cam6_3_159: Diagnostic rainbows and GPU fixes in PUMAS microphysics #702
Conversation
Katetc
requested review from
andrewgettelman,
cacraigucar,
peverwhee and
nusbaume
Katetc
changed the title
Katetc
changed the title
cacraigucar
changed the title
| do i = 1, ncol | ||
|
|
||
| ! Rainbows currently calculated on the grid, not subcolumn specific | ||
| do i = 1, ngrdcol |
There was a problem hiding this comment.
If I do not use SCM or SILHS (e.g., just the default F2000 compset), does ngrdcol equal to ncol in this case?
There was a problem hiding this comment.
If subcolumns are not turned on (SILHS or any other subcolumn source) then ngrdcol is equal to ncol. The SCM issue was a result of the ccs_config tag, but I'm not sure why.
There was a problem hiding this comment.
Thanks for the clarification. That is good to know.
Externals.cfg
Outdated
| @@ -1,5 +1,5 @@ | |||
| [ccs_config] | |||
| tag = ccs_config_cesm0.0.99 | |||
| tag = ccs_config_cesm0.0.85 | |||
There was a problem hiding this comment.
By changing the tag from ccs_config_cesm0.0.99 to ccs_config_cesm0.0.85, the PUMAS GPU code won't be compiled correctly. Will the ccs_config_cesm0.0.99 tag be brought separately later?
There was a problem hiding this comment.
The ccs_config_cesm0.0.99 tag was causing the single column model test to fail. I'm not sure why, and I don't think it's in the scope of this tag to figure it out. When you work with Brian to bring in the GPU test suite, hopefully you can find and address the problem. Or, if somebody else updates the externals for CAM, they will need to fix this test. I will update the description to note this.
There was a problem hiding this comment.
Oh, thanks for the clarification. Do you have the steps to reproduce the failed single column model test?
|
@Katetc - Which test is the failing scm test? I would suggest you add a prealpha test for it as well, so that when @fischer-ncar updates the ccs_config (which I think he has already done, or may be doing with the next alpha tag), that the test fails at that level. I assume that when you say scm, this is a SCAM test. We want to make sure that we don't break SCAM with this next beta tag. |
HI Cheryl, this test fails with the newest ccs_config tag: SCT_D_Ln7.T42_T42_mg17.QPC5.derecho_intel.cam-scm_prep and passes with the current one. So I'm rolling back the ccs_config tag in this PR. I'll open an issue about this. And I can add a prealpha marking to this test as well. |
|
I have added this issue to be addressed regarding the ccs_config tag: #1017 |
|
I have added this issue to be addressed regarding unexpected test DIFFs for HEMCO: #1018 |
Merge pull request ESCOMP#702 from PUMASDevelopment/katetc/rainbows_pr cam6_3_159: Diagnostic rainbows and GPU fixes in PUMAS microphysics Fixes ESCOMP#683 Add Diagnostic Rainbow Calculation Partially addresses ESCOMP#1007 Broken PUMAS GPU code and GPU regression test by bringing in an updated PUMAS external tag with fixed GPU code. However, the needed ccs_config tag ( ccs_config_cesm0.0.99 ) caused a failure in the single column tests right now. So the updates to the ccs_config external will be left for the next GPU fixing tag. RAINBOWS!!! I'M NOT KIDDING!!!!111! Also a new PUMAS tag that fixes GPU OpenACC calls. B4b Does not change answers. ESCOMP commit: 46ff80c
Merge pull request ESCOMP#702 from PUMASDevelopment/katetc/rainbows_pr cam6_3_159: Diagnostic rainbows and GPU fixes in PUMAS microphysics Fixes ESCOMP#683 Add Diagnostic Rainbow Calculation Partially addresses ESCOMP#1007 Broken PUMAS GPU code and GPU regression test by bringing in an updated PUMAS external tag with fixed GPU code. However, the needed ccs_config tag ( ccs_config_cesm0.0.99 ) caused a failure in the single column tests right now. So the updates to the ccs_config external will be left for the next GPU fixing tag. RAINBOWS!!! I'M NOT KIDDING!!!!111! Also a new PUMAS tag that fixes GPU OpenACC calls. B4b Does not change answers. ESCOMP commit: 46ff80c
Fixes #683 Add Diagnostic Rainbow Calculation
Partially addresses #1007 Broken PUMAS GPU code and GPU regression test by bringing in an updated PUMAS external tag with fixed GPU code. However, the needed ccs_config tag ( ccs_config_cesm0.0.99 ) caused a failure in the single column tests right now. So the updates to the ccs_config external will be left for the next GPU fixing tag.
RAINBOWS!!! I'M NOT KIDDING!!!!111!
Also a new PUMAS tag that fixes GPU OpenACC calls. B4b
Does not change answers.