[gmx-developers] GROMACS OpenCL on Gallium

[Szilárd Páll]

On Mon, Dec 7, 2015 at 1:34 PM, Mark Abraham <mark.j.abraham at gmail.com>
wrote:

> Hi,
>
> regressiontests are fine for me with OpenCL. But if you are still testing
> release-5-1 code against master-branch regressiontests, that won't work,
> whether for OpenCL or not.
>

http://jenkins.gromacs.org/job/Gromacs_Gerrit_5_1-test-opencl-slave/14

*12:06:31* -----------------------------------------------------------*12:06:31*
Building using versions:*12:06:31* gromacs:
refs/heads/release-5-1
860a5922326ff56394d6c87d065388881dda57c4*12:06:31* regressiontests:
 refs/heads/release-5-1
e3e5b95c15d22780540874e2543e270da5afc6c9*12:06:31* releng:
 refs/heads/5.0.0
91fda692d97c057ea02a65e9ab96ed765600879a

*12:06:31* -----------------------------------------------------------


Three tests segfault with 5.1 release code and 5.1 regressiontests and the
above seems to be consistent with the the tests triggered in the
"Gromacs_Gerrit_5_1" project.

A previous trigger did indeed use master regressiontests because that's
what the jenkins "retrigger" offered and I thought for a second that may
not be unusual given that there seemed to be no releng branch named
consistently with the source code-base either.

--
Szilárd



>
> Mark
>
> On Mon, Dec 7, 2015 at 10:17 PM Szilárd Páll <pall.szilard at gmail.com>
> wrote:
>
>> On Mon, Dec 7, 2015 at 12:15 PM, Szilárd Páll <pall.szilard at gmail.com>
>> wrote:
>>
>>> On Fri, Dec 4, 2015 at 7:54 PM, Szilárd Páll <pall.szilard at gmail.com>
>>> wrote:
>>>
>>>> On Fri, Dec 4, 2015 at 7:26 PM, Vedran Miletić <rivanvx at gmail.com>
>>>> wrote:
>>>>
>>>>> 2015-12-04 18:13 GMT+01:00 Szilárd Páll <pall.szilard at gmail.com>:
>>>>> > Sounds good, is there anything on our side we can help with?
>>>>> >
>>>>> > Cheers,
>>>>> >
>>>>> > --
>>>>> > Szilárd
>>>>>
>>>>> Just making sure OpenCL works correctly with AMD and NVIDIA
>>>>> proprietary drivers is enough, so I can consider it a perfectly
>>>>> working black box.
>>>>>
>>>>
>>>> I hope that those cases are covered by our CI setup, but I better
>>>> verify.
>>>>
>>>
>>> Unfortunately we still don't have CI verification of OpenCL builds.
>>> There is a manual-trigger test project, but nothing gets tested
>>> automatically.
>>>
>>
>> Additionally, some of the regressiontests don't even pass which is
>> worrying. I'll get back to this later.
>>
>>
>>>
>>> I'll try to look into this, but I'm very up-to-date with the state of
>>> the jenkins setup.
>>>
>>>
>>>>
>>>>>
>>>>> When I get GROMACS to run, I will require help with testing
>>>>> correctness and performance measurements.
>>>>>
>>>>
>>>> I can certainly help with that. If the regressiontests pass (make
>>>> check) most likely everything is fine.
>>>> One more tip in case if you need it: for a quick correctness check I
>>>> suggest running a few steps (>0, e.g. 1-5) and diff-ing the energy terms
>>>> printed in the log file (increase frequency using nstlog in the mdp)
>>>> against a known to be correct mdrun. Slightly more elaborate is to, instead
>>>> of diff-ing, use gmxcheck to check output (e.g. energy or force) against a
>>>> reference output.
>>>>
>>>> Cheers,
>>>> --
>>>> Szilárd
>>>>
>>>>
>>>>> Regards,
>>>>> Vedran
>>>>>
>>>>> --
>>>>> Vedran Miletić
>>>>> http://vedranmileti.ch/
>>>>> --
>>>>> Gromacs Developers mailing list
>>>>>
>>>>> * Please search the archive at
>>>>> http://www.gromacs.org/Support/Mailing_Lists/GMX-developers_List
>>>>> before posting!
>>>>>
>>>>> * Can't post? Read http://www.gromacs.org/Support/Mailing_Lists
>>>>>
>>>>> * For (un)subscribe requests visit
>>>>> https://maillist.sys.kth.se/mailman/listinfo/gromacs.org_gmx-developers
>>>>> or send a mail to gmx-developers-request at gromacs.org.
>>>>>
>>>>
>>>>
>>> --
>> Gromacs Developers mailing list
>>
>> * Please search the archive at
>> http://www.gromacs.org/Support/Mailing_Lists/GMX-developers_List before
>> posting!
>>
>> * Can't post? Read http://www.gromacs.org/Support/Mailing_Lists
>>
>> * For (un)subscribe requests visit
>> https://maillist.sys.kth.se/mailman/listinfo/gromacs.org_gmx-developers
>> or send a mail to gmx-developers-request at gromacs.org.
>
>
> --
> Gromacs Developers mailing list
>
> * Please search the archive at
> http://www.gromacs.org/Support/Mailing_Lists/GMX-developers_List before
> posting!
>
> * Can't post? Read http://www.gromacs.org/Support/Mailing_Lists
>
> * For (un)subscribe requests visit
> https://maillist.sys.kth.se/mailman/listinfo/gromacs.org_gmx-developers
> or send a mail to gmx-developers-request at gromacs.org.
>
-------------- next part --------------
An HTML attachment was scrubbed...
URL: <http://maillist.sys.kth.se/pipermail/gromacs.org_gmx-developers/attachments/20151207/3f299a7b/attachment.html>