Bug when merging patches #54
Comments
|
Note that fuse_2_patches is called two times in the code, and the other time is not followed by a deallocate. so we are reasonably confident that removing that line of code is in fact the right thing to do as it makes the call more consistent. Note that @rgknox still thinks its weird to deallocate the argument to a subroutine within the subroutine, which I tend to agree with. So the alternate strategy is to reinstate that deallocate line that follows the call to fuse_2_patches in terminate_patches, add a line to deallocate the patch after the call to fuse_2_patches in fuse_patches, and delete the deallocate call in fuse_2_patches. |
|
OK, so perhaps that solution does not in fact work. The model now hangs in year 77, which seems like the kind of thing it would do if not-deallocated patches were building up... |
|
OK, so now I'm confused. It turns out that there are some write statements already around the offending piece of code: Looking at the log files, the output from the first one of these is the following: fusing patches because one is too small 9.989301881637238E-004 Note that this is the first time I see this phrase in any of my log files, so this is definitely an infrequent occurrence. before the code changes in 59e6e6e, it would crash before reaching the next write statement. After the code changes in 59e6e6e, it did reach the write statement, and reported the following: after fusion 6483.79540623672 1.129761864022804E-026 but it seems that something is wrong as the model is then hanging two years later (unless the hanging is completely unrelated, and is just the appearance of the next bug that it would have encountered anyway, which I won't rule out completely.) Since this only happens once, it doesn't seem like it could be a memory buildup issue, but I don't really understand the logic here. @rosiealice, I'm guessing this bit of code is wrapped in write statements because it has misbehaved in the past. Any insight into what's going on here? |
|
OK, I did another test and now I think the subsequent hang is unrelated. The test that I did to arrive at this conclusion was to go back to the master code, and modify only min_patch_area in EDTypesMod.F90 to be smaller by two orders of magnitude, and then re-run from the restart before the original crash occurred. What this meant is that the relevant code in terminate_patches was never triggered and the patches were both kept as-is without fusing. I confirmed this by searching for the string "fusing patches because one is too small" in the log file and it did not appear. The result of the test is that the code proceeds past the original crash point encountered with unmodified master code (in year 74, month 12), and then still hangs at the same timestep (# 1340545, in year 77, month 06) as it does with the changes in 59e6e6e. This would seem highly unlikely to occur if the bug were being encountered in the same gridcell, since the patch structure, and therefore everything's internal state, is different as a result of changing min_patch_area. So, the hang must therefore be happening in a different gridcell. So: what this means is that: (a) the original code change in 59e6e6e is likely not the culprit, and may be the correct solution. I believe it is in fact correct, but I have no idea how to confirm that it is correct, other than by staring at the code. it will most definitely not be encountered in any of our standard tests, since it requires an older ecosystem than is possible by running for a short period starting from bare ground. (b) we have another, unrelated, bug that is causing the model to hang with no other diagnostic clues 77 years into a 4x5 simulation. this would seem to be quite difficult as I have no idea how to go about identifying such a bug. (c) this all underscores the fact that our current testing infrastructure is not detecting these edge cases that occur under slightly obscure, yet still regularly encountered, situations. I think we need to think about adding some sort of long-term stability test that may not be in the regular suite but that we still run with some regularity to find these things. 200 years at global 4x5 without crashing would seem like a decent criteria to me, and one that we cannot yet reach. |
|
Hi Charlie, Thanks for this. I agree on the need for long term tests. Given how common Apologies that this isn't more helpful, but nee work on eliminating the -Rosie On 18 April 2016 at 12:34, Charlie Koven [email protected] wrote:
Dr Rosie A. Fisher Terrestrial Sciences Section |
|
Rosie, Thanks, thats a good point that we can't completely rule out the possibility of it being in the same patch if its just a matter of something happening due to the small patch size. I guess the other thing I could try would be to increase the value of min_patch_area by an order of magnitude or so from the standard case, and then run a new case from year 1 to see if it still hangs at the same point. Presumably if the bus is related to something in that gridcell, either in an unmerged patch or in a post-merged patch, then it ought to trigger at some earlier date is the merge happened earlier due to the patch crossing the min_patch_area threshold earlier? I can try that but that still strikes me as pretty unlikely. re: the hanging bugs; yeah, this seems pretty difficult. In your experience, do these tend to be associated with infinite while loops around the patch/cohort structure? I'm quite used to making CLM crash, but not really hang, so it seems to me that this kind of thing might be a unique feature of the while loops in the ED code. One possibility (inelegant but may get the job done) would be to put some sort of extreme-debug-only write statement inside every single one of these. I count about 100 or so of these loops (142 counts of "while" in grepping through the ED subdirectory, but many of these are comments to the enddo statements) spread across 10 or so files: so a lot, but not more than an hour or so in an emacs macro could handle. That way we could bring the model just up to the edge of the hang, and then turn on the write statements to see if we are caught in such a loop. This would seem like the kind of thing that CPP tokens are actually kosher for, as they would add some debuggability to the heavy use of while loops that the ED structure requires, at no runtime cost... |
|
Hi. So, my experience was that the crash happens outside the ED code. To figure On 18 April 2016 at 13:02, Charlie Koven [email protected] wrote:
Dr Rosie A. Fisher Terrestrial Sciences Section |
|
OK, so I put in write statements at the ED entry and exit points (i.e. where they currently are, but not just for masterproc), and based on my count of entry and exit stamps in the output log, the hang is happening within the ED daily timestep. So on to plan B: assign a random string to every while loop and output that string every loop iteration to see if one of them is infinite. |
|
That's good, actually, since it means that it's likely not the same as the On 18 April 2016 at 18:27, Charlie Koven [email protected] wrote:
Dr Rosie A. Fisher Terrestrial Sciences Section |
|
agreed. for my own future reference, and possibly as a team resource if/when the next time this kind of thing crops up, attached an emacs function I found via google for writing a random string, since there is (slightly shockingly) not such a feature by default. put in ~/.emacs.d/init.el ...
|
|
well, that appears to have worked. it seems to be caught in an infinite loop within this area loop: looks like there are a bunch of cohort while loops within that, and none of them are triggering write statements, so perhaps it is in a patch with no cohorts... |
|
nice job! On 18 April 2016 at 21:38, Charlie Koven [email protected] wrote:
Dr Rosie A. Fisher Terrestrial Sciences Section |
|
OK, so I think I sorted this out. the error is in the canopy structure code, and happens when patch area is too small. So I put in some logic to only do the canopy structure calculation when patch%area is greater than min_patch_area. this is now in tag 9d0dd5d Also, this is unrelated to the first bug I encountered because I checked and they are happening in different gridcells, though both are consequences of small patch areas when things age. So two questions remain: |
|
OK, now the model crashes again a few years later. I suspect that what's going on here is that the min_patch_area criteria is not being obeyed. to me the key line is this: The hedge in that line is that it doesn't force the merge if either (a) the tiny patch is the oldest, or (b) the tiny patch is the youngest. I get why it needs to not force youngest patches to merge (since they will have area added to them via disturbance, whereas the old patches ought only get smaller). But I suggest we more strongly enforce the min_patch_area threshold at the old end. I will try out some logic to force the merge of a tiny oldest patch into its younger neighbor as a first step, and see if that helps out the stability of the model as things get older. |
|
OK, the change to better enforce the min_patch_area criteria (bb8ac55) avoids that subsequent crasher that I had run into. The model fused a bunch of patches in the first timestep after restart and for now appears to be chugging along past where it crashed before. |
|
With the most recent code change in place (bb8ac55), the model now seems like its stable on long-term runs. It has run to 120 years now and things seem to still be running smoothly. @rosiealice: I am not sure whether to undo the changes in 9d0dd5d or leave them in. There is still the potential for small patches to go through the canopy construction step if they are the youngest patch on a site, as the fusion logic still lets those persist. Checking the log files, this does seem to still be possible, though after a while a younger patch appears, and once the younger neighbor is in place the tiny patch immediately fuses into it. But it is an unlikely condition, as it has only happened once in the 50 years at the global 4x5 conditions since I fixed the fusion logic. My guess is that the canopy structure mod is acting as the canary in the coalmine for small-patch numerical weirdness because of the construction of the while loop that has a small-but-not-that-small number in the logic: But If I do leave the logic in place, I am still worried that by avoiding the whole block of code, key variables are not being set that could produce other types of weirdness. Is there anything that must be set there, and if so, what is the least dangerous value to set it to? |
|
Back from panel, trying to catch up to you guys again! The parameter min_patch_area is newly introduced. It is possible that there are locations in the code where patch area is compared to a hard-coded threshold that should be re-named min_patch_area, but I did not catch it when adding the parameter. I think we also need fates constant file for stuff like this. |
|
sorry but I have to ask, what about the "mostly" part? |
|
like I said above, if the youngest patch on a site is tiny it still doesn't get merged. in practice this doesn't appear to occur very often, though in theory it can. hence my inclination to leave that check in place for the canopy construction code to avoid issues there. |
Merge PR #58, branch 'ckoven/fusebug' The model was exhibiting instability when trying to run for global (4x5 resolution) long-term runs. some of these instabilities were coming from old patches that had extremely small areas. This was a fairly uncommon occurrence in single-point code and was not being frequently encountered, but when it was encountered, there were a series of issues that could crash the code. One issue is that it was trying to deallocate something twice. Another issue is that when it merged patches, it would still allow a patch to exist even when its size was below the minimum threshold if it was the oldest patch on the column; that is now changes so that if a tiny patch is the oldest it merges it into the younger neighbor, whereas if it is neither the oldest nor the youngest patch, it fuses it into the older neighbor. It will still allow a patch to be smaller than the minimum threshold if it is the youngest patch, since that is a situation that can and may occur transiently. In that case, however, it now bypasses a section of the canopy construction code for the tiny youngest patch that was particularly vulnerable to minimum patch size numerical issues (and, worse yet, responded by going into an infinite loop). Fixes: #54 User interface changes?: No Code review: ckoven, rgknox Test suite: ed, lawrencium-lr2 intel; yellowstone intel, gnu, pgi Test baseline: 0471ef9 Test namelist changes: none Test answer changes: should be climate changing, but only after many years Test summary: passes except for expected failures: f19 and f09 restarts #14, and gnu restarts #43.
a5a3c95 Get SPL working 790d13c Merge tag 'cime4.5.14' into gswp3drvflds b631529 Merge remote-tracking branch 'upstream/master' f4c90e5 Merge pull request #447 from billsacks/vertical_gradient_matrix_solve 50a06a7 Merge pull request #449 from fischer-ncar/usermods_inputdata 8d812c5 Merge remote-tracking branch 'upstream/master' into usermods_inputdata 42696c9 Removed inputdata files that referenced glade from usermods. 7e01d2b Merge branch 'gswp3drvflds' of https://github.com/ekluzek/cime into gswp3drvflds e76f958 Add SPL test 490e221 Merge branch 'gswp3drvflds' of https://github.com/ekluzek/cime into gswp3drvflds c177e80 WW3 description 3286612 Add ability to handle WW3 DATM b6d38ad Correct options for flux_epbal to off,ocn 5ad0fe2 Merge pull request #446 from bertinia/update_st_archive 380eb15 Minor fix to st_archive to exclude removing rpointer files 743836c update to st_archive to correctly check if interim restart file duplicates should be removed 2db6710 Add README file 9c32895 Merge pull request #443 from jedwards4b/esp_assignment_fix 7800cc7 fix issue with esp_present on restart 5396091 Merge remote-tracking branch 'upstream/master' 00ca59b Merge pull request #441 from fischer-ncar/remove_logan 2ff3b12 Merge pull request #440 from jedwards4b/jedwards/nersc_update 684431a Merge pull request #438 from mnlevy1981/mlevy/compare_namelist 00a572b Merge pull request #437 from jedwards4b/shr_orb_mod aab40b3 Remove reference to logan from config_compilers 6cb731f fix a bug or two 775ddea Add a comment cef70eb Back out changes related to vertical_gradient_calculator_continuous 6bd1820 Merge tag 'cime4.6.0' into gswp3drvflds 9147cb9 Don't remove existing drv_flds_in with the driver merging the versions from the components that doesn't need to happen bb9c2a9 Add in ESP component 3f832c1 Merge tag 'cime4.5.9' into gswp3drvflds 29d1b17 Merge branch 'master' into mlevy/compare_namelist 3cb7c2a re add #118 correction to shr_orb_cosz 68bcf07 Merge branch 'master' into gswp3drvflds 8ff70f4 Merge remote-tracking branch 'upstream/master' 0d7f4f3 Merge pull request #436 from fischer-ncar/revert_Intel 2b7bbfd Reverting Intel compiler version on Yellowstone. 1d5c94f Merge remote-tracking branch 'upstream/master' into revert_Intel 7239f97 Add a bunch of tests with real output 2087fb7 Go back to setting y range ourselves c928198 Output initial guess, unlimited and limited for each test cb4f44b Move tests that do output to their own file 78ea018 Clean up plots a0ab8f0 Make routine public 14173df Use factory method, call print method 0ab8f0f Print various statistics ad6bb83 Refactor limit_gradients eb6149f Specify initial guess and limit gradients 5caac60 Change the stub calculator to allow specifying any gradients 58b8535 Separate computation stage of gradient calculators 25bcadb Move stuff out of vertical_gradient_test_utils f5c1ce8 Simplify vertical_gradient_calculator_continuous constructor 93c1920 Simplify vertical_gradient_calculator_2nd_order constructor f65b5cb Remove need to pass in nelev b827905 Remove unnecessary ability to specify min_elevation_class 8eb4a3c Add start of create_vertical_gradient_calculator_2nd_order be9126f Begin work on factory to create gradient calculators 6d901d8 Add place-holder for limiting gradients 2be8d78 Add more tests for plotting, and tweak plotter 9484ff6 Merge pull request #434 from fischer-ncar/SSP_testreporter_fix 3571f14 Merge remote-tracking branch 'upstream/master' into SSP_testreporter_fix 7d94209 Merge pull request #433 from jedwards4b/jedwards/tuning 78099a2 Merge pull request #432 from gold2718/PGIbugfix b976c24 Add test functionality report to SSP test 3601e78 For problematic topo values, set gradient to 0 rather than NaN 351fb92 Handle topo values out of bounds 5a0c561 Add assertion for expected gradients 0a0f504 update intel compiler on yellowstone 7979df5 Switch production code to use new gradient calculator f647995 Output gradients to file bcc5167 Remove unused variables 281bb95 Add a basic smoke test of the new calculator d280f3b Move some shared test utility functions to a new file 8a63e33 Add vertical_gradient_calculator_continuous d828818 Add a script to plot gradients 706e52b Change tridiagonal inverse test to have more unique values 8d0ed86 tune b1850 f09_g16 default 0fb6c67 Workaround for seq_infodata_mod PGI compiler bug 369d3bd Add a function to invert a tridiagonal matrix 1f444f1 Merge remote-tracking branch 'upstream/master' into shr_frz_mod_fix 27155d4 Merge pull request #428 from gold2718/ESPbugfix fc51d10 Fixed ESP errors found in prealpha testing 535181f Change CLM_ACCELERATED_SPINUP for CLM SSP test. Add merge of drv_flds_in namelists from different components. Add die_on_conflict option to Namelist.pm merge_nl a3d25af fix bug in string statement, load balance B1850 f19_g16 c4bda40 Add merge of component drv_flds_in into driver, and also add a die_on_conflict option to Namelist.pm merge_nl 2287638 Fixed max_alarms parameter c6c4979 Wav clock offset fix ea2279e Merge pull request #426 from fischer-ncar/wav_clock_fix e3c1fcf Fix wav clock offset in the driver. ede0eda Esp compset description fix c5721bd Merge pull request #425 from jedwards4b/esp_compset_description_fix 0bec14e Ensure that elevation class bounds are monotonically increasing d30f22c Comment out call to check_topo e9e7e70 Add a test demonstrating non-monotonicity 0cd0cde Revert "Demonstrate that it works to have topo going high to low" f93e16a Demonstrate that it works to have topo going high to low 5867dbc Add multi-point test, and a comment in production code 9950ba0 Tweak ECbottom unit test 25fe31d Add code to limit the gradient, and associated unit tests fda0e5a Add more tests that should almost (but not quite) trigger the limiter 4f82e95 Rename tests 685515e Add a test where gradient is almost limited 6f692a8 Make a parameterized test method 4538e74 Refactor: change where statement to loop with conditional a24c07e Finish changes for adding elevation class bounds to gradient calculator 823b635 Add elevation class bounds to vertical gradient calculator bf715bb Remove dependence of vertical_gradient_calculator on glc_elevclass_mod 3a3ac1c Add a routine to glc_elevclass_mod: glc_all_elevclass_strings 039673e a workaround but not a fix b307ac9 fix esp compset description to match on no compset provided 9f075fd Turn off diurnal flux, and other fixes. f32e2fe Merge pull request #424 from fischer-ncar/diurnalfluxoff 2f7dbcd Merge pull request #423 from jedwards4b/configcompsetfixagain 4ba7f8e Fix Intel module name for corip1, should be 16.0.2.181 instead of 2016.0.2.181 7cbe05f remove blank line ; 16fcbd6 remove comment line 7a7ae4e Turn off diurnal flux, fix B1850Cw compset, move SMS_D.f19_g16.B1850 to yellowstone, update cori Intel compiler. c0a99da another fix for compset logic 8a8be68 Merge remote-tracking branch 'upstream/master' 7de299f Merge branch 'apcraig-diurnal_fix1' b0b2008 Merge branch 'diurnal_fix1' of https://github.com/apcraig/cime into apcraig-diurnal_fix1 8a80287 fix shr_flux_atmocn_diurnal restart problem related to cold start initialization logic fa75de9 Merge remote-tracking branch 'mydev/user_compset_fix2' merge pr 8e4252d Merge branch 'gold2718-DART' d54e632 Merge branch 'DART' of https://github.com/gold2718/cime into gold2718-DART merge to master c44bacd Added comment in component_run for ESP model use 77a0d0a need an additional fix for user compsets that do not match any defined alias 3fc6a38 Merge pull request #419 from fischer-ncar/fixes 024d973 For the B1850s compsets that are based off or the B1850 compsets, same versions of CAM,CLM,POP,CICE.... The CROP option was removed, and %ECO and BGC%BDRD was added. This was done to make the B1850 compsets more consistent. 032abf2 Added ESP component and SESP (stub ESP model). 29e0582 Merge remote-tracking branch 'upstream/master' into fixes 2fe1ac3 Set flux_diurnal to true, update debug smoke test for B1850 270de50 Set the default flux_diurnal to true in the driver. Changed the smoke debug test for B1850 to f19_g16. bfe2f44 Merge remote-tracking branch 'upstream/master' into fixes 4812f7f Merge pull request #415 from fischer-ncar/fixes aa3167c Merge pull request #414 from bertinia/st_archive_fix 96446da Merge pull request #413 from fischer-ncar/test_dev 24d94d1 Merge pull request #411 from billsacks/glc_mapping_files 36d31d4 Merge pull request #410 from jedwards4b/user_compset_nextstep 5f463fa Merge pull request #406 from billsacks/jg_compset 8866fee Merge pull request #405 from fvitt/waccm_cime4.4.9 5a0080f Removed ic ice_ic from the usermods, this is now being set as the defualt. Add Hb=min(Hb,0.0) in shr_flx_mod to ensure Hb is <=0.0. If Hb is > 0.0 there is a floating point error. 89a13fc Update st_archive to fix CIME issue #412 e2e4189 Add 5 day debug prebeta smoke test for f09_g16 B1850 2883636 Change RTM tests to MOSART,fix long line, b03d7cc Add mapping files for new glc-lnd grid combinations a23c7bc Define new grids that include an active glc grid 49679bb allows user_compset and alias compset to get same result 97ed005 add usage message for -keepexe flag in create_clone dec0d59 Merge pull request #404 from billsacks/pr_template 7430725 Rename compset alias "BW1850" as "B1850Cw" 728eb9d Revert "Revert my commits on master" 579b4b6 Merge pull request #407 from jedwards4b/user_compset_fix 0a1e76b no warning if user provides compset de7b8c6 Revert "Validate the drv namelist" 3d47d92 Validate the drv namelist 2184961 fix issue with user_compset option 1f1eca7 Add some comments pointing to open issues 98307f8 Add WACCM compset BW1850. Remove "CCMI" B compsets (C2R). 315e717 Add a J compset test to the prebeta test suite cb918ec Add note about CPL_ALBAV 7bec46c Add pull request template f2262ab Change some comments in driver's config_component.xml 4fb2923 Merge remote-tracking branch 'origin/jg_compset' into jg_compset 4e967f4 Change J compset layout to be reasonably load balanced 45f47d9 Make compset match more specific for ROF_NCPL 122732f Change config_pes block so that it matches b05f748 Make compset match less greedy for PE layout matching BGC dc78ff7 Add a new J compset: all active except data atmosphere e460c93 Merge pull request #399 from fischer-ncar/test_dev 5cfc0d0 Merge pull request #398 from billsacks/pelayout_b1850_x dd5e256 Change "X" PE layout for B1850-f09 on yellowstone 5acd97b add user_nl files for b1850 usermods 0bc39f9 add user_mods for b1850 bacf592 redefine B1850 13e7bd5 add hybrid start info for B1850 eb13a18 Merge remote-tracking branch 'upstream/master' into test_dev 7e18dc0 Fix testreporter.pl status, use defaultio testing 57b0150 Merge branch 'pySECTupdate' update to pyCECT 2.0.2 8599ce1 fix merge conflicts 1acd234 bring in pyCECT 2.0.2 from cisl 3e74c29 remove old pyCECT a02f110 Merge branch 'salinity-temp' 3eed7c6 Merge branch 'salt' of https://github.com/dabail10/cime into salinity-temp 48c17dd Merge branch 'clean_build_comment' of https://github.com/billsacks/cime into salinity-temp 6cb2d71 Revert my commits on master e48aee0 Revert "Add setting of BARRIER for I compsets" f59d63a Add only clauses. 3028024 Add setting of BARRIER for I compsets 0f3323e Merge branch 'master' of https://github.com/ekluzek/cime a2ca7a5 Fix syntax error. b0c94a5 Merge branch 'master' of https://github.com/CESM-Development/cime f82337f Add in drv_flds namelist definition 1556bfd Add a comment about why we don't clean the shared clm library in a test 4e4c39c Merge pull request #392 from jedwards4b/RESUBMIT_SETS_CONTINUE_RUN 17a0263 Merge pull request #391 from bertinia/testreporter_fix c18d16a Merge pull request #388 from jedwards4b/Makefilecleanup d75ba3f fix testreporter.pl, SMS and PFS test output, cori queue settings 8917e74 address issue #386 with this new variable 34e9111 correct syntax in clean_build 908c619 do not clean shared land library in a test 6d2accc Merge pull request #60 from NCAR/ejh_23 88a6eb6 shortened lines so that NAG compiler will not complain 7c39f57 correct comparison error bf72eec More cleanup. 1e26343 fix clean build issue in Makefile bug in xmlchange 50cad45 Cleaning up freezing point function. 6e3fecb Updated freezing point module. f06007d Merge remote-tracking branch 'cesm-master/master' e02e632 Change name of Solar for GSWP3 data b5956b9 Updated freezing point code. 2b67a27 Merge pull request #383 from jedwards4b/documentCOMP_RUN_BARRIERS ebd5c62 Merge pull request #59 from NCAR/ejh_21 762c0b0 improve documentation of COMP_RUN_BARRIERS variable 10be4d1 improve documentation of COMP_RUN_BARRIERS variable 46785d0 Add a comment. b9ba8a1 Add some bounds on the calculation of the freezing point. 70e680a fixed handling of errors in set_chunk_cache functions 9751a59 fixed handling of errors in set_chunk_cache functions c0f2990 Merge pull request #381 from jedwards4b/buildchanges 015fa98 Merge pull request #377 from bertinia/tests_update f00b753 Merge pull request #376 from Katetc/master e9dc09b new treatment of user_compset, bug fix for make clean 2fd4a49 fixes based on Jim's comments to pull request f738fad fixes based on Jim's comments to pull request 6d398bf removed printf statement 9904cfc New csm_share moduled to compute the freezing point of salt water. 7e0ea19 using PIO_Offset instead of size_t in API 578919c got tests working with var and file chunk caches. 8edfbf5 more fixes for chunk cache functions 2437220 further development of fortran set_chunk_cache tests d8a042f further development of fortran set_chunk_cache tests b3b4e08 minor update for testing ccc56c4 Breaking up four long lines into 8 shorter ones to statisfy Nag compiler requirements. 77ebc2b Merge pull request #1 from CESM-Development/master d0191ec Modify DOCN and DICE to use salinity dependent freezing point. 8e2f9c3 Merge remote-tracking branch 'origin' d1bf8cc Merge branch 'master' of https://github.com/CESM-Development/cime 727cffb [ 50 character, one line summary ] cd9f1f2 Merge remote-tracking branch 'cesm-master/master' eb4ad98 Merge pull request #374 from apcraig/cplhisthgt 416e9b3 Modify seq_io_mod.F90 so an avect field named "hgt" is never written to the history file. This is a temporary workaround until models initialize the hgt field in the domain datatype. The exact restart test was failing in cases where hgt was not initialized (and not used) simply because it appeared on the coupler history file and was not defined uniquely in those cases. 0426d0d Merge pull request #369 from brian-eaton/grid_mod 657335a Merge pull request #370 from bertinia/remove-expectedFails 42ebbc4 Merge pull request #372 from bertinia/batch_update 43fbb2a Merge pull request #373 from bertinia/hobart_pes_update d60b543 rest of nc4 functions in fortran API 3a98a06 more testing 854eb87 more testing 1a1c246 more tests for nc4 performance functions 0326776 update hobart pe layout for all active f19_g16 db73024 Merge pull request #58 from NCAR/ejh_20 cd1c596 set chunk cache function now working 0d1f692 further testing, other variations of fortran functions 04165b5 update config_batch.xml to remove pbs directives d979112 test cleanup 4af979f Add correct value of ATM_NCPL for T42 grid. 791d9d1 fixed fletcher32 filter inq and documentation 1d15d18 fixed fletcher32 filter inq and documentation 21b3940 Merge pull request #57 from NCAR/ejh_18 9410243 test cleanup 929a091 fixed test output 93c8375 fixed bug in inq_var_deflate, got fortran version and tests working be8e885 Remove ExpectedFails.xml file from CIME and update testreporter.pl c70d96e added inq_szip fortran function 0056ace fixed name of inq_var_deflate in fortran API 452fd3a more tests for nc4 functions f839012 fortran tests for def_var_deflate turned back on 1ada5bd add T85_T85 to config_grids.xml 191d036 Merge branch 'master' of https://github.com/Katetc/cime Fixes output issues with check_input_data dbd94c4 Merge branch 'constance' of https://github.com/apcraig/cime Update Machine constance f3cfb20 This branch is called mlevy/southern_ocean in the apcraig fork. There are two tags, southern_ocean_n01 which is just the streams4d update and southern_ocean_n02 which adds the stream open/close and the BARRIER_OPTION features. These changes add support for reading 4d stream data (3d space + time) and single column data, both which were tested in POP. The stream open/close feature reduces the number of open and closes done on stream input files to improve performance. The BARRIER_OPTION adds a new alarm to the driver to synchronize all tasks at a frequency specified by the BARRIER_OPTION/BARRIER_N setting. This is needed in some concurrent cases (datm/clm) where the datm model can advance in time asychronously from the rest of the model and cause MPI memory issues as data is waiting to be received. 157d136 Merge branch 'southern-ocean' merging to master 5b7af03 Merge branch 'mlevy/southern_ocean' of https://github.com/apcraig/cime into southern-ocean Merging to cime master 25aad26 Changed logger calls to prints in check_input_data to make sure the output capture in checkinputdata actually works, and missing files are appropriately downloaded. 89e1b54 add BARRIER_OPTION to driver and new logic for opening and closing files in the streams input to improve performance. d6aeb17 fixed comment b45360c Merge pull request #56 from NCAR/ejh_17 778a243 fixed potential memory leak in error condition, other error handling code a7e2b4d fixed other nc4 inq functions so now result is bcast 748f357 even better handling of null pointers d28ef0b better handling of null pointers e649334 removed print statement 785cd2a added c99 flags to unit test directory 912c392 removed redundant broadcast 29f0abd removed unneeded test program f09d69b fixed some nc4 code ce3e95f fixing def_var_deflate issues 65df1e8 Merge pull request #54 from NCAR/ejh_15 6ff466c moved name null-terminator fixes to perl code which generates pio_nc.c 1259f43 Update compsets and grids for single column pop capability. f2691ba test and fixes for name null terminator problem in processors not rank=0 70c007b getting test_names ready 93d4305 avoid fill pass with netcdf4p 260864d fix pio_inq_unlimdimid and add a test 8cf3d63 Merge pull request #52 from NCAR/ejh_14 45fb0f7 fixed compile problem when timing turned off e2db73b clean up debug write 81deb7b Merge remote-tracking branch 'origin/master' into mlevy/southern_ocean a3e7501 update cime for pop single column and strm4d capabilities. ae4829f Minor updates to the testlist_allactive.xml 1b79608 Update machines files for constance. b751502 Revert pull request cf010df146114aa88ba772bfc5953b5573783999 1405acf Merge remote-tracking branch 'cesm-master/master' 524496f Merge branch 'jedwards4b-constance' d0f655c Merge branch 'constance' of https://github.com/jedwards4b/cime into jedwards4b-constance 94e2837 Merge branch 'bertinia-machine_updates' b86840c Merge branch 'machine_updates' of https://github.com/bertinia/cime into bertinia-machine_updates a8ec7d7 Merge branch 'cacraigucar-geotrace_branch_cime' - added first file from geotracer branch 8bac184 modified: driver_cpl/bld/build-namelist - added first file from geotracer branch f396cce batchredirect default should be empty rather than UNSET 592dec0 update constance machine setup 1e21ecc update to config_machines.xml for bluewaters upgrade; update testlist_allactive.xml 3c93918 Merge branch 'jedwards4b-pioduet' Add pio1 and pio2 subdirectories and change build to support both. 66c5321 changes so that PIO_REARRANGER setting works 2e9dfc8 Merge commit 'd95f3bed85738b7834262204b9ad19fb39f8fca7' into pioduet practice pull from subtree d95f3be Squashed 'externals/pio2/' changes from 0cf9924..dc374a1 dc374a1 latest changes from cesm dev e405d8e Add 'externals/pio1/' from commit 'f713b941947be742b2f7c159517e7810b24b6f95' 1d335c9 Add 'externals/pio2/' from commit '0cf9924548b5ac4b3c785985bef4e535af788610' 896f495 remove pio directory 8994b91 modify build to handle two pio libraries 4a4b5d5 Fix gnu mpi-serial compiler error and mpi-serial pio defaults ea2b9fe improve pio performance with netcdf 0cf9924 minor bug fix f60c4a7 separate serial and parallel in pio_darray 647e71f allocate requests in chunks dd39a87 Merge branch 'jedwards4b-piofixesforalpha05' Merge to trunk for testing 1953598 Merge branch 'cism_nag_port' of https://github.com/billsacks/cime into jedwards4b-piofixesforalpha05 Merge branches for testing d5758ad Revert "Remove no-longer-needed CISM_GNU CMake option" f547459 get the option set correctly ab851ad align cime with latest pio, including fixes for alpha05a testing 3de3528 Add pio_rearr_box to stub pio to get cime unit tests to pass 8cf46fe Merge pull request #355 from CESM-Development/revert-354-cime4.2.3_cntlexp a9f8f02 Revert "Cime4.2.3 cntlexp" 0617ff1 Merge pull request #354 from cecilehannay/cime4.2.3_cntlexp a6a748c fix typo in refcase name 27afe34 Merge remote-tracking branch 'cesm-master/master' 0871843 Updates for pio and archiving 466819f update archive xml to include mosart e207dcc Merge branch 'st_archive_update' of https://github.com/bertinia/cime into bertinia-st_archive_update merge in archiver changes 9669053 Merge branch 'jedwards4b-open_mpi_pio_workaround' merge in pio changes e2dd525 openmpi workaround pio defaults fix dd02a91 correct a bug in settng pio defaults 6daa145 rework workaround b2e0025 works around MPI_DATATYPE_NULL issue in openmpi 4f99653 branch for winter workshop control experiment 872e46b Change intent(out) to intent(inout) for a pointer 8078281 update st_archive, supporting files and XML schemas 9410f8a Merge pull request #49 from Katetc/master 3878373 A couple more tweaks to get output just right. 69d9ee3 remove unneeded include e232d1c pio netcdf serial file sync issue, cimeteststatus, G compset coupling b6f2dbd pio/netcdf serial issue / unused xml cleanup / cimeteststatus dcd02a0 Merge branch 'ncpl' of https://github.com/dabail10/cime into cime_merge Merge for precommit testing. f85ec1b Merge branch 'piosyncfix' of https://github.com/jedwards4b/cime into cime_merge merge branches for testing. 3146313 Remove no-longer-needed CISM_GNU CMake option c59b447 remove unused xml fix timer issue in coupler 5aa4bd1 Changes to performance test output to make it a bit easier to parse for plots and analysis. 5cdd5cc Reverting runcdash-nersc-cray to the main repo form. Debugging paused. e5c73d3 Merge pull request #3 from NCAR/master 7622c83 fix a sync issue with serial netcdf io eb441c4 Merge branch 'master' of https://github.com/NCAR/ParallelIO Fix the sync operation so that the iobuf is freed for all libraries not just pnetcdf 435868e need to flush for all not just pnetcdf db53954 Changes required to get examples to build on Yellowstone. fce4639 Cori queue/walltime change, PFS test fix ee28d29 Merge pull request #348 from jedwards4b/edison_port a72f97d Merge remote-tracking branch 'cesm-master/master' c3a9c74 port to edison slurm b61460a Merge pull request #48 from NCAR/ejh_12 07e9e10 fixed memory leak afc2cb3 Changes to test nightly building on cori 3c0859b Merge pull request #2 from NCAR/master 93b5f1a Merge branch 'master' into ejh_12 826c7c4 added emacs temp files 9d709f3 correct path for ctest on cori f41aa0e try salloc 5304311 another cori test change 9e4d6e9 update for slurm ce46244 another nersc test update 00b4802 updating ctest on cori 01173f3 update nersc ctest 685bf50 update nersc ctest a44eaa0 add code to support serial version of the test 2c47019 update nersc batch to slurm 54b28ff updates for nersc 2616613 mpimod test should be tried even if include path is not found d11eeed update pgi compiler version on yellowstone 98a5543 update gnu compiler version on yellowstone 7e1a644 more work on netCDF-4 functions 8eef91d remove c99 style from tests, updates from cesm testing ff0b899 more work on netCDF-4 code and tests e009bec more work on netCDF-4 functions ed20df5 adding and testing netCDF-4 functions 3093bab adding and testing netCDF-4 functions 6c6c8b3 more testing for netCDF-4 functions d11b265 Merge branch 'master' of https://github.com/PARALLELIO/ParallelIO merge to trunk for testing 666f4d8 Merge pull request #47 from NCAR/ejh_11 2ec9a52 fix issue with string padding in fortran c0d1220 fixed c99 issues a09fe44 removed by version of nc_inq_var_fil() 3feb775 merged master 4244cb5 adding memcheck test 85b9447 Merge pull request #45 from jedwards4b/docupdate 29e1e17 fix some documentation in pio_nc.c 14416c5 updates documentation in pio_nc 6ce1ff2 remove debug print statements 8a36ae5 unlimdim in dof seems to be working now 0e23b35 fixed test code 6e8b33d fixing problems with test_nc4 on non-netcdf-4 builds b40221e removed MPI error handler setting which caused errors on some platforms ce6a5bc fix issue in inq_unlimdim 43f5043 adding support for unlimdimindof 17634e9 starting to isolate some memory leak issues f8daed1 Merge pull request #44 from Katetc/master 7f852c3 Updated MPI Serial to stub in a call to MPI_Errhandler_set. Rebuilt library and pointed nightly mpiserial tests on yellowstone to the new MPI-serial library. 01afba2 feedbacks from cesm 5f583c1 Merge pull request #43 from Katetc/master 9bfcf05 Removed for loop variable declarations so we don't have to muck with CMakeList compiler options. 6b6f179 Merge pull request #1 from NCAR/master b79bb73 added test for inq_var_deflate 43178dd more nc4 function work dae0d82 starting to add other netcdf-4 functions 80b7ee8 Merge pull request #42 from NCAR/ejh_10 2804de3 added tests for def_var_chunking/inq_var_chunking 26a5736 Merge remote-tracking branch 'cesm-master/master' 983a450 Work on CLMGSWP3, and remove drv_flds_in in preview_namelist 78c96aa adding test of nc4 chunking functions dc46914 Revert "adding unit tests for def/inq chunking function" 318a1c6 adding unit tests for def/inq chunking function 6599e88 added inq function 3d12ca7 added function to define chunksizes b562100 now using chunksizes for netCDF-4 files 0393927 Merge pull request #41 from NCAR/ejh_8 3a69130 build file fix 4fc5c97 more documentation changes 0d7ea52 doc cleanups, more work on example2 70313c0 added use of mpe logging to example2 6903e74 more work on example2 e494c22 Change the default coupling interval for G compsets. c44afe1 more development of example2 0f10186 fixing MPI leaks d13a119 further development of example2 7aaee0a further development of example2 6b3edf9 working on example2 f8513df Merge pull request #40 from NCAR/ejh_7 3eb3252 example now uses all four output flavors, and plain old netcdf to check the result files 89107ea fixed some doxygen warnings ba6976c continued development of example1.c 2df0d8d clean up repo 73dd592 Merge branch 'master' of https://github.com/PARALLELIO/ParallelIO pull then push f1ace47 update version 2697c9c Merge pull request #39 from jedwards4b/fillval_test_fix 54a124d precount outstanding requests so that request and status are sized properly 431ea2b Merge remote-tracking branch 'cesm-master/master' 6d28fa9 Merge pull request #38 from NCAR/ejh_doc3 0071d69 more discussion of examples in documentation 1bd9b3b testing memory issues f902f0b fix for leaking MPI_Groups 7d5ebf8 fixing memory leaks and uninitialized value identified by valgrind 6e865b0 added config.h ecef0f8 Revert "added config.h file, added PIO_USE_MALLOC to it, added resulting config.h to c files that needed it." b770fc9 added config.h file, added PIO_USE_MALLOC to it, added resulting config.h to c files that needed it. 8f4361b added example1, improved docs, finalized timing library to please valgrind. 41f4570 adding beginning of example1.c d202032 Merge branch 'master' of https://github.com/CESM-Development/cime 15ed685 updates from cesm testing 5325fc5 adding to documentation 4e82091 adding documentation 2f1a6f8 Merge pull request #37 from NCAR/ejh_doc2 df59de4 fixed version number 80aa58e documentation fixes 901632d Merge pull request #36 from jedwards4b/put_requests_fix 85eee34 fix an issue with put requests bb0094c Merge pull request #35 from Katetc/master 7de57b0 Removed duplicate definition of loop counter (i) that gave the nag compiler fits. a82aed3 Merge remote-tracking branch 'cesm-master/master' 93b26cb improving documentation, C example now checks for correct number of processors 6810404 Merge pull request #34 from NCAR/ejh_doc 809f3bf updated documentation 5c4bc27 removed unneeded includes f946bdb added verbose option, also now check data after read 3df8a6e fixed bug in compdof array, added more documentation 08e1a27 further work on documentation fa69758 added documentation to C example code c88521b Add in GSWP3 for CLM 41d8486 Merge pull request #33 from NCAR/ejh_examples 3c24efe added examples to documented build targets 417a2ae fixed bug 9335536 turned on example build, initialized GPTL in C example 4d72cad updated build instructions a bit 918706d clean up and add a little documentation f397173 add gptl init to tests, fix issues with last commit c34292a Merge pull request #32 from jedwards4b/request_redo 16cfbea rework request list as allocatable to reduce mem usage 1ece5b2 Merge pull request #31 from jedwards4b/master 092216a increase PIO_MAX_REQUESTS and add a test to make sure it isnt exceeded 4e0e98c Merge pull request #10 from NCAR/master f3d0f74 Merge pull request #30 from NCAR/ejh_adding_build_docs 3af2ef4 updated modified date 088f0fc added build instructions for Linux, from source code e9e999b update pnetcdf library and compiler versions on yellowstone 680701f Merge pull request #28 from jedwards4b/cmakegptl_fix 0334df1 change HINTS to PATHS in find_path b9a2357 Merge pull request #27 from jedwards4b/cmakegptl_fix 7936fa7 need to mv check for gptl library 68d0e8e add a little documentation 6f363e4 Merge pull request #26 from Katetc/master 3dd101a Added ifdefs around the MPIAlltoall call in the case of OPENMPI which doesn't seem to like one of the MPI types used. This allows all of the nightly tests on Hobart to complete. 15a2f4e Merge pull request #25 from Katetc/master 9691ee2 Changed runcdash back to correct URL for nightly tests. 77ab8bc Fixes to get the nightly tests to run through on Caldera with serial MPI. Better handling of bad task configurations at init time, plus added logic to the test that prevents asking for bad configurations when running serially. d683d34 Changing checkout url so I can test problems with nightlies on Caldera 7471a1f Updated comments to support doxygen. 077ef28 Merge pull request #9 from NCAR/master 48ade07 Updated URL in README.md to point to correct location under the new NCAR domain. a744be9 Merge pull request #24 from NCAR/katesdocbranch 5acd796 Updates to doxygen comments to fix the API/User Interface documentation on the website. d71a484 Merge pull request #8 from NCAR/master 6506d65 Merge pull request #23 from Katetc/master 628dd8c Added code to attempt to open files with regular NetCDF if Parallel-NetCDF or NetCDF4 fails. d81c899 Added documentation hooks to help the doxygen api page pick up on darray functions. d3cbe86 forgot a file in previous commit b62188b Merge pull request #21 from jedwards4b/errstrfix 5c7d994 make errstr in pio_nc.c allocatable and minimal 935d2cd Merge pull request #20 from Katetc/master 8996a50 Going to test out nightlies with the main repo. 9a7c3c3 Hobart nightly debugging. still. 9371ca0 Merge branch 'master' of https://github.com/Katetc/ParallelIO 46b95ea Testing a git issue for nightlies on Hobart d423a7a Merge pull request #7 from PARALLELIO/master 9fce2d1 Merge pull request #19 from Katetc/master e8b29cf Rebuilt pio_nc.c with pnetcdf 1.6.0 which is what is used on most machines now. 1ca6def split 6-hourly SOFS streams into multiple files f742ae1 split daily SOFS streams into multiple files 3c76036 split NCEP streams into multiple files e314c1f one last gitignore update b04f58a didn't mean to commit .gitignore at cime/ level 4f518aa update .gitignore in mapping tools 58b5fb1 Updating ncparser.pl and pio_nc.c to support a newer version of the pnetcdf libraries. 46ca777 Added the full path to qsub and qstat on Hobart and removed the extra PIO_DASHBOARD_SITE export call in runctest-cgd.sh 3fae34e Updating Edison documentation to reflect changes in the cce module. 17ed7e8 Merge pull request #18 from Katetc/master d51e5fb NERSC removed the cce/8.4.0.219 module we were using and replaced it with cce/8.4.0.223. I updated runcdash-nersc-cray.sh so our nightly tests will build and run again. 9044637 Merge pull request #17 from Katetc/master f385846 Finished debugging! Cdash should work for Hobart now. Changed paths back to original locations for the git repo and Jim's dashboard directory. d9455ff lowercase cgd 4abbae7 Still debugging afe0cfe I think I'm almost done debugging. 04a69f1 More debugging 513aaa1 more debugging 6a66167 More debugging 879ff76 More debugging 40ed8af More debugging 2fd1f2e More debugging Hobart 0e11285 Reverting to regular CTestScript so I can merge to the main branch. 132f4a7 More debugging c64cd46 More debugging 9e38451 more debugging e3d42f5 more debugging 86d5cb0 More hobart debugging. 283885b more debugging fff902e More Hobart ctest debugging. d4b60bd More hobart ctest debugging. 26aca80 More hobart ctest debugging. 4ec1fc3 Debugging changes to Hobart tests. 4f92e91 Merge branch 'master' of https://github.com/Katetc/ParallelIO ffe2c30 Testing changes for overnight tests on Hobart. 8660f6d Merge pull request #6 from PARALLELIO/master a2354e4 Generating pio_nc.c with a newer parallel-netcdf library broke the build. I am checking in the previous version of the file temporarily until we get a more permanent fix in place. 2c53a65 Merge pull request #5 from PARALLELIO/master f630c6e Updates to ncparser.pl to add a more specific link to the NetCDF documentation for the doxygen documentation of each of the wrapped functions in pio_nc.c c7acf2c Updated documentation for building with PGI on Blue Waters. 6cfc67e Merge pull request #16 from PARALLELIO/templatingTarget 75fc329 Simplifying (doesn't appear to work, yet) 2ee7196 No SOURCES tag in add_custom_command 4f61f3f Changing to add_custom_command to fix dependencies 64d807b Attempting to change 'execute_process' to 'add_custom_target' d922581 More things to work on testing on hobart. 81e85ec Trying stuff for the hobart nightly tests 1ae09c7 Pull source from my branch for testing for now. 72d0e5f Testing out changes to get the hobart cdash tests running. 14364ed Merge pull request #4 from PARALLELIO/master 58c6211 Removed -x from the script. 7f13253 Added a runctest file for Hobart. Hopefully this helps the nightly tests run through... though needs testing. cf12369 Bug fix for compare_namelist.pl 53f5ad7 Merge branch 'master' of https://github.com/PARALLELIO/ParallelIO 207d705 Changes to cmake and ctest to run tests on Blue Waters. Updates to machine walkthrough documentation for Blue Waters (updated netcdf modules required to run ctest). 20ebb23 Added a link to the cdash site for nightly tests 7e59004 Added build instructions for Intel compiler on Blue Waters. a8cc829 iframe is really, truely no longer ALLOWED by github. Will have to find a different way? 69ac68d Trying again to add an iframe to embed the nightly test results on the GitHub page. 855ff8f Trying to add an iframe to embed the nightly test results on the GitHub page. 6a63311 Added a better handler for NAG compiler -mismatch_all 27214fc update to use $ENV ddbbe33 add hobart to supported test systems 171aa70 add ctest scripts for hobart 8c5b4ea Merge pull request #15 from Katetc/master 9365829 Cleaning up files to get ready to push bug fix changes. 5d8aa2d Debugging and fixing problem with more indexes than number of elements in the array. One test case now works, but the original one still explodes. e3e770e Merge pull request #13 from Katetc/master 3a04958 Improved output and added DEBUG flags 0dff92d Added and improved some doxygen documentation for fortran support functions. 9a0870d Patch to allow RESUBMIT to work af95400 Update C1DIAF streams 6f49384 Merge pull request #9 from Katetc/master 09c3132 Added link to new github.io website and doxygen documentation to main page. b7071c8 A few last minute changes to mach-walkthrough bd28794 Added Install walk-throughs for several super computers. 719a448 needed to build without timing e016b30 Deleting commented-out code 777bbba Setting return variables in parent scope (instead of caching) c4aaf21 Reverting last change (uncommenting PKGCOMP_FOUND check) cf053f9 Removing check on PKGCOMP_FOUND 985f6f2 Changing NAG checks to STREQUAL (instead of MATCHES) just to be sure 4207106 Making compile options private 78eae8c Adding new function to initialize search paths from include/lib lists 5e4f0f4 Merge pull request #3 from PARALLELIO/master 347892e Formatting only. 3cd06f3 Allowing for all-caps MPI.mod when searching for Fortran module b965a21 Bumping back up to cce version 8.4.0.219 to fix compile errors 0766adb Moving back to default mpich and bumping down cce version to 8.3.12 06df635 Bumping up version of mpich 3818db0 Updating cray compiler to most recent version 42d570e Allowing mpif.h for Fortran component of mpi-serial 135bd1f MPISERIAL enabling variable misnamed 31c2f3e Adding dashboard runscript for MPISERIAL build edf8130 Removing unnecessary GNU run script e4541b5 Typo fix. Need prefix path, not lib path. 94dcf85 Locking software versions and cleaning environment on ALCF 1fbe81d Manually setting environment variables for ALCF IBM compiler 643f228 Adding .project (Eclipse project file) to ignores f5f047b Merge pull request #8 from Katetc/master 1e21cfd Changes to CMakeLists for NAG compiler and shortening a line in pioperformance.F90 for NAG. 05160cd Merge pull request #7 from Katetc/master ae84e0f Shortened two lines to make the Nag compiler happy. 880cbff Merge branch 'master' of https://github.com/katetc/ParallelIO 695877e Merge remote-tracking branch 'piomain/master' 8efad12 Merge remote-tracking branch 'ktfork/master' 06f0d74 Fix for MPI_Fortran includes in checks 133d24a Fixing bug when getting gnu compiler version 45394bc Disabling add_mpi_test when using mpi-serial 396cdcb Removing debugging statements. ccda7e6 Debugging statements. 791b5f7 Bugfix. Need endforeach (not endif). 3723d51 Adding MPISERIAL find package d6c842f Adding new find package capabilities to support MPISERIAL 8ed1096 Changing comments when disabling PnetCDF b7d7a3b Updating Introduction and Install documentation. 5400fdb Wrapping CheckMPIreturn method with NO_MPIMOD checks. 81f6409 Removing MPI_OFFSET definition from pio_kinds (just don't use module) 8d2bea2 Setting system and compiler directives. 6254b53 Adding includes/libraries for MPI only when using mpi-serial e503ad3 Defining MPI_OFFSET for modules that do not define it 0ba152d Making NO_COMM_F2C private 8062688 Better checks on MPI to test/enable builds with mpi-serial 9a9e66d Merge branch 'master' of https://github.com/PARALLELIO/ParallelIO.git c889bec put pnetcdf specific code in ifdef block e9b6263 Move checkfunctionexists include to top-level e5a35d3 Adding checks for PAPI, gettimeofday and MPI_Comm_f2c ea720dc Defining preprocessor directives for system/compiler names 537d11f Adding a FindPAPI module for the internal GPTL 6a28ac5 .gitignore html/ folder (contains gh-pages branch) 80cdddc Merge branch 'master' of https://github.com/katetc/ParallelIO ef790f0 Merge branch 'master' of github.com:PARALLELIO/ParallelIO updating local 385eeb2 remove debug print statement f713b94 add support for optional old error handle 8e597f5 Adding initial ALCF IBM-compiler dashboard run script b134aea Adding specific environment setting for ALCF f8f3a90 Automatically getting compiler version on Edison. 87e06fb Adding build group/model (Nightly, Experimental, etc) to build dir name d2f97b3 Getting compiler version automatically on Yellowstone d942dba Simplify the dashboard run process. Guarantee build/tests jobs same. d8f87dc Adding CDash run script for Cray compiler at NERSC 00f7e48 Adding "nid*" hostnames to NERSC identifiers 0ded31f Adding GNU CDash run script for NERSC 43285b5 Adding more basic NERSC modules 1ebdef9 Need to load MPI 6dc107a Adding initial CDash run script for NERSC 22d93b2 Fix for NO_MPI_MOD --> NO_MPIMOD 0a261a9 Properly read environment variables e613366 Setting environment for NERSC runs. d4749d4 Adding NERSC environment and ctest run scripts. 2329452 Passing compiler ID to parallel job environment 22b86f0 Reading Dashboard model name from command line cd5326b Adding CDash test scripts for NWSC/Yellowstone 244187c Adding pronghorn to NWSC name designations e681f38 Removing test-run output redirect 78783a6 Re-enabling submission. d299ffc Removing debug output. 6626b34 More debugging output. d70e2d9 More debugging output. 5722156 Minor change to compiler info in build name. 7bfcb98 Disabling submission. Adding debug output. b20cc7f Bugfix. Wrong prefix variable for options. 867b383 Don't need PREFER_STATIC on Yellowstone. Removing. 1d742ca Adding PATHS options to the find_package func (searched after hints) 79812dd Adding debug statements 9080fb1 Removing debug lines c96f2b8 Adding debugging output. 463e22e Removing debug output. 9164459 Debugging output. Making execute_process command a string. 96fccd9 Testing PREFER_STATIC with Yellowstone GNU 71d9899 Allowing compiler ID to be sent to CTest via env variable 67e4061 Executing tests from binary directory d94bbe0 Bugfix. Using "--envs" with runjob, not "--env" fb20503 Setting dashboard root environment variable on ALCF ffc90e8 Set PREFER_STATIC to TRUE 246fae5 Trying PREFER_STATIC instead 15eeadf Explicitly setting HDF5_PATH for ALCF 91f0029 Preferring static on ALCF 47569d5 Removing debug statement 895897a Adding debug statement 918cdce Nope. Prepending doesn't change behavior. Going back to append. 20bc020 Need to prepend to module path to over-ride existing Find* modules 18aa3a2 Quick datm update for 01col IAF a615399 Bugfix. Need to pipe extra commands to runctest script. 1d83ce0 Adding ALCF test scripts 75c70ea Need to APPEND to tag, after all 63efe69 Simplifying binary dir location to make portable 5c374d0 Using improved site-name a7a0d7d Improved site-name format 4540983 Add C1DIAF compset / update datm streams f55b07c Updating NWSC dashboard scripts 3e12c33 Updating NWSC run script to match new model e518ab8 Removing limit on number of tests 10232cf Bugfix to prevent 2 submissions when attempting only 1. b62fba8 Bugfix. Changed test script name. 16b8579 Creating a common header-script for inclusion in main and test scripts 7ece6f6 Normal ctest_test command in Appendix script 16d3d0d Need to run appendix script from source directory 1c8a7b4 Try to append tests to same dashboard tag d1e7368 For testing, just enable tests 5-8 c0f61d5 Try to just run ctest 4e19c06 Try to use the CTestScript for testing only... e4d511d Renaming, again, and requiring PIO_DASHBOARD_ROOT variable for control ae92705 Assuming source exists and all ENV variables are set before ctest e5e0e53 Adding scripts to run on Yellowstone 4b821ef Bugfix. Can't use Bourne-shell commands in CMake! 7d3e91b Enabling dashboard submission 0008917 Adding script-driven CTest running 775a045 Setting up configure/build system for nightly tests. a9e4446 Adding environment setting script for default machines 990a8e0 Renaming and turning off submission for testing purposes. ed6034f Testing fb0b65d Starting to general the CTest dashboard script 29ed55c Re-enabling git update step, now that bugfixes are in place 88c79eb Using add_definitions to set the directory-level compile options e3e2431 If GNU Fortran, setting tests to free-format line length 3c742c2 BUGFIX: Incorrect check for Fortran_CSIZEOF f57b4fb BUGFIX: Incorrect global find/replace fixed! fc1f5a8 Merge pull request #1 from PARALLELIO/master 2a22680 Adding first attempt at CTest Dashboard scripting 75d1f81 Adding installation instructions (and other info) to README 0f5f3af Adding the ability to install PIO in CMAKE_INSTALL_PREFIX location 7f65ed9 Merge pull request #4 from PARALLELIO/gptlExternal e4af78c More updates to the README f4619ab Adding options for backward compatibility (see below) e496345 Fixing more references to *netCDF_DIR (should be *netCDF_PATH) 814e82c Adding messages about GPTL during build 4caf864 Properly handling QUIET and REQUIRED arguments to find_package e2d40b3 Merge remote-tracking branch 'origin/master' into gptlExternal d4a8f82 Adding link to external/internal GPTL if set and found 9612eb1 Changing the name of the component to Fortran_Perf e1ab44b Removing unnecessary CPP directive f58ee83 Adding a component for perf_mod.mod e60327c Switch back to gptl.mod. The perf_mod has nothing to do with gptl. a5423dc Switch back to perf_mod.mod include file. 55d9750 Reverting to previous revision. Unnecessary. 545c9eb Updating the Fortran interface to external/internal GPTL versions 4a0bf6d Adding find_package check for GPTL and setting CPP directives cc0eb76 Adding find_package check for GPTL dependency 063c81c Adding a new Find module for GPTL. d6c78e9 typo in LND_DOMAIN_FILE for 01col 21af6df Update datm defaults eb23669 Reverting needless variable name change (back to USER_CMAKE_MODULE_PATH) 3cb6107 Create first pass at README fd09d4f A few tweaks needed to run 01col grid 4df82aa Adding component-specific environment variables to find_package searches 80afe02 Merge pull request #3 from kmpaul/simpleCMake 6106433 Formatting only. c53a7fc Adding INQ_VAR_FILL and DEF_VAR_FILL enum values ec09bf0 Renaming template-generation flag to PIO_GENERATE_SOURCES_FROM_TEMPLATES a8533fc A little execute_process cleanup 3fd011b Adding template source generation to pioc CMakeLists 3a756a1 Moving template source back into src. Just makes more sense 6999389 Adding environment variable definitions to ALCF mpiexec command d86bf72 Moving nc*parser.pl scripts into templates directory, too 283fb30 Moving templates into their own directory f0b72eb Moving santos.c into _UNUSED_ until I know better where it goes 7a48e36 Moving prune_decomps.pl into top-level scripts directory 25edc19 Adding "tests" and separate "check" targets to control testing 4c282cd Doxygen cleanup. 69b6662 A little doxygen cleanup. db451fb Moving output up one directory 47d84ec Adding "check" target to emulate autotools "make check" 2e42142 Remove unnecessary file 26dc48f Bumping up CMake minimum version to 2.8.12 and adding doc sourcedir 4bb8cee Excluding all tests from ALL 03ef90d Moving testpio example into examples directory f0f82bd Moving all doc source into doc directory (and restructuring) bf5a839 Finished changing _DIR variables into _PATH variables 94b6f10 Removing version files since try_run cannot be used b0d153a Fixed check_version function to NOT try_run 2ce5199 Setting compiler options in pioperf specifically. 15db8dd Making compile options private. May need to add to tests individually. 0fcdddf Moving back to 2.8.12 syntax 8a09383 Update factorfn default for 01col grid 8eca84c Including external macro not so easy... 1afa8fe Adding option to include an external macro script (e.g., from cime) 52dfee1 Fixed naming problems 05aabee Changing name of unit test, too 35a024a Changing name of performance test target 4d8bd8b Removing more debugging statements 7976cc6 Removing debugging statements 7983240 BUGFIX: Forgot to update add_mpi_test syntax in unit tests 5687bfc Debugging statements 6b0c666 On LSF, need to move number of tasks to end of command 721cf09 BUGFIX: Need to specify "unknown" for other platforms 0fde4cd Changing add_mpi_test to take separate ARGUMENTS and use platform execs 6cb90ad Use single column datm forcing for 1D_1D res 19a4b8a BUGFIX: Typo (num_proc -> num_procs) e2e35d4 Adding basic CTestConfig.cmake file for CDash submissions f1f419f BUGFIX: Need elseif statements, not if 401ff9d Creating MPI launch commands assuming running from interactive session 566998f Much faster to run CTest from within an interactive session 1bba202 Simplifying UCAR/NCAR test runs. f67f089 Remove check_file function (not needed if using NetCDF 4.3.3+) 756b273 Changing _DIR's to _PATH's for convenience fbd3209 Fixed platform name function 5328d3b Adding check for PREFER_STATIC (in addition to PREFER_SHARED) a8785dd Adding check for file (instead of a direct CPP macro check) e227f72 Fix for MPIEXEC on all other platforms a6da680 Adding platform determining function e054c4e Removing SOURCES keyword for CMake 2.8 syntax 4e19a39 Fix for UCAR MPIEXEC command 8cd74ec Fixed transition to CMake 2.8. 4e3d036 Update to support 01col grid 3bf8ce7 Moving to CMake 2.8 syntax. 952d48a Switching to CMake version 2.8.11- syntax 7577418 Reverting changes and bumbing CMake required to 2.8.12 3db0f0a Test out lower version of CMake d4d9732 Bumping down to 2.8.11 7bded5a Bumping up required version for target_ commands 3dcdbff Missed GPTL 76031cf Change version in all libs c0b5e41 Lowering CMake required version further bd3fb50 Try reducing required CMake version 0ef8c3d BUGFIX: Need scoping for try variables 8eeb37e Adding HDF5 check for optional SZIP 1b89a53 Renaming check_package to check_property...makes more sense adb4c68 Cleanup. Moving mpi-related functions into LibMPI be839f0 Simplifying check_* calls with LibCheck library functions 7e280ac Renaming LibFind for helper functions for Find* macros ec83bfc Fix case in TryMPIMod.f90 file name 8a07afa Fixing bug with MPIEXEC not found abb0e02 Enabling CTests (except for performance test) d3ad6c9 First pass at enabling tests 5963c2b Adding MPI Fortran module check 532de83 Using CMake_Fortran_utils for testing purposes 50b5029 Removing required specification for NetCDF C f0d55a4 Enabling performance test build aec7cd4 Adding Fortran compiler checks and made compiler options public 33a5f9e Making NetCDF_C required e12ff5e Adding big endian test d662558 Renaming directory (to be consistent) 1594879 Adding unit test directory to build a8c043e Formatting only 5c95933 Adding automated GenF90 download, if GENF90_DIR not set bc60418 Checking cache variables for "if defined" instead of just "if" 8f7c7c7 Fixing version checking (one version per package) 3be528e Including check on shared libraries 56ac227 Including PnetCDF fortran libraries 80dccaa Attempting to find appends 3b2a736 Appending PnetCDF to NetCDF_Fortran, too, if needed 8237d5b Bugfix ae373d3 Adding PnetCDF to NetCDF if required ffec202 Removing include for CheckHDF5 module 392a45a No need for CheckHDF5 module e9158fe Moving package checks to separate modules 19562b5 Debugging output 891187d Bugfix: extra endif removed b78e05f Looking for all components before doing dependencies and checks 531ca7e Adding all include dirs for Fortran version checking via tryCompile a8f76a9 Updating version requirements 374c1d5 Adding PnetCDF version checking f2bb41b Checking NetCDF Fortran version 1a0b326 Renaming variables to ensure scoping is safe e3e3b43 Removed debug checks. NetCDF_Fortran version check works on mac. 3b76f49 Adding PNETCDF support check and debugging output f5d386c Output in case version could not be found. e8e60c3 Get NetCDF Fortran version string a4c1ccc Removing debug statements e564fe2 Adding NetCDF version checking analysis 741ee52 Adding version checking for NetCDF 9c9bc76 Fix for is_shared_library function b8cef51 No up-front declaration of prefer shared or static variables 0959575 Use default find_library first, then search for shared if necessary 56dfdd8 Turning PREFER_SHARED on by default 98a1865 For testing purposes, adding tests back to all b5b94d8 Removing old CMakeLists file c3da0be Fixing MPI includes/libraries for HDF5 HL and Fortran_HL searches b924600 Removing unnecessary inputs to find_package_component function ccb16d3 Modifying find_package_component to assume variables pre-defined 41bb342 Fixed other foreach loop bugs. ffa9e27 Removing debug statements. Fixed foreach loop bug. 7cc4151 debug output b4b1bc3 Debug output ad2b73f Debugging output. 9c0a760 Updating find modules to use new helper functions 27dc859 Adding define_package_component and find_valid_components functions 6a8489b Adding find_valid_components function. 993bc8c Moving includes (removing redundant includes) 022b6c1 Removed unnecessary finds after already found 51d58b1 Manually adding libdl to NetCDF libraries when static 4dc561b Fixes to try_compile source files to check macro values are 1 91e5882 Debugging output 584f7c7 Debugging output fea5a12 Debugging output cbf6b17 Debugging output 59257c2 Adding tests for DAP support and Parallel support da67f4d Adding modules for HDF5 and LIBZ by hand 25eea60 Changed NetCDF Fortran dependency d9a371e Changing dependency of NetCDF Fortran to HDF5 Fortran e2ebf0e Disable static libraries c134cec Cleaner include/library hints ec5515a Appending EXTRAs to the end of each target 8ef5ffe Using static libraries in HDF5 if set globally 3b31b01 Fix GNU compiler options ce3c78d Correcting target and variable names in piof f5193e2 Adding the ability to pass in extra includes, libraries, etc 8dc6d6f Removing CURL dependency. Only needed with DAP support. ac40e1a Use static libraries variable 0e4dff4 Remove scoping from used shared libraries variable 7c9f491 Renaming variables to provide PIO scope 7051cc6 Remove blank line 25b36d8 Removing REQUIRED from NetCDF-C dependencies 1906a44 Fixes for find_package_component macro 4a78f3d Adding new macro function to search for includes/libraries 0349b93 Starting to move common find functionality to library macro 582b108 Typos. 3a67912 Checking MPI wrapper for Fortran includes/libraries 7f45e43 BUGFIX: Removed extra endif 7ab8e4b First attempt to check MPI includes/libraries from wrapper first 45d871a BUGFIX: Typo in name of PnetCDF Fortran library d752f73 Debugging statements. 609cb39 Using component-based find modules d39c6db Forgot to include component-based find module for PnetCDF 2af042d Component-based versions of find modules in place bc6102c Adding component-based NetCDF find module e2af06a Including explicit checks for shared libraries first. f47818e Removing PnetCDF explicit dependency on MPI 3eb6a54 Removing commented out code. f211199 Re-publicizing NetCDF and PnetCDF link libraries 7560b46 Removing debug statement a025a65 Debugging output 5580c3e More code cleanup 3fd1068 Removing commented out code. 2f37d25 More debugging. Re-enabling "plural" include/link variables 13ab96c Debugging: Using single include/library from NetCDF and PnetCDF 0d50a19 Debugging: Commenting out unnecessary link to GPTL in tests 4981bf9 Debugging. Commented out explicit MPI link and includes b53c086 For testing purposes, making linked libraries private 69022b5 Fixed target libraries as PUBLIC 0f26432 Adding additional support to detect static libraries cdad8e0 Look for shared variable and do not append options for dependencies 78b2d00 BUGFIX: Corrected library name 1597ba9 Changing PnetCDF module name to include language 8575a34 Adding SHARED variables to netCDF modules 4d614d3 Comments only 0f47be3 Splitting the PnetCDF module into separate languages. 772617a Re-enabling builds of tests 373a2dd Pulling out unnecessary compiler flags 92e5602 Moving VARN check into target specs, out of PnetCDF module 3bc5719 Removing separate PnetCDF_<lang> modules 25aaf33 Another fix to make only 1 library variable 1d3bf55 BUGFIX: Should only be 1 library 099ae58 Using common PnetCDF module 24fae32 Removing debug statements, adding common PnetCDF module f5942c8 Removing debug statements c05bf3d BUGFIX: Hack for bug in CMake v3.0; manually set normal-case found var 88ca52f More debug statements. 2560491 Look for shared as well as static in LIBM d0dcd71 Debugging statements 469aee1 Debugging statements 770ccd6 Remove unnecessary -D from compiler definitions adb82f1 Debugging statements. c8e3175 Forgot to include CheckFunctionExists 724e0f5 Updating new find_package modules cc3ddd5 Updating to use new find_package modules 30827bc No need to remove duplicates. ddd80d0 Removing duplicates in lists. Fixing check on environment variables. 1d9fbf5 Adding C library find_package modules 25f544d First test of new find_package modules 8f06cac Disabling tests for work with new find_package modules 7b00877 Adding most-recent LibFindMacros-based find_package modules 80804a5 Formatting only. 5c440b6 Enabling tests 71b560d "Privatizing" options and definitions (not include directories) 46c84d7 Whoops! Actually do need the pioc library for Fortran tests 62f02ec Adding binary directory to public include directories for piof 7c22a12 Removing pioc dependency in Fortran tests 0615d19 Enabling Fortran language in tests 2b9dd0b Adding first tests to CMake e224000 Moving (apparently) unused source out of the way 63c823e BUGFIX: Removed PUBLIC keyword from target_link_libraries 168ec68 BUGFIX: PUBLIC keyword doesn't belong in target_link_libraries 29627cd NetCDF and PnetCDF dependencies use environment variables f22942e Adding comments 0bd858d Adding gitiginores 1de1fda New CMake build structure. f8ff531 Restructuring the CMake build, using more subdirectories. 9afbdda Some debugging output 21c91e6 Changing library links from names to absolute paths 169010c Adding current source dir to include path 6df09af Bugfix: Need PRIVATE/PUBLIC/INTERFACE descriptor 0bdf02f Attempting CMake 3 target-specific definitions 775c1ea Cleaning out old CMakeLists and starting very simple 141f6da Merge pull request #3 from PARALLELIO/master a337c92 add idealized block and roundrobin decomps 69222aa fixed issue with missing data in doml_mask at 900 tasks by reversing iobuf and missing data writes 11cfd54 add some documentation and improve timers. 76b6651 Merge pull request #1 from PARALLELIO/master e21308e another rework of the request handling mechanism 121cef1 adress an issue with writing partial not decomposed fields and saving requests abb2d76 Merge branch 'master' of github.com:PARALLELIO/ParallelIO update to latest 42a86f0 fix issue in flush_output_buffer 2fa9e88 Merge branch 'master' of github.com:PARALLELIO/ParallelIO merge in changes from cetus development e1e5b3c remove CXX check as not needed 2ad54b7 Merge branch 'master' of github.com:PARALLELIO/ParallelIO pull in changes from alcf development f9773af mv list alloc from malloc to bget 916918c rework of memory management for iput 9389cba fix rate estimate in pioperformance.F90 improve hole algorythm f7b491e fix logic in mpio_onesided clause dab77bb options for wait_all call 5c6ceb2 use PIO_REQ_NULL to maintain sync between io tasks e592aae Merge branch 'cetusdebug3' bring in mpio_onesided optimzation c68843a remove debug statements, force flush on record change 0561c91 add back MPIO_ONESIDED code, bug was found in pnetcdf c1cdccf no map of 0 length required 6a217f3 still debuggin 7dd4626 improved handling of _FillValue d20bfe2 acknowledge Worley contribution2 2c1fc72 acknowledge Worley contribution 143b545 add a check for fill values 083eaa9 Merge branch 'master' of github.com:PARALLELIO/ParallelIO into jedwards/perftune trying to get this on master and probably doing it wrong. fd8d3d0 more extensive testing df1a52b add PIO_CNBUFFER_LIMIT to supported environment variables, value should be a positive integer optional followed by M or K which multiple by 1000000 or 1000 respectivly 56742a7 add environment variable to control swapm settings, PIO_SWAPM expects a : seperated list of three values nreqs:handshake:isend where nreqs is an integer and handshake and isend are boolean values 9866ce2 update timers to conform to newly established standard for cesm update author name 9834318 conflict resolved a758d32 conflict resolved 53bb021 remove debug print statement bda85e1 remove crusty mpiserial code 4b0ceb5 reduce print statements 7cb87ff merge changes from master ef426e4 clean up 67c6971 remove mpiserial test, no longer needed with mpiserial update 81c9c21 Merge branch 'jedwards/perftune' of github.com:PARALLELIO/ParallelIO into jedwards/perftune b19e9c3 dont test nreqs=1 a10082a hand mrge a1e2322 remove cmake test for mpi-serial c7bee49 Merge branch 'master' of github.com:PARALLELIO/ParallelIO 95a8c91 remove MPISERIAL cpp flags, no longer needed d93c410 Merge branch 'jedwards/perftune' of github.com:PARALLELIO/ParallelIO into jedwards/perftune fix the conflict a68b9f4 Merge branch 'master' of github.com:PARALLELIO/ParallelIO into jedwards/perftune is this really what I wanted? 96040db redo perf check to avoid slow configurations 597d71a fix include path for gptl.h add headers for gptl and math correct call to io2comp 815aec3 add performance tuning c764740 fixed an issue with size of rindex in box rearranger code 1e24916 move tests to common subdirectory to prevent tests building in cesm ebf0519 Need to always write non-decomposed fields from io task 0 instead of rotating through tasks b6b400d look for NETCDF_C_DIR or NETCDF_DIR f3d6aef remove debug print statement b431d8e add a workaround so that BGQ machines will link correctly ae26ae8 remove some debug print statements 7a2d36d update nightly build script to use git repo 658bd07 no need to sync readonly files 3c9819a add bectl call and allow for direct use of malloc with PIO_USE_MALLOC cpp flag, fix double free issue 34ceebe Making sure that we print the failure output from rank 0 only when comparing float/double arrays. This check is already there for checking int arrays 4296efe undo last e252797 try removing purge on goldbach 50927cc try turning of experimental part f5ee4fb cleaned up code ac4b542 fix for no data issues bad9aa7 update for mira build 3d59ca7 initialize holegridsize ae07fba fix nightly build issues 4556501 port to bluewaters 4182177 Fix an issue with identifying NETCDF NETCDF4 7c1bea1 Fix an issue with identifying NETCDF b3a2fa1 try yellowstone again ; 03e5dd2 nominally working on cetus 216df9a onesided mpiio flags for mira a4a94e9 try new macros file 406daf9 mail results of cetus testing 2c79831 fine tuning cetus testing 6b8f14f tests are working on cetus now c85c9cf fix flush command for netcdf4p, netcdf4c dd7ad9f try again on cetus d80aa03 port to cce compiler bd384b6 update externals 6c61a1b more cmake work 0778e5b fix typo e937cde policy was a bad idea 6617cb3 add support for netcdf c and fortran libraries installed seperatly e010df3 update build flags for mira 0143589 Adding some fillvalue tests for reads. These are similar to the write tests added in r1485 and r1488 6194e3f Remove debug prints and re-introduce deletion of test output files - added in r1489 42aad2a fixed fillval test f31b63e Fixing bug in the fillvalue tests. Making sure that we read the complete data before checking the results. The previous versions used the same iodesc for read and write, so the explicit or implicit holes in the write iodesc (write compdof) were also present in the read iodesc (read compdof). Hence the reads were not reading the complete array e337fb3 make wbuf allocatable 4f7119b Adding a test for writing data with fillvalues where the data holes are implicitly specified (rather than filling the compdof with 0s for the invalid data values, the compdof for invalid values is not specified). The old test where data holes were explicitly specified (by filling corresponding compdofs with 0s) is now renamed appropriately 0ce01f7 fix issue in netcdf4p interface 9daa924 fix for openmpi: MPI_Datatype is not an integer and cannot be used in switch statements 518547a Adding a write test with fillvalues. The compdof is defined with holes (0s) and the user specifies the fillvalue (-2) explicitly during the write. f332eb6 work around for xlc bug? be44462 need to find maxregions across all iotasks 9930a99 merge in fillval branch 2e88734 fixes from mira c7a2818 try this on cetus 110bee2 not working yet, but close 68eb5cf fillvalue experiment branch 43f46c3 remove an unneeded use statement 9194276 a little cleanup ae40fdc remove restriction on filename string length acdb1b5 use 64bit data for pnetcdf 94568d2 nothin special d13161f netcdf4 works b38167e reduce number of flushes 4efd607 fix for fill in box abae181 bug fixes in mpi type declaration 536f2c8 fix an issue with netcdf4p open test 7b4997f back to mira da07fc1 fix membug issue b0ad555 work around nf90_def_var_fill missing 0b534dc need to increase search size for mpas a076839 try this 2872185 modify mpi_allreduce to work on ibm? 488dc6b still trying to handle memory issues 54dea03 try again 0c774fe revert nag compiler for goldbach build a443cce bug fixes found in cice5 port d21ca1c more memory management d7c4e56 more mem management fixes 2355c73 initialize compute buffer in init call d58c83c add memory managment 4a5b75a fix an out of bounds error e5e6f59 fix an out of bounds error a0abba9 updates to nightly build stuff 420a78f updates to nightly build stuff 14e41a0 updates to nightly build stuff 4b8d438 fix a cmake issue a210d76 improved handling of halo regions on read e67e5ae improved handling of halo regions on read cd25733 fix some cmake issues 40f7a9c more error information for get_att b730203 try this 43d4210 fix build issue with openmpi (basetype not integer) 7cfbc49 add fill value support for incomplete field write 62cad14 more bug fixes c443d51 need to set PIO_WRITE flag on files opened with create 250f4d9 fix failing unittest by adding check for file in read only mode 47af1c7 more debugging 4a7c6f4 fix unittests b5ad175 correct cpp da17b49 fix issues in unit tests e479dc0 replace mpi_request with int 1e6be53 MPI_Request is wrong - should be int 3fce73c more multi bug fixes 9b4db9f buffered writes d1f510c fix a character read issue bdb3de4 another multi …
Summary of Issue:
There appears to be a crashing bug in EDPatchdynamicsmod. The proximal issue is that it is crashing when fuse_2_patches is called. The underlying issue is that the same patch was trying to be deallocated twice.
Expected behavior and actual behavior:
I tried to run a global 4x5 run and it crashed after 74 years. Expected behavior is that it not crash. Actual behavior is that it does crash. Was able to repeat the crash, both with restarts and with a second identical case. The error log pointed to an issue in Patchdynamics:
end run_mct
forrtl: severe (174): SIGSEGV, segmentation fault occurred
Image PC Routine Line Source
cesm.exe 0000000000FE6511 Unknown Unknown Unknown
cesm.exe 0000000000FE49C7 Unknown Unknown Unknown
cesm.exe 0000000000F86A54 Unknown Unknown Unknown
cesm.exe 0000000000F86866 Unknown Unknown Unknown
cesm.exe 0000000000F1BBF6 Unknown Unknown Unknown
cesm.exe 0000000000F26F2D Unknown Unknown Unknown
libpthread.so.0 00002B1F65EC5710 Unknown Unknown Unknown
cesm.exe 00000000007D5127 edpatchdynamicsmo 1267 EDPatchDynamicsMod.F90
cesm.exe 00000000007CAC61 edmainmod_mp_ed_d 209 EDMainMod.F90
cesm.exe 0000000000504FEC clm_driver_mp_clm 811 clm_driver.F90
cesm.exe 00000000004F5862 lnd_comp_mct_mp_l 436 lnd_comp_mct.F90
cesm.exe 000000000042AA8D component_mod_mp_ 1044 component_mod.F90
cesm.exe 00000000004143AF cesm_comp_mod_mp_ 2415 cesm_comp_mod.F90
cesm.exe 000000000042838D MAIN__ 93 cesm_driver.F90
cesm.exe 000000000041208E Unknown Unknown Unknown
libc.so.6 00002B1F660F2D5D Unknown Unknown Unknown
cesm.exe 0000000000411F99 Unknown Unknown Unknown
Steps to reproduce the problem (should include create_newcase or create_test command along with any user_nl or xml changes):
standard 4x5 global ED run, run for at least 74 years.
What is the changeset ID of the code, and the machine you are using:
have you modified the code? If so, it must be committed and available for testing:
yes, https://github.com/ckoven/ed-clm/commit/59e6e6e90dee77cceb37e7ebc31be0fb6fa216bd
The original code in the fuse_2_patches had a warning in it from Stefan that that bit of code seemed fragile, so Ryan rewrote it. That code is now a bit more clear, in that it has fewer inline pointers to pointers, and better documented, but with the same functionality. So that didn't actually fix the problem. What did fix the problem is to remove the subsequent line to the one that called fuse_2_patches, which tried to deallocate a piece of memory that is already deallocated in fuse_2_patches. That appears to have fixed the problem, and the code now proceeds onwards with its business.
Screen output or output files showing the error message and context:
The text was updated successfully, but these errors were encountered: