Update sandiatoss3 pe-layouts for failing tri-grid tests #3295
Conversation
|
I don't think I should review this. I am unfamiliar with pe-layouts. |
| <grid name="a%r05_l%r05_oi%null_r%r05_g%null_w%null_z%null_m%oEC60to30v3"> | ||
| <mach name="sandiatoss3"> |
There was a problem hiding this comment.
I think it'd be good to set <mach name="any"> for both grids, tests with these grids are running into out-of-memory errors on multiple machines.
There was a problem hiding this comment.
Thanks for looking, @amametjanov . Do all the machines we test on have the resources? I can't remember how big melvin is...
There was a problem hiding this comment.
@wadeburgess can you try
./cime/scripts/create_test SMS_P256.r05_r05.I1850CLM45CN SMS_D_Ld1_P256.ne30_r05_oECv3.A_WCYCL1850
on Melvin?
There was a problem hiding this comment.
SMS.r05_r05.I1850CLM45CN.melvin_gnu passes without changing the layout, so it's not having the memory issues....
There was a problem hiding this comment.
IIRC, Melvin is a single node with 48 cores and very large memory. Oversubscribing (256 tasks on 48 cores) is allowed.
Another option is to set
<mach name="sandiatoss3|cori-knl|cori-haswell|theta|anvil|bebop">
There was a problem hiding this comment.
Melvin has a ton of memory but not 256 cores (we oversubscribe up to 64 on melvin)
There was a problem hiding this comment.
Thanks for all your help, @amametjanov and @jgfouca. I will change it to Az's second option, listing all the machine names...
Update sandiatoss3 pe-layouts for failing tri-grid tests This PR adds two specific pe-layouts for sandiatoss3 and other platforms to run two failing tri-grid tests: * SMS.r05_r05.I1850CLM45CN.sandiatoss3_intel * SMS_D_Ld1.ne30_r05_oECv3.A_WCYCL1850.sandiatoss3_intel These layouts were tested in SES-875 and now pass. [BFB] for all previously tested configurations
|
merged to next |
Update sandiatoss3 pe-layouts for failing tri-grid tests This PR adds two specific pe-layouts for sandiatoss3 and other platforms to run two failing tri-grid tests: * SMS.r05_r05.I1850CLM45CN.sandiatoss3_intel * SMS_D_Ld1.ne30_r05_oECv3.A_WCYCL1850.sandiatoss3_intel These layouts were tested in SES-875 and now pass. [BFB] for all previously tested configurations
|
merged to master |
These variables should get reasonable initial values if the namelist isn't read. Test suite: scripts_regression_tests.py had testing of SMS_Ld5.f09_g17.B1850G.cheyenne_intel.allactive-cism-test_coupling Test baseline: Test namelist changes: Test status: bit for bit Fixes #3294 User interface changes?: Update gh-pages html (Y/N)?: Code review:
This PR adds two specific pe-layouts for sandiatoss3 to run two failing tri-grid tests:
These layouts were tested in SES-875 and now pass.
[BFB] for all previously tested configurations