Re: [gmx-users] Log output of GPU accelerated GROMACS

2015-01-26 Thread Mark Abraham
No, it stems from a CUDA limitation, where if two+ ranks share a GPU then
mdrun can't get clean timing numbers.

Mark
On Jan 8, 2015 3:10 PM, Ebert Maximilian m.eb...@umontreal.ca wrote:

 Dear list,

 I was wondering why the log file does not always contains the same
 information. In one of my configurations I got the following information:

 GPU timings

 -
  Computing: Count  Wall t (s)  ms/step   %

 -
  Pair list H2D   1251   0.5630.450 0.2
  X / q H2D  50001   6.9980.140 2.9
  Nonbonded F kernel 47500 212.1184.46686.7
  Nonbonded F+ene k.  1250   9.3717.497 3.8
  Nonbonded F+ene+prune k.1251   9.7597.801 4.0
  F D2H  50001   5.8740.117 2.4

 -
  Total244.6834.894   100.0

 -

 Force evaluation time GPU/CPU: 4.894 ms/4.012 ms = 1.220
 For optimal performance this ratio should be close to 1!

 But I never got this information in any other configuration using GPUs. Is
 this only part of the output if there is a problem with too much work for
 the GPU?

 Thank you very much,

 Max
 --
 Gromacs Users mailing list

 * Please search the archive at
 http://www.gromacs.org/Support/Mailing_Lists/GMX-Users_List before
 posting!

 * Can't post? Read http://www.gromacs.org/Support/Mailing_Lists

 * For (un)subscribe requests visit
 https://maillist.sys.kth.se/mailman/listinfo/gromacs.org_gmx-users or
 send a mail to gmx-users-requ...@gromacs.org.

-- 
Gromacs Users mailing list

* Please search the archive at 
http://www.gromacs.org/Support/Mailing_Lists/GMX-Users_List before posting!

* Can't post? Read http://www.gromacs.org/Support/Mailing_Lists

* For (un)subscribe requests visit
https://maillist.sys.kth.se/mailman/listinfo/gromacs.org_gmx-users or send a 
mail to gmx-users-requ...@gromacs.org.


[gmx-users] Log output of GPU accelerated GROMACS

2015-01-08 Thread Ebert Maximilian
Dear list,

I was wondering why the log file does not always contains the same information. 
In one of my configurations I got the following information:

GPU timings
-
 Computing: Count  Wall t (s)  ms/step   %
-
 Pair list H2D   1251   0.5630.450 0.2
 X / q H2D  50001   6.9980.140 2.9
 Nonbonded F kernel 47500 212.1184.46686.7
 Nonbonded F+ene k.  1250   9.3717.497 3.8
 Nonbonded F+ene+prune k.1251   9.7597.801 4.0
 F D2H  50001   5.8740.117 2.4
-
 Total244.6834.894   100.0
-

Force evaluation time GPU/CPU: 4.894 ms/4.012 ms = 1.220
For optimal performance this ratio should be close to 1!

But I never got this information in any other configuration using GPUs. Is this 
only part of the output if there is a problem with too much work for the GPU?

Thank you very much,

Max
-- 
Gromacs Users mailing list

* Please search the archive at 
http://www.gromacs.org/Support/Mailing_Lists/GMX-Users_List before posting!

* Can't post? Read http://www.gromacs.org/Support/Mailing_Lists

* For (un)subscribe requests visit
https://maillist.sys.kth.se/mailman/listinfo/gromacs.org_gmx-users or send a 
mail to gmx-users-requ...@gromacs.org.