[gmx-developers] lots of new SIMD code but very poor jenkins coverage!

Thu Mar 10 16:56:47 CET 2016

On Wed, Mar 9, 2016 at 4:43 PM, Mark Abraham <mark.j.abraham at gmail.com>
wrote:

> Hi,
>
>
> On Wed, Mar 9, 2016 at 3:42 PM Szilárd Páll <pall.szilard at gmail.com>
> wrote:
>
>> Hi,
>>
>> No, I don't. I triggered it the other day because I did not know the main
>> 5.1 matrix does the job already.
>>
>
> OK, I'll delete it tomorrow.
>
>
>> What about master, additional releng support is needed I guess? Also, is
>> there anything holding us back from adding pre- and post-submit configs
>> to admin//builds/*matrix.txt.
>>
>
> See my other email to gmx-developers just now. I think pre-submit should
> cover only the highest-value targets. Post-submit for lower-value targets
> and/or tests that take more time.
>

I agree. A dozen or so configs to cover the most common
software/parallelization configurations should be enough. Having looked at
the Gromacs_Gerrit_master_nrwpo configs, I propose the following
additions/tweaks:
- OpenCL (I suggest using AMD GPUs and adding both release and debug, error
handling is rather poor and OpenCL API erros are mostly handled with
assert's)
- an SSE4.1 (i.e. 4-wide SIMD) build unless  - AFAICT VM-based builds are
not already doing that;
- some runs that execute tests with 2D/3D DD (if 3D is hw resource-wise an
issue could be left for post-submit),

Additionally:
- Has valgrind testing been discontinued? (Teemu mentioned on #1815 that
valgrind configs got accidentally removed at some point?)
- How about TSAN? I've seen no confgis use it (but we've been using it for
5.1, right?)

--
Szilárd

>
> Mark
>
>
>> Cheers,
>>
>> --
>> Szilárd
>>
>> On Wed, Mar 9, 2016 at 3:32 PM, Mark Abraham <mark.j.abraham at gmail.com>
>> wrote:
>>
>>> Hi,
>>>
>>> Sorry, I wrote this email at least a month ago, but didn't send it...
>>>
>>> It's not strictly related, but now that the code is fixed, I have copied
>>> four of the five OpenCL configs from
>>> http://jenkins.gromacs.org/job/Gromacs_Gerrit_5_1-test-opencl-slave/
>>>  into http://jenkins.gromacs.org/job/Gromacs_Gerrit_5_1/.
>>>
>>> "Compiler=gcc CompilerVersion=5 GMX_GPU=ON GMX_USE_OPENCL=ON
>>> AMDAPPSDK=3.0 CMAKE_BUILD_TYPE=Release host=bs_nix-amd_gpu"
>>> "Compiler=gcc CompilerVersion=5 GMX_GPU=ON GMX_USE_OPENCL=ON
>>> AMDAPPSDK=3.0 GMX_THREAD_MPI=off CMAKE_BUILD_TYPE=Debug host=bs_nix-amd_gpu"
>>> "Compiler=gcc CompilerVersion=4.7 GMX_GPU=ON GMX_USE_OPENCL=ON CUDA=6.5
>>> GMX_SIMD=SSE4.1 CMAKE_BUILD_TYPE=Debug host=bs_nix1204"
>>> "Compiler=gcc CompilerVersion=4.8 GMX_GPU=ON GMX_USE_OPENCL=ON CUDA=7.5
>>> GMX_MPI=on CMAKE_BUILD_TYPE=Debug host=bs_nix1310"
>>>
>>> So unless e.g. Szilard still wants it, we can delete
>>> http://jenkins.gromacs.org/job/Gromacs_Gerrit_5_1-test-opencl-slave/
>>>
>>> Mark
>>>
>>> On Wed, Jan 6, 2016 at 2:48 PM Szilárd Páll <pall.szilard at gmail.com>
>>> wrote:
>>>
>>>> Thanks for the explanation!
>>>>
>>>> --
>>>> Szilárd
>>>>
>>>> On Mon, Jan 4, 2016 at 9:21 PM, Teemu Murtola <teemu.murtola at gmail.com>
>>>> wrote:
>>>>
>>>>> Hi,
>>>>>
>>>>> The new pre-submit job output looks nearly identical to the old matrix
>>>>> job (it is still a matrix job). Before it is fully taken into use, you can
>>>>> push changes with "verify-matrix" as their topic, and the new job is also
>>>>> triggered for them; all releng changes currently also trigger both jobs.
>>>>>
>>>>> The Jenkins configuration is a bit more complicated with two
>>>>> interconnected jobs, but you will only see the actual matrix linked from
>>>>> Gertit job as long as everything works, with a few caveats:
>>>>> * To retrigger a build, you need to navigate to the upstream job
>>>>> instead of using the build linked from Gerrit.
>>>>> * If you push changes to the matrix that make it invalid, you will see
>>>>> the non-matrix launcher job getting linked from Gerrit instead of the
>>>>> actual matrix job.
>>>>> * And if someone manually aborts the matrix job, there can be some
>>>>> glitches (I think that currently, the build is reported as failed instead
>>>>> of aborted).
>>>>>
>>>>> Best regards,
>>>>> Teemu
>>>>>
>>>>>
>>>>> On Mon, Jan 4, 2016, 21:34 Szilárd Páll <pall.szilard at gmail.com>
>>>>> wrote:
>>>>>
>>>>>> Hi,
>>>>>>
>>>>>> Admittedly, I forgot about the pre-submit configs will be stored in
>>>>>> the repository. Can we see the output of those tests anywhere? Is the
>>>>>> matrix view retained or will it behave like the old "workflow" project's
>>>>>> output?
>>>>>>
>>>>>> I'll give a thought to the configs and will review Mark's version
>>>>>> when it gets pushed up.
>>>>>>
>>>>>> The comment-based post-submit trigger trigger sounds pretty useful!
>>>>>> With this feature, at some future point it may even be useful to allow
>>>>>> triggering subset of tests from a the larger collection of e.g. weekly test
>>>>>> set.
>>>>>>
>>>>>>
>>>>>> --
>>>>>> Szilárd
>>>>>>
>>>>>> On Thu, Dec 31, 2015 at 5:07 PM, Mark Abraham <
>>>>>> mark.j.abraham at gmail.com> wrote:
>>>>>>
>>>>>>> Hi,
>>>>>>>
>>>>>>> Yes the current pre-submit configuration should be good to go, I
>>>>>>> tried it out again last week. IIRC I/we thought that letting Gerrit rebase
>>>>>>> over that content naturally for a few weeks would let us do a clean switch
>>>>>>> to new Jenkins jobs. I'd been deferring actually doing it until I was back
>>>>>>> in Stockholm after a month in Australia :-)
>>>>>>>
>>>>>>> I have significant content intended for a post-submit matrix in
>>>>>>> Gerrit right now (but am reworking it this week). We probably also want to
>>>>>>> cover some differ combinations in post-submit.
>>>>>>>
>>>>>>> Mark
>>>>>>>
>>>>>>> On Wed, 30 Dec 2015 14:40 Teemu Murtola <teemu.murtola at gmail.com>
>>>>>>> wrote:
>>>>>>>
>>>>>>>> Hi,
>>>>>>>>
>>>>>>>> On Wed, Dec 30, 2015 at 4:11 PM Szilárd Páll <
>>>>>>>> pall.szilard at gmail.com> wrote:
>>>>>>>>
>>>>>>>>> Unless I'm mistaken, in the current config it's still as simple as
>>>>>>>>> adding a new line in the matrix for Gromacs_Gerrit_master-new-releng.
>>>>>>>>> However, I'd prefer to not do this based on what I fancy but rather based
>>>>>>>>> on what we agree is important.
>>>>>>>>>
>>>>>>>>
>>>>>>>> I'd strongly prefer that we instead start using the configuration
>>>>>>>> in admin/builds/pre-submit-matrix.txt. That also partially solves the
>>>>>>>> problem of agreeing on the things, since the configuration can be discussed
>>>>>>>> in Gerrit before applying any changes. Everything on this front seems to
>>>>>>>> have stopped after 5249 <https://gerrit.gromacs.org/5249> got
>>>>>>>> merged, though; I don't see why we shouldn't have switched to that a month
>>>>>>>> ago.
>>>>>>>>
>>>>>>>> Not sure how to add new configs for the post-submit tests. Do we
>>>>>>>>> have jenkins project for that?
>>>>>>>>>
>>>>>>>>
>>>>>>>> There is no post-submit tests, since there is no useful content for
>>>>>>>> them. If people create a set of desired configurations in
>>>>>>>> admin/builds/post-submit-matrix.txt, I can easily create a config that
>>>>>>>> actually builds using it.
>>>>>>>>
>>>>>>>> Teemu
>>>>>>>> --
>>>>>>>> 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.
>>>>>
>>>>>
>>>>> --
>>>>> 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.
>>>
>>
>> --
>> 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/20160310/523221bc/attachment-0001.html>