[gmx-developers] GROMACS OpenCL on Gallium

[Roland Schulz]

On Thu, Nov 26, 2015 at 3:23 PM, Szilárd Páll <pall.szilard at gmail.com>
wrote:

> On Thu, Nov 26, 2015 at 9:01 PM, Roland Schulz <roland at utk.edu> wrote:
>
>>
>>
>> On Thu, Nov 26, 2015 at 2:51 PM, Szilárd Páll <pall.szilard at gmail.com>
>> wrote:
>>
>>>
>>> On Thu, Nov 26, 2015 at 8:25 PM, Roland Schulz <roland at utk.edu> wrote:
>>>
>>>>
>>>>
>>>> On Thu, Nov 26, 2015 at 2:07 PM, Szilárd Páll <pall.szilard at gmail.com>
>>>> wrote:
>>>>
>>>>> Hi,
>>>>>
>>>>> Besides FPGA folks, what about Apple, embedded and mobile platforms
>>>>> (Qualcomm, ARM, Samsung, etc.)?
>>>>>
>>>> All the embedded GPUs are for the foreseeable futures uninteresting for
>>>> HPC.
>>>>
>>>
>>> Note that HPC does, in most cases, not drive programming
>>> language/framework/standard adoption.
>>>
>>
>> But we only have influence in HPC. Thus if that is true, then we don't
>> have any influence anyhow. But I don't think that's always true. Especially
>> if you look at OpenACC/OpenMP/Fortran (even some new things in C++).
>>
>
> Let me not look at OpenACC, please. :)
>

Maybe HPC shouldn't have influence on languages: Fotran, OpenACC, ... ;-)

I agree that OpenCL still matters outside of HPC. I only meant that it
>> might not matter in the future for HPC, which is the only area where our
>> decision have any influence.
>>
>
> Actually, I would argue that if you consider the number of users, GROMACS
> likely has most likely more use on commodity computing resources (that
> includes laptops, workstations, and university basement clusters) than on
> actual HPC iron - even without counting F at H. In terms of cycles burnt
> that may not be the case. Also, I'm not entirely sure what the balance is
> when in terms of where the science output is produced (and I won't even try
> to weight that by quality). However, in terms of raw publication numbers a
> lot of science may not come from GROMACS running on "proper HPC"
> facilities. Than there is the education impact of GROMACS where the
> situation is very non-HPC.
>

I agree. But I think what I wrote is true even if we use HPC very
broadly.Like NVidia sometimes uses it for anything with a single Tesla.
That still excludes anything embedded or Iris and mostly APU too. And
according to that broad use of HPC, most of GROMACS is HPC.


> However, this is a bit of a chicken-egg problem and that's why I believe
> it matters whether we take the back seat or put some effort into having an
> honest alternative! This does not have to be super-tuned, but it does have
> to work, have mostly complete feature support, for further extensions,
> porting, or performance tuning when an opportunity arises, would you not
> agree?
>

It certainly makes sense for us to try new things, to explore technologies
in an early phase, and find out whether they are useful and drive research.
But that isn't the issue with OpenCL. I think few people claim that there
are inherit issues with the OpenCL standard which makes it impossible to
get good performance. It has to do with the effort put into the
implementation. As one can see with JavaScript, performance is more in the
implementation than in the language. So I don't think we need a proof of
usefulness at this point but the business case for the implementers to put
(significant) more resources into it. Not sure how we can do that. Because
like I said, until we actually drop CUDA we aren't an important OpenCL user.

Roland


>
> --
> Szilárd
>
>
>>
>> Roland
>>
>> --
>> 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.
>>
>
>


-- 
ORNL/UT Center for Molecular Biophysics cmb.ornl.gov
865-241-1537, ORNL PO BOX 2008 MS6309
-------------- next part --------------
An HTML attachment was scrubbed...
URL: <http://maillist.sys.kth.se/pipermail/gromacs.org_gmx-developers/attachments/20151126/90fd3f7a/attachment-0001.html>