Speed-up depends what you measure against, what precision GROMACS is using, and what your AVX hardware is. There is no single number worth mentioning. GROMACS has different AVX kernels for single vs double, and Intel Sandy Bridge / Ivy Bridge vs AMD's Opteron / Bulldozer. I haven't seen any numbers because the author of the kernels hasn't produced them yet, but they will be considerably faster than anything you might compare them with.
Mark On Fri, Dec 7, 2012 at 4:37 AM, Yorquant Wang <[email protected]> wrote: > Hi Roland: > Based on your test, how many speed-up you get? > Yorquant > > 2012/12/7 Roland Schulz <[email protected]> > > > On Thu, Dec 6, 2012 at 9:42 PM, Yorquant Wang <[email protected]> > wrote: > > > > > Hi Mark: > > > There is a new instruction set architecture (*Advanced Vector > > > Extensions > > > * (AVX)) that can make float calculation in Intel CPU faster two times > > > compared with the old instruction set. I want to know if the Gromacs > > > developers have a plan to make GMX support AVX. if the GMX can support > > GMX, > > > I think maybe two times speed-up in GMX will be gotten immediately. > > > > > Yes AVX is supported. But you won't see a 2x speedup. > > > > Roland > > > > > > > > > > Best > > > > > > Yukun Wang > > > PhD candidate > > > Institute of Natural Sciences && College of Life Science, Shanghai Jiao > > > Tong University > > > Cell phone: 13621806236. > > > China Shanghai > > > > > > > > > 2012/12/7 Mark Abraham <[email protected]> > > > > > > > Hi all, > > > > > > > > We've updated the GROMACS beta version to fix some bugs both you and > we > > > > found. We've also added the Adaptive resolution scheme (adResS) to > our > > > list > > > > of new features (though we still have yet to publish a complete list > of > > > > those!). adResS couples two systems with different resolutions by a > > force > > > > interpolation, which can be used to speed-up atomistic simulations. > The > > > new > > > > source package can be found at > > > > ftp://ftp.gromacs.org/pub/gromacs/gromacs-4.6-beta2.tar.gz. > > Installation > > > > instructions still here > > > > http://www.gromacs.org/Documentation/Installation_Instructions > > > > > > > > Please try it out, particularly if you haven't done so already! Also, > > if > > > > everything is smooth sailing, please drop us a line on gmx-users just > > to > > > > say that. We can't tell whether silence is "worked great, nothing to > > say" > > > > or "haven't tried it yet". That will help us judge when things are > > stable > > > > enough to make a real release! > > > > > > > > Speaking of that, we are keen to make that final release soon. While > we > > > > can't pick a date yet, we promise that if you give us feedback by > > > December > > > > 21, then we will make a sincere effort to incorporate the results of > > that > > > > feedback in the final release. In particular, if it's an issue on our > > > > Redmine bug report database http://redmine.gromacs.org, then it will > > be > > > > sure to get our attention and consideration. You'll need to register > an > > > > account to make a bug report (so that we can get back to you), but > > that's > > > > free and easy. > > > > > > > > We hope to release a current version of our regression test suite > next > > > > week, and a benchmark set soon. > > > > > > > > Cheers, > > > > > > > > Mark Abraham > > > > GROMACS development manager > > > > -- > > > > gmx-users mailing list [email protected] > > > > 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 [email protected]. > > > > * Can't post? Read http://www.gromacs.org/Support/Mailing_Lists > > > > > > > > > > > > > > > > -- > > > Yukun Wang > > > PhD candidate > > > Institute of Natural Sciences && College of Life Science, Shanghai Jiao > > > Tong University > > > Cell phone: 13621806236. > > > China Shanghai > > > -- > > > gmx-users mailing list [email protected] > > > 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 [email protected]. > > > * Can't post? Read http://www.gromacs.org/Support/Mailing_Lists > > > > > > > > > > > > > > > > > > > > > -- > > ORNL/UT Center for Molecular Biophysics cmb.ornl.gov > > 865-241-1537, ORNL PO BOX 2008 MS6309 > > -- > > gmx-users mailing list [email protected] > > 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 [email protected]. > > * Can't post? Read http://www.gromacs.org/Support/Mailing_Lists > > > > > > -- > Yukun Wang > PhD candidate > Institute of Natural Sciences && College of Life Science, Shanghai Jiao > Tong University > Cell phone: 13621806236. > China Shanghai > -- > gmx-users mailing list [email protected] > 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 [email protected]. > * Can't post? Read http://www.gromacs.org/Support/Mailing_Lists > -- gmx-users mailing list [email protected] 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 [email protected]. * Can't post? Read http://www.gromacs.org/Support/Mailing_Lists
