Socretes kernels extended - #723
Benjamin Went (MetBenjaminWent) wants to merge 9 commits into
Conversation
|
PR and issue on pause due to lack of testing methods currently |
|
Tthese are used in the regional model so it would be good to put some OMP in there. |
|
| "plot_str": "plot_map.py -ral $NODAL_DATA_DIR/lfric_diagnostics.nc $PLOT_DIR", | ||
| }) %} | ||
|
|
||
| {% elif task_ns.conf_name == "ral3-2T-seuk_MG" %} |
There was a problem hiding this comment.
Just note here that care will need taking when merging with #764 to ensure names etc are correctly changed from ral3 to ral4
James Manners (mo-jmanners)
left a comment
There was a problem hiding this comment.
This is fine by me. Note the segment size tuning may be different for these increment calls than for the full SW/LW calls as these have far fewer bands, so less work to do per profile.
James Manners (@mo-jmanners), roger that, thanks, that's good to know. I can change this now given we know they'll be different, get ahead for future tuning? Or wait till then? As CO which would be preferable for you? |
Do you mean have a sw_inc_seg_limit_size separate from sw_seg_limit_size (and lw)? Yes - probably good to add in now. |
Matthew Hambley (MatthewHambley)
left a comment
There was a problem hiding this comment.
Code owner (build system) review: Lists additions in line with current usage.
PR Summary
Extent Source changes and Transmute scrits to other Socrates Domain kernels:
lw_inc_kernel_modsw_inc_kernel_modThese schemes are run by the RAL jobs, which get run with ukv jobs.
I've tested some smaller jobs which show the OMP gains at the smaller res for lw_inc, but still need some better results for sw_inc as it's currently too small to tell. However given both work like the existing functional schemes, these should be good.
run_lfric_atm_ral3-**_n_**T-seuk_MG_ex1a_cce_production-32bitThis is without the segment size tuning for these jobs, which is currently tied to the existing LW/SW segmentation sizes.
This might be okay, longer term, or given the differing configs which control these schemes, could not be.
However these should be tuned with further performance testing.
Sci/Tech Reviewer:
Code Reviewer: Matthew Walker (@mattatmet)
Issue: #619
Umbrella: #424
Code Quality Checklist
Testing
trac.log
Test Suite Results - lfric_apps - socretes_kernels_extended/run14
Suite Information
Task Information
✅ succeeded tasks - 13
Test Suite Results - lfric_apps - socretes_kernels_extended/run16
Suite Information
Task Information
❌ failed tasks - 36
⌛ waiting tasks - 19
Security Considerations
Performance Impact
AI Assistance and Attribution
Documentation
PSyclone Approval
Sci/Tech Review
(Please alert the code reviewer via a tag when you have approved the SR)
Code Review