[ecco-support] [EXTERNAL] Re: Cost function calculation problems when changing number of cores in ECCOv4-r4
Wang, Ou (US 329B)
ou.wang at jpl.nasa.gov
Wed Dec 30 16:55:26 EST 2020
Hi Dan,
Thanks for posting STDOUT.0000.
I don’t see anything obviously wrong in STDOUT.0000. The modified namelist for data.exch2 also looks good to me. I wonder if the following files for your runs with 360 or 192 cores were corrupted;
mdt2_dtu17.bin
mdt_dtu13_error_m_areascaled.bin
sshv4_scale_1p5points.bin
Regards,
Ou
From: <ecco-support-bounces at mit.edu> on behalf of Dan Jones <dcjones.work at gmail.com>
Reply-To: "ECCO support list, wider membership" <ecco-support at mit.edu>
Date: Wednesday, December 30, 2020 at 10:35 AM
To: "ecco-support at mit.edu" <ecco-support at mit.edu>
Subject: [EXTERNAL] Re: [ecco-support] Cost function calculation problems when changing number of cores in ECCOv4-r4
[Reposting, forgot to change subject line in last email]
Hi Ou and Patrick,
Thanks for your quick replies. I've been modifying data.exch2 (i.e. uncommenting and commenting out the relevant sections) with respect to the different core numbers, so hopefully that part is okay.
Here's one of the STDOUT files from a run with 360 cores:
https://raw.githubusercontent.com/DanJonesOcean/myscripts/master/STDOUT.0000<https://urldefense.us/v3/__https:/raw.githubusercontent.com/DanJonesOcean/myscripts/master/STDOUT.0000__;!!PvBDto6Hs4WbVuu7!cUNrxHVuOjHouyqPmInRluNudUHCf5VKsRGTUHHo5ND5zlrTlrfVwrzsgoIDxGyKA7c$>
Thanks,
Dan
On Wed, Dec 30, 2020 at 5:58 PM <ecco-support-request at mit.edu<mailto:ecco-support-request at mit.edu>> wrote:
Send ecco-support mailing list submissions to
ecco-support at mit.edu<mailto:ecco-support at mit.edu>
To subscribe or unsubscribe via the World Wide Web, visit
http://mailman.mit.edu/mailman/listinfo/ecco-support<https://urldefense.us/v3/__http:/mailman.mit.edu/mailman/listinfo/ecco-support__;!!PvBDto6Hs4WbVuu7!cUNrxHVuOjHouyqPmInRluNudUHCf5VKsRGTUHHo5ND5zlrTlrfVwrzsgoIDYL-Yhi8$>
or, via email, send a message with subject or body 'help' to
ecco-support-request at mit.edu<mailto:ecco-support-request at mit.edu>
You can reach the person managing the list at
ecco-support-owner at mit.edu<mailto:ecco-support-owner at mit.edu>
When replying, please edit your Subject line so it is more specific
than "Re: Contents of ecco-support digest..."
Today's Topics:
1. Re: Cost function calculation problems when changing number
of cores in ECCOv4-r4 (Dan Jones)
2. Re: [EXTERNAL] Re: Cost function calculation problems when
changing number of cores in ECCOv4-r4 (Wang, Ou (US 329B))
3. Re: [EXTERNAL] Re: Cost function calculation problems when
changing number of cores in ECCOv4-r4 (Patrick Heimbach)
----------------------------------------------------------------------
Message: 1
Date: Wed, 30 Dec 2020 17:27:39 +0000
From: Dan Jones <dcjones.work at gmail.com<mailto:dcjones.work at gmail.com>>
Subject: Re: [ecco-support] Cost function calculation problems when
changing number of cores in ECCOv4-r4
To: ecco-support at mit.edu<mailto:ecco-support at mit.edu>
Message-ID:
<CAPj3iHT55=rFszYS1E0rF4gmc=B39EbiEMQSD+wWMe9Bf3E0OQ at mail.gmail.com<mailto:B39EbiEMQSD%2BwWMe9Bf3E0OQ at mail.gmail.com>>
Content-Type: text/plain; charset="utf-8"
Hi Hong,
So, turning off the "profiles" package helped with some of the runs, but
not with others.
In some of the runs, I noticed that the "sshv4-mdt" term of the cost
function is NaN. Does that term also depend on the grid? It doesn't cause
problems with the 96 core version, but it does cause an issue with the 192
and 360 core versions.
Best,
Dan
-------------- next part --------------
An HTML attachment was scrubbed...
URL: http://mailman.mit.edu/pipermail/ecco-support/attachments/20201230/2cbaf1d6/attachment-0001.html<https://urldefense.us/v3/__http:/mailman.mit.edu/pipermail/ecco-support/attachments/20201230/2cbaf1d6/attachment-0001.html__;!!PvBDto6Hs4WbVuu7!cUNrxHVuOjHouyqPmInRluNudUHCf5VKsRGTUHHo5ND5zlrTlrfVwrzsgoID-9OwMMM$>
------------------------------
Message: 2
Date: Wed, 30 Dec 2020 17:46:51 +0000
From: "Wang, Ou (US 329B)" <ou.wang at jpl.nasa.gov<mailto:ou.wang at jpl.nasa.gov>>
Subject: Re: [ecco-support] [EXTERNAL] Re: Cost function calculation
problems when changing number of cores in ECCOv4-r4
To: "ECCO support list, wider membership" <ecco-support at mit.edu<mailto:ecco-support at mit.edu>>
Message-ID: <AAF31FEB-9653-4CA8-A603-EE365D4C4958 at jpl.nasa.gov<mailto:AAF31FEB-9653-4CA8-A603-EE365D4C4958 at jpl.nasa.gov>>
Content-Type: text/plain; charset="utf-8"
Dan,
The ?ssh-mdt? term of the cost function shouldn?t depend on the grid. Could you post your STDOUT.0000 so I can take a look at it?
Thanks,
Ou
From: <ecco-support-bounces at mit.edu<mailto:ecco-support-bounces at mit.edu>> on behalf of Dan Jones <dcjones.work at gmail.com<mailto:dcjones.work at gmail.com>>
Reply-To: "ECCO support list, wider membership" <ecco-support at mit.edu<mailto:ecco-support at mit.edu>>
Date: Wednesday, December 30, 2020 at 9:28 AM
To: "ecco-support at mit.edu<mailto:ecco-support at mit.edu>" <ecco-support at mit.edu<mailto:ecco-support at mit.edu>>
Subject: [EXTERNAL] Re: [ecco-support] Cost function calculation problems when changing number of cores in ECCOv4-r4
Hi Hong,
So, turning off the "profiles" package helped with some of the runs, but not with others.
In some of the runs, I noticed that the "sshv4-mdt" term of the cost function is NaN. Does that term also depend on the grid? It doesn't cause problems with the 96 core version, but it does cause an issue with the 192 and 360 core versions.
Best,
Dan
-------------- next part --------------
An HTML attachment was scrubbed...
URL: http://mailman.mit.edu/pipermail/ecco-support/attachments/20201230/6d13a65a/attachment-0001.html<https://urldefense.us/v3/__http:/mailman.mit.edu/pipermail/ecco-support/attachments/20201230/6d13a65a/attachment-0001.html__;!!PvBDto6Hs4WbVuu7!cUNrxHVuOjHouyqPmInRluNudUHCf5VKsRGTUHHo5ND5zlrTlrfVwrzsgoIDGOu3w5o$>
------------------------------
Message: 3
Date: Wed, 30 Dec 2020 17:58:18 +0000
From: Patrick Heimbach <heimbach at mit.edu<mailto:heimbach at mit.edu>>
Subject: Re: [ecco-support] [EXTERNAL] Re: Cost function calculation
problems when changing number of cores in ECCOv4-r4
To: ecco-support <ecco-support at mit.edu<mailto:ecco-support at mit.edu>>
Message-ID: <9A726223-81E7-4146-898B-F56C690AF505 at mit.edu<mailto:9A726223-81E7-4146-898B-F56C690AF505 at mit.edu>>
Content-Type: text/plain; charset="utf-8"
You may want to check the blanklist in data.exch2 which is specific to the tiling.
p.
> On Dec 30, 2020, at 11:46 AM, Wang, Ou (US 329B) <ou.wang at jpl.nasa.gov<mailto:ou.wang at jpl.nasa.gov>> wrote:
>
> Dan,
>
> The ?ssh-mdt? term of the cost function shouldn?t depend on the grid. Could you post your STDOUT.0000 so I can take a look at it?
>
> Thanks,
> Ou
>
> From: <ecco-support-bounces at mit.edu<mailto:ecco-support-bounces at mit.edu>> on behalf of Dan Jones <dcjones.work at gmail.com<mailto:dcjones.work at gmail.com>>
> Reply-To: "ECCO support list, wider membership" <ecco-support at mit.edu<mailto:ecco-support at mit.edu>>
> Date: Wednesday, December 30, 2020 at 9:28 AM
> To: "ecco-support at mit.edu<mailto:ecco-support at mit.edu>" <ecco-support at mit.edu<mailto:ecco-support at mit.edu>>
> Subject: [EXTERNAL] Re: [ecco-support] Cost function calculation problems when changing number of cores in ECCOv4-r4
>
> Hi Hong,
>
> So, turning off the "profiles" package helped with some of the runs, but not with others.
>
> In some of the runs, I noticed that the "sshv4-mdt" term of the cost function is NaN. Does that term also depend on the grid? It doesn't cause problems with the 96 core version, but it does cause an issue with the 192 and 360 core versions.
>
> Best,
> Dan
> _______________________________________________
> ecco-support mailing list
> ecco-support at mit.edu<mailto:ecco-support at mit.edu>
> http://mailman.mit.edu/mailman/listinfo/ecco-support<https://urldefense.us/v3/__http:/mailman.mit.edu/mailman/listinfo/ecco-support__;!!PvBDto6Hs4WbVuu7!cUNrxHVuOjHouyqPmInRluNudUHCf5VKsRGTUHHo5ND5zlrTlrfVwrzsgoIDYL-Yhi8$>
-------------- next part --------------
A non-text attachment was scrubbed...
Name: smime.p7s
Type: application/pkcs7-signature
Size: 1865 bytes
Desc: not available
Url : http://mailman.mit.edu/pipermail/ecco-support/attachments/20201230/3ce86058/attachment.bin<https://urldefense.us/v3/__http:/mailman.mit.edu/pipermail/ecco-support/attachments/20201230/3ce86058/attachment.bin__;!!PvBDto6Hs4WbVuu7!cUNrxHVuOjHouyqPmInRluNudUHCf5VKsRGTUHHo5ND5zlrTlrfVwrzsgoIDXo3lTbA$>
------------------------------
_______________________________________________
ecco-support mailing list
ecco-support at mit.edu<mailto:ecco-support at mit.edu>
http://mailman.mit.edu/mailman/listinfo/ecco-support<https://urldefense.us/v3/__http:/mailman.mit.edu/mailman/listinfo/ecco-support__;!!PvBDto6Hs4WbVuu7!cUNrxHVuOjHouyqPmInRluNudUHCf5VKsRGTUHHo5ND5zlrTlrfVwrzsgoIDYL-Yhi8$>
End of ecco-support Digest, Vol 59, Issue 3
*******************************************
-------------- next part --------------
An HTML attachment was scrubbed...
URL: http://mailman.mit.edu/pipermail/ecco-support/attachments/20201230/d533f64e/attachment-0001.html
More information about the ecco-support
mailing list