Re: [gmx-users] Gromacs on GPU
I should add that while we had runs that completed, all we got were NaN for the coordinates and energies. Personally, I have attributed our inability to execute the GPU code cleanly to our C2050 Teslas [they are double precision]. (As I had the GPUized code working with an earlier card, although the options were pretty limited) Matt On Sat, Jan 28, 2012 at 11:32 AM, Benjamin Hall wrote: > Justin A. Lemkul wrote: > > > We get junk output with 4.5.4 if the box size is set to zero, for some odd > reason, though in principle it should be ignored with "pbc = no" in the > .mdp > file. Using any non-zero box results in usable results. Our OpenMM version > is > the same, but Cuda is in the 3.x series (can't remember off the top of my > head, > sorry). > > > I've tried setting the box size in the gro file but that doesn't seem to > work (100x100x100nm). Which compiler did you use? I realise that this is an > unlikely source of trouble but it's another variable to eliminate in > testing. > > I'll update issue 568 on redmine, which has been closed but appears to be > this problem. > > Thanks & best wishes > > Ben > > > -- > Dr Benjamin A Hall > Centre for Computational Science, Department of Chemistry, UCL > benjamin.a.h...@ucl.ac.uk > > > > > > -- > gmx-users mailing list gmx-users@gromacs.org > http://lists.gromacs.org/mailman/listinfo/gmx-users > Please search the archive at > http://www.gromacs.org/Support/Mailing_Lists/Search before posting! > Please don't post (un)subscribe requests to the list. Use the > www interface or send it to gmx-users-requ...@gromacs.org. > Can't post? Read http://www.gromacs.org/Support/Mailing_Lists -- gmx-users mailing listgmx-users@gromacs.org http://lists.gromacs.org/mailman/listinfo/gmx-users Please search the archive at http://www.gromacs.org/Support/Mailing_Lists/Search before posting! Please don't post (un)subscribe requests to the list. Use the www interface or send it to gmx-users-requ...@gromacs.org. Can't post? Read http://www.gromacs.org/Support/Mailing_Lists
Re: [gmx-users] Gromacs on GPU
Sorry for the delay, but I used gcc version 4.1.2 20080704 to compile our version of "GPU" enabled gromacs. Matt On Sat, Jan 28, 2012 at 11:32 AM, Benjamin Hall wrote: > Justin A. Lemkul wrote: > > > We get junk output with 4.5.4 if the box size is set to zero, for some odd > reason, though in principle it should be ignored with "pbc = no" in the > .mdp > file. Using any non-zero box results in usable results. Our OpenMM version > is > the same, but Cuda is in the 3.x series (can't remember off the top of my > head, > sorry). > > > I've tried setting the box size in the gro file but that doesn't seem to > work (100x100x100nm). Which compiler did you use? I realise that this is an > unlikely source of trouble but it's another variable to eliminate in > testing. > > I'll update issue 568 on redmine, which has been closed but appears to be > this problem. > > Thanks & best wishes > > Ben > > > -- > Dr Benjamin A Hall > Centre for Computational Science, Department of Chemistry, UCL > benjamin.a.h...@ucl.ac.uk > > > > > > -- > gmx-users mailing list gmx-users@gromacs.org > http://lists.gromacs.org/mailman/listinfo/gmx-users > Please search the archive at > http://www.gromacs.org/Support/Mailing_Lists/Search before posting! > Please don't post (un)subscribe requests to the list. Use the > www interface or send it to gmx-users-requ...@gromacs.org. > Can't post? Read http://www.gromacs.org/Support/Mailing_Lists -- gmx-users mailing listgmx-users@gromacs.org http://lists.gromacs.org/mailman/listinfo/gmx-users Please search the archive at http://www.gromacs.org/Support/Mailing_Lists/Search before posting! Please don't post (un)subscribe requests to the list. Use the www interface or send it to gmx-users-requ...@gromacs.org. Can't post? Read http://www.gromacs.org/Support/Mailing_Lists
Re: [gmx-users] Gromacs on GPU
Sorry to hear that, as I have had the same problem for over four months (OpenMM 3+, Gromacs 4.5.4+, Cuda 4, etc.). No one ever got back to me, and I just abandoned Gromacs for Amber (who's GPU accelerated code base compiles and works). I would be interested to hear if the intervening time, if a plan had been hatched to make the GPU code production quality. If that has indeed happened. Matthew On Fri, Jan 27, 2012 at 9:18 AM, Ben Hall wrote: > Hi > > I've been attempting to use the GPU enabled version of gromacs to run > implicit solvent simulations, but I've run into an odd problem. I have > downloaded and can run the benchmark simulations from the gromacs website, > and reproduce the speeds reported there. However, when I edit the MDP > files to write out data both the xtc and trr files contain junk data- the > first frame is correct but every coordinate in the following frames is the > same. These same MDP files work on the GPU disabled (default) gromacs and > produce reasonable trajectories. As the simulations aren't crashing I > assume that the issue relates to writing specifically; has anyone observed > the same problem, and is there a workaround or patch? > > The versions of each piece of software used are > gromacs 4.5.5 > openmm 3.1.1 > cuda 4.0 > and all compiled using intel compiler version 11.1/072 > > Thanks in advance > > Ben > > > -- > Dr Benjamin A Hall > Centre for Computational Science, Department of Chemistry, UCL > benjamin.a.h...@ucl.ac.uk > > > > > -- > gmx-users mailing list gmx-users@gromacs.org > http://lists.gromacs.org/mailman/listinfo/gmx-users > Please search the archive at > http://www.gromacs.org/Support/Mailing_Lists/Search before posting! > Please don't post (un)subscribe requests to the list. Use the > www interface or send it to gmx-users-requ...@gromacs.org. > Can't post? Read http://www.gromacs.org/Support/Mailing_Lists -- gmx-users mailing listgmx-users@gromacs.org http://lists.gromacs.org/mailman/listinfo/gmx-users Please search the archive at http://www.gromacs.org/Support/Mailing_Lists/Search before posting! Please don't post (un)subscribe requests to the list. Use the www interface or send it to gmx-users-requ...@gromacs.org. Can't post? Read http://www.gromacs.org/Support/Mailing_Lists
[gmx-users] Gromacs 4.6?
Hi, I am wondering when Gromacs will support OpenMM 3+ and my GPU architecture (S2050C Teslas)? I have been patiently waiting, and still haven't seen a new version (and yes, I have pulled back a development version from git). Still getting this nonsensical error from OpenMM2 and while OpenMM3 compiles I obviously get NaN's everywhere on the S2050C's since it is inappropriate to link them together. Am I doomed to use Amber forever? :) [Sorry, I just love Gromacs and have been waiting for four months for an update that would allow me to execute it on my rackmounted tesla blades...] Thanks in advance, Matthew -- gmx-users mailing listgmx-users@gromacs.org http://lists.gromacs.org/mailman/listinfo/gmx-users Please search the archive at http://www.gromacs.org/Support/Mailing_Lists/Search before posting! Please don't post (un)subscribe requests to the list. Use the www interface or send it to gmx-users-requ...@gromacs.org. Can't post? Read http://www.gromacs.org/Support/Mailing_Lists
[gmx-users] Tesla S2050 issues
Hi all, I have had no issues compiling OpenMM 2.0 and Gromacs 4.5.x (I have tried all flavors between 1 and 4), but once it comes time to execute mdrun I get the following error message: Error: invalid argument launching kernel kReduceForces The OpenMM group has been helpful in giving suggestions as to what to try, but I am at a loss now. I still get the same error. I have Cuda 3.1 on board and I have been able to compile both OpenMM 2 and OpenMM 3 without issue. I am also able to compile Gromacs, all recent flavors, linking in OpenMM without issue. I have also tried using the stock OpenMM libraries and have ended at the same place. Obviously I get a different error on exit when I link in OpenMM 3, Is it that my card is double precision? Thanks, Matt -- gmx-users mailing listgmx-users@gromacs.org http://lists.gromacs.org/mailman/listinfo/gmx-users Please search the archive at http://www.gromacs.org/Support/Mailing_Lists/Search before posting! Please don't post (un)subscribe requests to the list. Use the www interface or send it to gmx-users-requ...@gromacs.org. Can't post? Read http://www.gromacs.org/Support/Mailing_Lists
Re: [gmx-users] Re: gmx-users] Mopac7 / Gromacs 4.0.4 Issue
Gerrit, First, I realized that it didn't complete my thought about why the ./configure fails. ld is simply unable to find the -lmopac library. It seems that I have gotten around that by adding the mopac information into pkgconfig directory (by way of creating a .pc file and placing it into the /usr/lib/pkgconfig directory. Back to the real problem... So I have turned away from using the intel compilers, to eliminate that as a variable. Fortunately I have f2c already installed, so that isn't a problem. I modified the LIBS variable in the Makefile to match your line. Still I get the same error. I was hopeful that I would get Mopac to merge with Gromacs, as the Gaussian merger looked even more painful. But I think that I am done looking at getting Mopac in. Thanks for the suggestion! Matthew On Wed, Apr 8, 2009 at 11:45 PM, Gerrit Groenhof wrote: > Hi, > it should suffice to set the LIBS prior to compiling. > > With what compiler was mopac7 compiled? > I have thus far only been able to compile a working version with f2c/gcc. > > Not optimal. Using the intel fortran compiler, I also get a running mopac, > but the results are not correct and I don't know why. > > Anyway, after creating the library, I set LIBS '-lmopac -lf2c -lm', and > configure creates working makefiles. > > Hope this helps. > > Gerrit > > > Message: 3 > Date: Wed, 8 Apr 2009 16:24:34 -0700 > From: Matthew Lardy > Subject: [gmx-users] Mopac7 / Gromacs 4.0.4 Issues > To: gmx-users@gromacs.org > Message-ID: ><12741c010904081624s46a146d3kbf3519f600594...@mail.gmail.com> > Content-Type: text/plain; charset="iso-8859-1" > > Hi all, > I have been playing with Mopac7 (both version 7 and the more recent Mopac > 7-1.10 release). I have compiled both Mopac and > Gromacs independently without issue. But when I try to add > -with-qmmm-mopac > to my ./configure line, I am unable to compile and link them together. > > I have been looking at this for some time, and I have read the posts to > this > list. I have two issues that I think may be related. > > First, when I add the -lmopac to the LIBS environment variable, -DUSE_MOPAC > to CPPFLAGS and -L/(Path to Mopac Libs) to LDFLAGS, I don't make it out of > the configure alive: > > from config.log > ld: -lmopac > > So, I added the lmopac information in to pkgconfig... Once again no dice > with ./configure.So I then reset my environment and stepped through the > configure. Now after removing the references to mopac in my environment, > run the ./configure as I have done so previously, and then add the mopac > environment variables by hand, I get the following errors: > > icc -O3 -tpp6 -axK -ip -w -funroll-all-loops -o grompp grompp.o > -L/path/to/fftw-3.2.optim-thread/lib -L/vscrn/Packages/MopacOpt/lib > ./.libs/libgmxpreprocess.a ../mdlib/.libs/libmd.a > /path/to/gromacs-4.0.4/src/gmxlib/.libs/libgmx.a ../gmxlib/.libs/libgmx.a > -lnsl /path/to/fftw-3.2.optim-thread/lib/libfftw3f.a -lm -lmopac > /path/to/MopacOpt/lib/libmopac.a(lread.o)(.text+0x160): In function > `e_rsle': > : undefined reference to `__ctype_b' > /path/to/MopacOpt/lib/libmopac.a(lread.o)(.text+0x21f): In function > `e_rsle': > : undefined reference to `__ctype_b' > /path/to/MopacOpt/lib/libmopac.a(lread.o)(.text+0x296): In function > `e_rsle': > : undefined reference to `__ctype_b' > /path/to/MopacOpt/lib/libmopac.a(lread.o)(.text+0x2d0): In function > `e_rsle': > : undefined reference to `__ctype_b' > /path/to/MopacOpt/lib/libmopac.a(lread.o)(.text+0x8b2): In function > `e_rsle': > : undefined reference to `__ctype_b' > /path/to/MopacOpt/lib/libmopac.a(lread.o)(.text+0xb05): more undefined > references to `__ctype_b' follow > make[3]: *** [grompp] Error 1 > make[3]: Leaving directory `/path/to/gromacs-4.0.4/src/kernel' > make[2]: *** [all-recursive] Error 1 > make[2]: Leaving directory `/path/to/gromacs-4.0.4/src' > make[1]: *** [all] Error 2 > make[1]: Leaving directory `/path/to/gromacs-4.0.4/src' > make: *** [all-recursive] Error 1 > > I should add that I have downloaded and used Gerrit's mopac library, as > well > as patched a clean version of Mopac7. No dice. Oh yeah, I have tried the > GNU compilers as well. And, cleaned the application every time I have > tried > to rebuild. > > I figure that there is something simple that I have forgotten. > > Thanks in advance! > Matthew > > > ___ > gmx-users mailing listgmx-users@gromacs.org > http://www.gromacs.org/mailman/listinfo/gmx-users > Please search the archive at http://www.gromacs.org/search before p
[gmx-users] Mopac7 / Gromacs 4.0.4 Issues
Hi all, I have been playing with Mopac7 (both version 7 and the more recent Mopac 7-1.10 release). I have compiled both Mopac and Gromacs independently without issue. But when I try to add -with-qmmm-mopac to my ./configure line, I am unable to compile and link them together. I have been looking at this for some time, and I have read the posts to this list. I have two issues that I think may be related. First, when I add the -lmopac to the LIBS environment variable, -DUSE_MOPAC to CPPFLAGS and -L/(Path to Mopac Libs) to LDFLAGS, I don't make it out of the configure alive: from config.log ld: -lmopac So, I added the lmopac information in to pkgconfig... Once again no dice with ./configure.So I then reset my environment and stepped through the configure. Now after removing the references to mopac in my environment, run the ./configure as I have done so previously, and then add the mopac environment variables by hand, I get the following errors: icc -O3 -tpp6 -axK -ip -w -funroll-all-loops -o grompp grompp.o -L/path/to/fftw-3.2.optim-thread/lib -L/vscrn/Packages/MopacOpt/lib ./.libs/libgmxpreprocess.a ../mdlib/.libs/libmd.a /path/to/gromacs-4.0.4/src/gmxlib/.libs/libgmx.a ../gmxlib/.libs/libgmx.a -lnsl /path/to/fftw-3.2.optim-thread/lib/libfftw3f.a -lm -lmopac /path/to/MopacOpt/lib/libmopac.a(lread.o)(.text+0x160): In function `e_rsle': : undefined reference to `__ctype_b' /path/to/MopacOpt/lib/libmopac.a(lread.o)(.text+0x21f): In function `e_rsle': : undefined reference to `__ctype_b' /path/to/MopacOpt/lib/libmopac.a(lread.o)(.text+0x296): In function `e_rsle': : undefined reference to `__ctype_b' /path/to/MopacOpt/lib/libmopac.a(lread.o)(.text+0x2d0): In function `e_rsle': : undefined reference to `__ctype_b' /path/to/MopacOpt/lib/libmopac.a(lread.o)(.text+0x8b2): In function `e_rsle': : undefined reference to `__ctype_b' /path/to/MopacOpt/lib/libmopac.a(lread.o)(.text+0xb05): more undefined references to `__ctype_b' follow make[3]: *** [grompp] Error 1 make[3]: Leaving directory `/path/to/gromacs-4.0.4/src/kernel' make[2]: *** [all-recursive] Error 1 make[2]: Leaving directory `/path/to/gromacs-4.0.4/src' make[1]: *** [all] Error 2 make[1]: Leaving directory `/path/to/gromacs-4.0.4/src' make: *** [all-recursive] Error 1 I should add that I have downloaded and used Gerrit's mopac library, as well as patched a clean version of Mopac7. No dice. Oh yeah, I have tried the GNU compilers as well. And, cleaned the application every time I have tried to rebuild. I figure that there is something simple that I have forgotten. Thanks in advance! Matthew ___ gmx-users mailing listgmx-users@gromacs.org http://www.gromacs.org/mailman/listinfo/gmx-users Please search the archive at http://www.gromacs.org/search before posting! Please don't post (un)subscribe requests to the list. Use the www interface or send it to gmx-users-requ...@gromacs.org. Can't post? Read http://www.gromacs.org/mailing_lists/users.php
