019, 22:49:24 UTC |
a3aaada | Mark Taylor | 25 January 2019, 16:48:30 UTC | use -D_NO_MPI_RSEND for PIO builds, to prevent flow control hangs | 25 January 2019, 16:48:30 UTC |
c5555f0 | Mark Taylor | 25 January 2019, 02:12:54 UTC | tweaks to template run script | 25 January 2019, 02:12:54 UTC |
c896819 | Mark Taylor | 25 January 2019, 02:09:49 UTC | reduce non-scalable init memory | 25 January 2019, 02:09:49 UTC |
210fd00 | Mark Taylor | 24 January 2019, 21:32:54 UTC | adding makegrid-cori run script | 24 January 2019, 21:32:54 UTC |
4249f8e | Mark Taylor | 24 January 2019, 20:14:38 UTC | reorg run script for anvil | 24 January 2019, 20:14:38 UTC |
f2c8b0f | Jon Wolfe | 24 January 2019, 19:59:39 UTC | Changes to mpas-ocean and its driver to fix issues with salinity restoring | 24 January 2019, 19:59:39 UTC |
bc96cd2 | James Foucar | 18 January 2019, 20:02:45 UTC | Merge branch 'homme/preqx_kokkos-summit-mods1' into master (PR #2695) Homme/preqx kokkos summit mods1 This PR does the following: * Fix a compile error in the code meant to handle the case that MPI does not handle on-device pointers. * Work around a Cuda 9.1.* parse error in this same section of code. * Fix some deprecated Kokkos usage. * Fix a bug in Homme in which accum%IEner_wet was not being 0-initialized, leading to NaNs in repro sum in preqx_kokkos. [BFB] * ambrad/homme/preqx_kokkos-summit-mods1: Homme: Change means of 0-initing IEner_wet. Homme: Fix a warning. Homme: IEner_wet does not exist in all models, so safeguard it. Hommexx: Update a Kokkos call to the new name. Homme: Fix a bug in 0-initialization of energy diagnostic arrays. Work around Cuda 9.1.85 parse error. Fix syntax error. | 24 January 2019, 19:10:06 UTC |
09ab663 | Jon Wolfe | 24 January 2019, 18:12:55 UTC | Merge branch 'darincomeau/seaice/dib_ers_bugfix' (PR #2705) Bug fix so that compsets with data icebergs (DIB) pass E3SM ERS tests This PR fixes a bug where data icebergs were not being properly read in on restarts in E3SM. This PR only impacts cryosphere science campaign compsets with data icebergs (DIB). Tested with: * ERS.ne30_oECv3wLI.A_WCYCL1850-DIB_CMIP6.edison_intel [BFB] | 24 January 2019, 18:12:55 UTC |
d2dbbe5 | Junqi Yin | 24 January 2019, 15:51:10 UTC | Merge branch 'bishtgautam/lnd/add-2010-to-mksurfdata' (PR #2678) Extends supported years in mksurfdata Adds 2010 as a supported year [BFB] | 24 January 2019, 15:51:10 UTC |
fbe8704 | Darin Comeau | 23 January 2019, 19:18:25 UTC | Bug fix so that compsets with data icebergs (DIB) pass E3SM ERS tests. | 23 January 2019, 19:18:25 UTC |
d430595 | Mark Taylor | 23 January 2019, 16:32:09 UTC | bugfix to allow utilities to run with nmax=0 (no timesteps) | 23 January 2019, 16:32:09 UTC |
6add7f8 | Mark Taylor | 23 January 2019, 15:01:46 UTC | simplified makegrid.job script, namelist | 23 January 2019, 15:01:46 UTC |
b4fbc32 | Mark Taylor | 23 January 2019, 01:52:36 UTC | bugfix in area calculation - uninitialized variable | 23 January 2019, 01:52:36 UTC |
91747d3 | Gautam Bisht | 22 January 2019, 18:57:41 UTC | Merge branch 'bishtgautam/script/updates-for-lawrencium' (PR #2686) Updates machine files for lawrencium-lr2 and lawrencium-lr3, while new support for lawrencium-lr6 is added. Additionally, support for GNU compiler for lawrencium-lr6 is added. [BFB] | 22 January 2019, 18:57:41 UTC |
8352dd4 | James Foucar | 22 January 2019, 18:01:36 UTC | Merge branch 'jgfouca/cime/improve_bless_robustness' into master (PR #2703) Improve bless_test_results robustness Using xmlquery to get case values opens the possibility of permissions errors if the source repo is not visible to the current user. Instead, open a Case object which was already needed in some code paths anyway. [BFB] * jgfouca/cime/improve_bless_robustness: Improve bless_test_results robustness Update CIME to ESMCI cime5.7.6 (PR #2689) | 22 January 2019, 18:02:27 UTC |
9ae072a | James Foucar | 22 January 2019, 18:00:20 UTC | Improve bless_test_results robustness Using xmlquery to get case values opens the possibility of permissions errors if the source repo is not visible to the current user. Instead, open a Case object which was already needed in some code paths anyway. | 22 January 2019, 18:00:20 UTC |
6a6fcb6 | Azamat Mametjanov | 22 January 2019, 01:08:06 UTC | Update modules and paths on Summit | 22 January 2019, 01:08:06 UTC |
a82b935 | Mark Taylor | 21 January 2019, 23:22:22 UTC | fix integer overflow in compdof fixes #2697 compdof() in HOMME changed to integer*8 All calculaitons of compdof() values done with integer*8 precision with this change, was able to run HOMME with NE=2048,20L and produce output | 21 January 2019, 23:22:22 UTC |
49b5a03 | Balwinder Singh | 21 January 2019, 18:09:43 UTC | Merge branch 'brhillman/atm/update-rrtmgp-inputdata'(PR #2675) Update RRTMGP inputdata data and fix read issue Update RRTMGP inputdata with new absorption coefficient data from the main RRTMGP repo, and convert to proper format to be read on OLCF and NERSC machines. Fixes #2565. Should be BFB for all tests other than the RRTMGP test, which will produce a namelist compare error and a baseline compare error due to the differing absorption coefficients, but this is expected. [NML] [NBFB] * brhillman/atm/update-rrtmgp-inputdata: Add PGI flag for F2003-style allocate Reset fluxes and heating rates when no day columns Remove old copy of mo_load_coefficients.F90 Make sure all dimension names have same length Replace "merge" with if statement Handle case where no sunlit points in chunk Remove old version of mo_load_coefficients Pull in read routine from rte-rrtmgp source Update RRTMGP coefficients files Remove old version of mo_load_coefficients Pull in read routine from rte-rrtmgp source Update RRTMGP coefficients files | 21 January 2019, 18:13:09 UTC |
7816bdc | Mark Taylor | 21 January 2019, 17:40:21 UTC | netcdf4p support for making ne>512 SCRIP files | 21 January 2019, 17:40:21 UTC |
4a12f75 | Mark Taylor | 20 January 2019, 19:42:22 UTC | tweaks to HStopo NCL diagnostics | 20 January 2019, 19:42:22 UTC |
77bc343 | Azamat Mametjanov | 19 January 2019, 01:26:02 UTC | Add kind parameter to centroids in mesh_mod | 19 January 2019, 01:26:02 UTC |
3ac5c8e | James Foucar | 14 January 2019, 22:51:06 UTC | Merge branch 'homme/preqx_kokkos_build_fix' into master (PR #2665) HOMME: fixing some issues with the preqx_kokkos target #2665 One issue this PR fix is #2648 . The other issue it fixes (which was not opened on github) is the duplication of files between preqx and preqx_kokkos targets. In particular, the files in the folder preqx/share were copied verbatim in the folder preqx_kokkos. This causes maintenance issues, and led to a build error when some changes were made in the files in preqx/share but not in the corresponding files in preqx_kokkos. Instead, we now simply include the preqx/share folder content in the list of files to be compiled for the preqx_kokkos target. Note: there are a few files in preqx that are copied verbatim in preqx_kokkos. It would be tempting to also remove those files in preqx_kokkos, and simply use the files in preqx. This would work in a cmake standalone homme build, but it would be problematic with cime. The reason is that not all files in the preqx folder are copied in preqx_kokkos; for instance, prim_driver_mod.F90 is different. In cmake, we can manually specify the files we want to compile, but in cime, we can only specify the folders where the files are, and cime will glob all source files in those folders. For now, we are ok with the removal of some duplicate files. We will think about how to remove the other duplicates later. [BFB] * bartgol/homme/preqx_kokkos_build_fix: i need this file under version control and i cannot use master with preqx_kokkos, so, placing it here HOMME: Fixing preqx/preqx_kokkos duplication issue Fix #2648 | 18 January 2019, 20:00:38 UTC |
c164447 | rgknox | 17 January 2019, 19:07:05 UTC | typo fix on the lawrencium partition batch files | 17 January 2019, 19:07:05 UTC |
5947727 | Patrick Worley | 16 January 2019, 23:49:33 UTC | modify behavior of t_startstop_valsf The perf_mod command t_startstop_valsf (GPTL command GPTLstartstop_vals) is used to create a timer or add time to an existing timer when t_startf/t_stopf cannot be used to time the corresponding event. The motivating example is measuring the time spent in MPI_Init (before the timing library has been initialized). When the "call count" parameter callcount is set to 1, the max and min for the event are updated as expected. However, when callcount is 0 (adding time without incrementing the call count) or greater than 1 (adding time associated with multiple calls that are not recorded individually), max and min cannot be updated in the usual way. The current implementation is insufficient even given this basic ambiguity, and this PR modifies the min/max update logic to improve upon this. [BFB] | 16 January 2019, 23:49:33 UTC |
743d9d5 | Wuyin Lin | 16 January 2019, 23:39:24 UTC | Merge bogensch/atm/SCM_omegafix (PR #2690) Large scale vertical velocity bug fix in Eulerian SCM Bug fix to allow prescribed omega to propagate to state variable. The prescribed omega (wfld) is used directly by the dynamics and this implementation has always been correct. However, CLUBB (physics parameterization) uses the large scale vertical velocity as well. Thus, since wlfd was not being passed to state correctly CLUBB was never seeing the updated wfld, only the wfld at the first time step. CLUBB is not very sensitive to changes in wfld, so the SCM solutions are not very sensitive, but not bit for bit. In addition, after this fix, the outfld conditional call for OMEGA in cam_diagnostics.F90 is not necessary and has been removed. [BFB] except for SCM mode. | 16 January 2019, 23:39:24 UTC |
7f733eb | Po-Lun Ma | 16 January 2019, 21:04:22 UTC | Rename flag to use_sgv | 16 January 2019, 21:04:22 UTC |
450bb5e | Andrew M. Bradley | 16 January 2019, 20:13:59 UTC | Homme: Change means of 0-initing IEner_wet. Instead of using ifndef, move model-specifc, prim_state_mod- specific 0-initing to prim_printstate_init. | 16 January 2019, 20:13:59 UTC |
a023d65 | Andrew M. Bradley | 16 January 2019, 20:13:39 UTC | Homme: Fix a warning. Fix ``` prim_advance_mod.F90:1200:79: elem(ie)%state%w_i(:,:,k2,nt)*stens(:,:,k2,3,ie))/2 ) / +& 1 Warning: Extension: Unary operator following arithmetic operator (use parentheses) at (1) ``` | 16 January 2019, 20:13:39 UTC |
544221a | Gautam Bisht | 16 January 2019, 20:06:30 UTC | Adds support back for Lawrencium LR2/LR3 | 16 January 2019, 20:06:30 UTC |
f4a6d08 | James Foucar | 15 January 2019, 18:07:23 UTC | Update CIME to ESMCI cime5.7.6 (PR #2689) Squash merge of jgfouca/branch-for-to-acme-2019-01-14 into master Features: * Improve error/exception handling throughout CIME * Update pio1 * Update pio2 * Improve errors messages coming from generic_xml * Remove support for CLM4.0 * bring in latest nuopc developments Bug fixes: * docn: make sure optional arg exists before using * improve the logic of the continue run check * case.cmpgen_namelists needs to be able to write to the case * Fix Case context manager enforcement [BFB] | 16 January 2019, 18:21:34 UTC |
7a5e14a | Oksana Guba | 16 January 2019, 17:01:51 UTC | minor change, no repeated defs are found for *.nl in this folder | 16 January 2019, 17:01:51 UTC |
94f9600 | Oksana Guba | 16 January 2019, 05:32:44 UTC | now it also finds multiple declarations | 16 January 2019, 05:32:44 UTC |
d47742c | Oksana Guba | 16 January 2019, 02:06:35 UTC | line that compares 2 namelists indep. of order | 16 January 2019, 02:06:35 UTC |
d1fccb3 | Oksana Guba | 15 January 2019, 22:45:52 UTC | removing extra def of nu_q. however we\d better have a utility of some sort for doubles like this | 15 January 2019, 22:45:52 UTC |
cddc305 | James Foucar | 15 January 2019, 22:10:20 UTC | create_test: Add new single-exe option Allows for all cases to share a single build. [BFB] | 15 January 2019, 22:10:20 UTC |
027fc06 | Junqi Yin | 15 January 2019, 21:58:59 UTC | Merge branch 'bishtgautam/lnd/remove-domain-gen-code' (PR #2636) Removes domain generation scripts under the land model The duplicate domain generation code is removed under the land component as the updated version of those scripts are available under CIME at cime/tools/mapping/gen_domain_files. Fixes #2635 [BFB] | 15 January 2019, 21:58:59 UTC |
4a0771f | Andrew M. Bradley | 15 January 2019, 20:39:42 UTC | Homme: IEner_wet does not exist in all models, so safeguard it. | 15 January 2019, 20:39:42 UTC |
b1f0be9 | Oksana Guba | 15 January 2019, 20:16:33 UTC | fixing hv_subcy issue | 15 January 2019, 20:16:33 UTC |
8341b6c | Andrew M. Bradley | 15 January 2019, 19:45:53 UTC | Hommexx: Update a Kokkos call to the new name. Old function name is deprecated. | 15 January 2019, 19:45:53 UTC |
1d19226 | Andrew M. Bradley | 15 January 2019, 19:44:46 UTC | Homme: Fix a bug in 0-initialization of energy diagnostic arrays. IEner_wet was not being initialized to 0. This was causing NaNs in repro sum in the Hommexx code. | 15 January 2019, 19:44:46 UTC |
0429539 | Oksana Guba | 15 January 2019, 19:44:05 UTC | adding a note that rsplit and qsplit need to be divisors of nmax | 15 January 2019, 19:44:05 UTC |
6c88a3c | Oksana Guba | 15 January 2019, 19:39:27 UTC | adding SL namelists that hasnt been run. reorganized nls grouping certain parameters together. to make it more bug-proof it would be nice to have a utility that checks nls for duplicated definitions and compares nls just for different values (because diff does not work this way). | 15 January 2019, 19:39:27 UTC |
710872b | Peter Bogenschutz | 14 January 2019, 23:28:55 UTC | bug fix to allow prescribed omega to propagate to state variable. The prescribed omega (wfld) is used directly by the dynamics and this implementation has always been correct. However, CLUBB (physics parameterization) uses the large scale vertical velocity as well. Thus, since wlfd was not being passed to state correctly CLUBB was never seeing the updated wfld, only the wfld at the first time step. CLUBB is not very sensitive to changes in wfld, so the SCM solutions are not very sensitive, but not bit for bit. | 15 January 2019, 17:43:08 UTC |
33fba63 | Po-Lun Ma | 15 January 2019, 17:41:41 UTC | Add ixcldliq in clubb_surface | 15 January 2019, 17:41:41 UTC |
01edb12 | Andrew M. Bradley | 15 January 2019, 00:03:10 UTC | Work around Cuda 9.1.85 parse error. With the syntax error fixed, it became clear that nvcc was not parsing this typename alias. Remove the layers of abstraction to get nvcc to work. | 15 January 2019, 00:03:10 UTC |
e8963fe | Oksana Guba | 14 January 2019, 23:16:19 UTC | Start on adding *nl files for thetal. I still need to run them. Diffs for H version vs v1-ne120.nl are below. Do we want lim9 from now on? [onguba@blogin1 v1]$ diff v1-ne120-thetal-h-euler.nl v1-ne120.nl 19a20 > tstep_type = 5 21,24c22 < theta_hydrostatic_mode=.true. ! HYDRO < tstep_type = 5 ! HYDRO < !theta_hydrostatic_mode = .false. ! NONHYDRO < !tstep_type = 7 ! NONHYDRO --- > theta_hydrostatic_mode=.false. 32c30 < limiter_option = 9 ! this is diff from what's for preqx --- > limiter_option = 8 Changes to be committed: new file: v1-ne120-thetal-h-euler.nl new file: v1-ne120-thetal-nh-euler.nl | 14 January 2019, 23:16:19 UTC |
535884e | Andrew M. Bradley | 14 January 2019, 23:10:51 UTC | Fix syntax error. | 14 January 2019, 23:10:51 UTC |
17051a0 | Wuyin Lin | 14 January 2019, 21:41:42 UTC | Merge wlin/atm/eul-scm-omega (PR 2685) Restore how omega is output to history file for EUL-SCM PR # 2667 introduced a change to have a more consistent treatment of calling outfld to save OMEGA field that works for SE dycore. But after the change, the prescribed vertical velocity field would not be saved as is because of a currently unidentified issue in EUL-SCM concerning state%omega. This causes integration test SMS_R_Ld5.T42_T42.FSCM5A9 to have DIFF failure, though the simulation results are not affected. The original format of outfld call is restored until a fix for EUL-SCM on state%omega is found. [BFB] | 14 January 2019, 21:41:42 UTC |
56e3e36 | Oksana Guba | 14 January 2019, 17:33:55 UTC | i need this file under version control and i cannot use master with preqx_kokkos, so, placing it here | 14 January 2019, 17:33:55 UTC |
b025ffd | Gautam Bisht | 14 January 2019, 14:49:19 UTC | Minor fixes to WaterBudgetMod Incorporates minor fixes to WaterBudgetMod for IBM compiler as identified by Min Xu Fixes #2687 [BFB] | 14 January 2019, 14:49:19 UTC |
8ebffd4 | Gautam Bisht | 11 January 2019, 20:09:16 UTC | Adds support for GNU on Lawrencium Support for gcc/6.3.0 on Lawrencium-LBNL is added. | 11 January 2019, 20:09:16 UTC |
72bf966 | Luca Bertagna | 11 January 2019, 20:08:15 UTC | Merge remote-tracking branch 'origin/master' into homme/preqx_kokkos_build_fix | 11 January 2019, 20:08:15 UTC |
9f7688c | Wuyin Lin | 11 January 2019, 15:49:37 UTC | Restore how omega is output to history file for EUL-SCM PR # 2667 introduced a change to have a more consistent treatment of calling outfld to save OMEGA field works for SE dycore. But after the change, the prescribed vertical velocity field would not be saved as is because of a currently unidentified issue in EUL-SCM concerning state%omega. This causes integration test SMS_R_Ld5.T42_T42.FSCM5A9 to have DIFF failure, though the simulation results are not affected. The original format of outfld call is restored until a fix for EUL-SCM on state%omega is found. [BFB] | 11 January 2019, 15:49:37 UTC |
5e962cd | James Foucar | 10 January 2019, 21:49:59 UTC | Merge branch 'jgfouca/homme/allow_overload_mpi' into master (PR #2683) Allow MPI overload on 'melvin-type' machines for HOMME [BFB] * jgfouca/homme/allow_overload_mpi: Allow overload on 'melvin-type' machines Get rid of deprecated use of cmp for sorting | 10 January 2019, 23:47:42 UTC |