<div dir="ltr">Hi,<div><br></div><div>Thanks. If a difference of that magnitude can be seen, then it should also show up when running the regressiontests (e.g. cmake -DREGRESSIONTEST_DOWNLOAD=on and then make check) as a failure of complex/nbnxn-ljpme-LB (which is the only P-R test that can run on the GPU). If other tests fail, then the problem is actually more widespread.</div><div><br></div><div>It may be that there is some issue with some part of the Mac+clang+OpenCL stack - we didn't target it during development, and at the last minute when Erik was unexpectedly able to get it to compile. I don't know if he got tests to pass. Erik?</div><div><br></div><div>Mark</div></div><br><div class="gmail_quote"><div dir="ltr">On Wed, Jul 15, 2015 at 5:22 PM Carlo Camilloni <<a href="mailto:carlo.camilloni@gmail.com">carlo.camilloni@gmail.com</a>> wrote:<br></div><blockquote class="gmail_quote" style="margin:0 0 0 .8ex;border-left:1px #ccc solid;padding-left:1ex"><br>
Dear Mark and Szilard,<br>
<br>
thanks for your answer. I filed a bug in redmine but in the meantime I was running more tests and I am a bit scared by what I found:<br>
<br>
what I have done is the following I have performed a single step run with gmx51-rc1 compiled with cuda, again clang and so on<br>
and compared the forces on the first step with -nb cpu or not (I am using -pforce 1), the forces are identical:<br>
<br>
ie.:<br>
<br>
cuda-gpu<br>
<br>
step 0 atom 1 x 3.940 5.612 2.226 force 1.90839e+03<br>
step 0 atom 2 x 3.852 5.659 2.211 force 4.24845e+02<br>
step 0 atom 3 x 3.979 5.665 2.303 force 6.89472e+02<br>
step 0 atom 4 x 3.992 5.610 2.139 force 7.42053e+02<br>
<br>
<br>
cpu:<br>
<br>
step 0 atom 1 x 3.940 5.612 2.226 force 1.90839e+03<br>
step 0 atom 2 x 3.852 5.659 2.211 force 4.24845e+02<br>
step 0 atom 3 x 3.979 5.665 2.303 force 6.89472e+02<br>
step 0 atom 4 x 3.992 5.610 2.139 force 7.42053e+02<br>
<br>
if I do the same test on the version compiled with OpenCL<br>
<br>
cpu:<br>
<br>
(the former are done on my macbook pro avx2_256 this latter on a MacPro avx_256, this should<br>
explain the small differences in the forces)<br>
<br>
step 0 atom 1 x 3.940 5.612 2.226 force 1.90838e+03<br>
step 0 atom 2 x 3.852 5.659 2.211 force 4.24848e+02<br>
step 0 atom 3 x 3.979 5.665 2.303 force 6.89470e+02<br>
step 0 atom 4 x 3.992 5.610 2.139 force 7.42043e+02<br>
<br>
opencl-gpu:<br>
step 0 atom 1 x 3.940 5.612 2.226 force 1.48597e+03<br>
step 0 atom 2 x 3.852 5.659 2.211 force 6.26942e+02<br>
step 0 atom 3 x 3.979 5.665 2.303 force 8.44032e+02<br>
step 0 atom 4 x 3.992 5.610 2.139 force 7.92786e+02<br>
<br>
I am afraid there is something wrong in OpenCL kernels.<br>
<br>
I am using the topol-nvt-nogen.tpr I have uploaded on redmine.<br>
<br>
Best,<br>
Carlo<br>
<br>
<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>