'debug_blocks' only lists the block distribution on first proc since #602 #718
Assignees
Labels
Comments
|
This was done to have readable output. Otherwise, all pes are writing to the same file at the same time. I could probably modify this, so it loops over pes then barriers and flushes between pe writes. Or we could write out from each pe to it's own file. Or we can leave as is or something else. Preferences? |
|
I think looping over procs + barrier + flushing would be best. |
|
Working on a PR now to update the debug_blocks output. New output will include information about the overall decomposition as follows, |
apcraig
added a commit
to apcraig/CICE
that referenced
this issue
Aug 10, 2022
16 tasks
apcraig
added a commit
that referenced
this issue
Aug 15, 2022
* Update/improve debug_blocks output, see #718. * Add ICE_MEMUSE cice.settings flag for batch memory use Add set_env.memsmall, memmed, memlarge options To use, will require changes to the env machine files. Most machines will probably not use it. See #674. * Add setup_machparams.csh to compute batch/launch machine parameters Update cice.batch.csh and cice.launch.csh to use setup_machparams.csh See #650 * Update subroutine diagnostic_abort which calls print_state Update ice_transport_remap and ice_transport_driver to call diagnostic_abort during some errors. See also #622 * Update miniconda install information See #547 * Code cleanup based on compile with -Wall Code cleanup based on -std f2003 and f2008 checks Add -stand f08 to cheyenne_intel debug flags Add -std f2008 to cheyenne_gnu debug flags Code consistent with Fortran 2003 except for use of contiguous in 1d evp code. * Remove all trailing blank space with script * Update the cheyenne env so qc testing works Add configuration/scripts/tests/qctest.yml file Update documentation * Update Icepack * Clean up some output * fix comments * update print_state output
dabail10
pushed a commit
to ESCOMP/CICE
that referenced
this issue
Oct 4, 2022
* Update/improve debug_blocks output, see CICE-Consortium#718. * Add ICE_MEMUSE cice.settings flag for batch memory use Add set_env.memsmall, memmed, memlarge options To use, will require changes to the env machine files. Most machines will probably not use it. See CICE-Consortium#674. * Add setup_machparams.csh to compute batch/launch machine parameters Update cice.batch.csh and cice.launch.csh to use setup_machparams.csh See CICE-Consortium#650 * Update subroutine diagnostic_abort which calls print_state Update ice_transport_remap and ice_transport_driver to call diagnostic_abort during some errors. See also CICE-Consortium#622 * Update miniconda install information See CICE-Consortium#547 * Code cleanup based on compile with -Wall Code cleanup based on -std f2003 and f2008 checks Add -stand f08 to cheyenne_intel debug flags Add -std f2008 to cheyenne_gnu debug flags Code consistent with Fortran 2003 except for use of contiguous in 1d evp code. * Remove all trailing blank space with script * Update the cheyenne env so qc testing works Add configuration/scripts/tests/qctest.yml file Update documentation * Update Icepack * Clean up some output * fix comments * update print_state output
Sign up for free
to join this conversation on GitHub.
Already have an account?
Sign in to comment
Hi @apcraig, I did not have time to review this PR (#602) before it was merged.
Here:
CICE/cicecore/shared/ice_distribution.F90
Lines 184 to 189 in 2d81f56
I notice that you prevented the block distribution from being outputed for all procs. Was this intended ? If we are debugging decompositions, I would expect we want output from all procs....
I was using
debug_blocksjust now to debug a segfault and was very confused as to why all blocks were listed on proc 1, and I dug up this PR by looking at the history ofcreate_local_block_ids...Originally posted by @phil-blain in #602 (comment)
The text was updated successfully, but these errors were encountered: