<div dir="ltr">That's great to hear, Mirco.<div><br></div><div>A message to stdout before and after compilation does sound like a good idea, so users waiting on the terminal know that this is normal.</div><div><br></div><div>Mark</div></div><br><div class="gmail_quote"><div dir="ltr">On Thu, Jun 25, 2015 at 11:09 PM Mirco Wahab <<a href="mailto:mirco.wahab@chemie.tu-freiberg.de">mirco.wahab@chemie.tu-freiberg.de</a>> wrote:<br></div><blockquote class="gmail_quote" style="margin:0 0 0 .8ex;border-left:1px #ccc solid;padding-left:1ex">On 25.06.2015 17:02, Mark Abraham wrote:<br>
> I got local access to a machine with two AMD GPUs, and ran into the same<br>
> initialization issues I'd seen with Nvidia runtimes, so I have updated<br>
> <a href="https://gerrit.gromacs.org/4314" rel="noreferrer" target="_blank">https://gerrit.gromacs.org/4314</a> so that in 5.1 GROMACS will never use<br>
> more than one OpenCL device per node. We can probably fix that in the<br>
> longer term, but it seems like issues of host-side concurrency are not<br>
> treated in the OpenCL standard (or at least the implementations we've<br>
> targeted), and we didn't do a good enough job of anticipating and<br>
> managing that.<br>
<br>
In this moment I completed compilation and run some first tests on<br>
my AMD box (AMD Phenom II X6 1090t + R9-270X Pitcairn GCN 1.1).<br>
<br>
Compilation/Installation:<br>
I fetched refs/changes/14/4314/38 against the current git master,<br>
compilation (VS 2013/MSVC 18.00.31101, AMD App-SDK 3.0.0-beta)<br>
went smooth, not a single error (built against a local fftw 3.3.3),<br>
so that looks really promising ...<br>
<br>
First run tests (for posting the status here):<br>
- single water box (2916 SPC)<br>
- NVT + RF ok.<br>
- NVT + PME ok.<br>
- NPT + RF ok.<br>
- NPT + PME ok.<br>
<br>
All runs did complete without any problem!<br>
Run times are promising too on this<br>
very small system (ns/day):<br>
<br>
cpu only - rf/npt: 32.482, pme/npt: 20.971<br>
cpu/gpu - rf/npt: 63.530, pme/npt: 51.429<br>
rf/nvt: 64.001, pme/nvt: 51.125<br>
<br>
The gpu kernels do get compiled in this version<br>
at the start of every run and are located in the<br>
%TEMP% directory - and deleted after loading, So they<br>
are not longer cached/stored. The time for gpu kernel<br>
compilation is >30 sec on this box (maybe this rewards<br>
an extra message). The structure of this file is<br>
(looks like) an ELF-binary?<br>
<br>
As it looks, the OpenCL extension could be a<br>
very useable attribute for Gromacs users. I'm<br>
impressed.<br>
<br>
<br>
Thanks & Regards<br>
<br>
M.<br>
<br>
--<br>
Gromacs Developers mailing list<br>
<br>
* Please search the archive at <a href="http://www.gromacs.org/Support/Mailing_Lists/GMX-developers_List" rel="noreferrer" target="_blank">http://www.gromacs.org/Support/Mailing_Lists/GMX-developers_List</a> before posting!<br>
<br>
* Can't post? Read <a href="http://www.gromacs.org/Support/Mailing_Lists" rel="noreferrer" target="_blank">http://www.gromacs.org/Support/Mailing_Lists</a><br>
<br>
* For (un)subscribe requests visit<br>
<a href="https://maillist.sys.kth.se/mailman/listinfo/gromacs.org_gmx-developers" rel="noreferrer" target="_blank">https://maillist.sys.kth.se/mailman/listinfo/gromacs.org_gmx-developers</a> or send a mail to <a href="mailto:gmx-developers-request@gromacs.org" target="_blank">gmx-developers-request@gromacs.org</a>.<br>
</blockquote></div>