[gmx-developers] 2018-beta1: Performance of membrane simulations

Jochen Hub jhub at gwdg.de
Fri Dec 1 23:20:43 CET 2017


Hi Magnus,

many thanks, and impressive, that clarifies my question.

Since UB has such as drastic effect on performance, maybe you can 
convince the other developers to make an exception and get a UB-SIMD 
patch into 2018?

I understand, normally only bug fixes after beta release, but is 33% 
performance loss (50% gain) not close to a bug?

Cheers,
Jochen

Am 01.12.17 um 22:54 schrieb Magnus Lundborg:
> Hi,
> 
> I'm running simulations with the CHARMM forcefield, which also uses UB 
> and experienced similar things. Apparently the flops count in the first 
> table is not the actual time for the calculations, if I understood the 
> explanations correctly. So it's the Force row in the second table that's 
> bonded forces (with long range and PME on GPU). So I tried making a SIMD 
> version of UB (only standard angles are SIMD optimised) and got almost a 
> 50% performance gain. Making also bonds using SIMD only have an 
> additional 1 or 2%. My patch is just a draft as it's not clear what 
> future SIMD functions should look like, but ill share it with you so 
> that you can try it. However, it won't be in the next release, I guess.
> 
> Cheers,
> 
> Magnus
> 
> 
> Den 1 dec. 2017 22:34 skrev "Jochen Hub" <jhub at gwdg.de 
> <mailto:jhub at gwdg.de>>:
> 
>     Dear developers,
> 
>     I started a thread in the user list yesterday (and Szilard already
>     gave a quick answer) but I felt this point is relevant for the
>     developers list.
> 
>     We did some benchmarks with the 2018-beta1 with PME on the GPU -
>     overall fantastic (!!) - we just don't understand the performance of
>     lipid membrane simulations (Slipids or Charmm36, with UB
>     potentials). They contain roughly 50% lipid, 50% water atoms. Please
>     see here:
> 
>     http://cmb.bio.uni-goettingen.de/bench.pdf
>     <http://cmb.bio.uni-goettingen.de/bench.pdf>
> 
>     As you see in the linked PDF, the Slipid simulations are limited by
>     the CPU up to 10 (!) quite strong Xeon cores, when using a GTX 1080.
>     Szilard pointed out that is is probably due to bonded UB
>     interactions - however, they make only 0.2% of the Flops, see the
>     log output pasted below, for ntomp=4 or 10 (for 128 Slipids system
>     with 1nm cutoff). The Flops-Summary is nearly the same for ntomp=4
>     or 10, so only the ntomp=4 is shown below.
> 
>     In contrast, protein simulations (whether membrane protein or purely
>     in water) behave as one hopes, showing that we can buy a cheap CPU
>     when doing PME on the GPU.
> 
>     So my question is: Is this expected? Is this really due to
>     Urey-Bradley? Or maybe due to Constraints? In case that UB is
>     limiting, are there any plans to port this also onto the GPU in the
>     future?
> 
>     This has also impact on hardware: Depending on whether you run
>     protein or membrane simulation, you need to buy different hardware.
> 
>     Many thanks for any input, and many thanks again for the fabulous
>     work on 2018!
> 
>     Jochen
> 
> 
>       Computing:                               M-Number         M-Flops 
>     % Flops
>     -----------------------------------------------------------------------------
>       Pair Search distance check             151.929968        1367.370 
>         0.0
>       NxN Ewald Elec. + LJ [F]            157598.160192    10401478.573 
>        97.2
>       NxN Ewald Elec. + LJ [V&F]            1623.781504      173744.621 
>         1.6
>       1,4 nonbonded interactions             200.360064       18032.406 
>         0.2
>       Shift-X                                  1.553664           9.322 
>         0.0
>       Propers                                246.449280       56436.885 
>         0.5
>       Impropers                                1.280256         266.293 
>         0.0
>       Virial                                   7.657759         137.840 
>         0.0
>       Stop-CM                                  1.553664          15.537 
>         0.0
>       P-Coupling                               7.646464          45.879 
>         0.0
>       Calc-Ekin                               15.262464         412.087 
>         0.0
>       Lincs                                   74.894976        4493.699 
>         0.0
>       Lincs-Mat                             1736.027136        6944.109 
>         0.1
>       Constraint-V                           226.605312        1812.842 
>         0.0
>       Constraint-Vir                           7.614336         182.744 
>         0.0
>       Settle                                  25.605120        8270.454 
>         0.1
>       Urey-Bradley                           144.668928       26474.414 
>         0.2
>     -----------------------------------------------------------------------------
>       Total                                                10700125.072 
>       100.0
>     -----------------------------------------------------------------------------
> 
> 
>           R E A L   C Y C L E   A N D   T I M E   A C C O U N T I N G
> 
>     On 1 MPI rank, each using 4 OpenMP threads
> 
>       Computing:          Num   Num      Call    Wall time       
>       Giga-Cycles
>                           Ranks Threads  Count      (s)         total
>     sum    %
>     -----------------------------------------------------------------------------
>       Neighbor search        1    4         51       0.260         
>     2.284   2.4
>       Launch GPU ops.        1    4      10002       0.591         
>     5.191   5.4
>       Force                  1    4       5001       7.314       
>       64.211  67.1
>       Wait PME GPU gather    1    4       5001       0.071         
>     0.626   0.7
>       Reduce GPU PME F       1    4       5001       0.078         
>     0.684   0.7
>       Wait GPU NB local      1    4       5001       0.017         
>     0.151   0.2
>       NB X/F buffer ops.     1    4       9951       0.321         
>     2.822   2.9
>       Write traj.            1    4          2       0.117         
>     1.026   1.1
>       Update                 1    4       5001       0.199         
>     1.749   1.8
>       Constraints            1    4       5001       1.853       
>       16.270  17.0
>       Rest                                           0.085         
>     0.743   0.8
>     -----------------------------------------------------------------------------
>       Total                                         10.907       
>       95.757 100.0
>     -----------------------------------------------------------------------------
> 
>     ********************************
>     ****** 10 Open MP threads ******
>     ********************************
> 
>     On 1 MPI rank, each using 10 OpenMP threads
> 
>       Computing:          Num   Num      Call    Wall time       
>       Giga-Cycles
>                           Ranks Threads  Count      (s)         total
>     sum    %
>     -----------------------------------------------------------------------------
>       Neighbor search        1   10         51       0.120         
>     2.625   2.3
>       Launch GPU ops.        1   10      10002       0.580       
>       12.731  11.3
>       Force                  1   10       5001       2.999       
>       65.828  58.4
>       Wait PME GPU gather    1   10       5001       0.066         
>     1.459   1.3
>       Reduce GPU PME F       1   10       5001       0.045         
>     0.980   0.9
>       Wait GPU NB local      1   10       5001       0.014         
>     0.308   0.3
>       NB X/F buffer ops.     1   10       9951       0.157         
>     3.453   3.1
>       Write traj.            1   10          2       0.147         
>     3.224   2.9
>       Update                 1   10       5001       0.140         
>     3.067   2.7
>       Constraints            1   10       5001       0.814       
>       17.867  15.9
>       Rest                                           0.053         
>     1.161   1.0
>     -----------------------------------------------------------------------------
>       Total                                          5.135       
>     112.703 100.0
>     -----------------------------------------------------------------------------
> 
> 
> 
>     -- 
>     ---------------------------------------------------
>     Dr. Jochen Hub
>     Computational Molecular Biophysics Group
>     Institute for Microbiology and Genetics
>     Georg-August-University of Göttingen
>     Justus-von-Liebig-Weg 11, 37077 Göttingen, Germany
>     <https://maps.google.com/?q=Justus-von-Liebig-Weg+11,+37077+G%C3%B6ttingen,+Germany&entry=gmail&source=g>.
>     Phone: +49-551-39-14189 <tel:%2B49-551-39-14189>
>     http://cmb.bio.uni-goettingen.de/ <http://cmb.bio.uni-goettingen.de/>
>     ---------------------------------------------------
>     -- 
>     Gromacs Developers mailing list
> 
>     * Please search the archive at
>     http://www.gromacs.org/Support/Mailing_Lists/GMX-developers_List
>     <http://www.gromacs.org/Support/Mailing_Lists/GMX-developers_List>
>     before posting!
> 
>     * Can't post? Read http://www.gromacs.org/Support/Mailing_Lists
>     <http://www.gromacs.org/Support/Mailing_Lists>
> 
>     * For (un)subscribe requests visit
>     https://maillist.sys.kth.se/mailman/listinfo/gromacs.org_gmx-developers
>     <https://maillist.sys.kth.se/mailman/listinfo/gromacs.org_gmx-developers>
>     or send a mail to gmx-developers-request at gromacs.org
>     <mailto:gmx-developers-request at gromacs.org>.
> 
> 
> 
> 

-- 
---------------------------------------------------
Dr. Jochen Hub
Computational Molecular Biophysics Group
Institute for Microbiology and Genetics
Georg-August-University of Göttingen
Justus-von-Liebig-Weg 11, 37077 Göttingen, Germany.
Phone: +49-551-39-14189
http://cmb.bio.uni-goettingen.de/
---------------------------------------------------


More information about the gromacs.org_gmx-developers mailing list