Excessive runtime spent in pair search on 18240

[arisu]

This may be nothing, but I thought I'd report it anyway in case it could be useful information. I do know how to use GROMACS, but this may be something that is expected because this project is testing out experimental force fields. Anyway, please forgive me if I am just making noise.

Almost all WUs with project 18240 spend over 11% of their computation time in NS. The cutoff-scheme for this project is verlet and not group so increasing nstlist can improve performance with no affect on accuracy. The log files also indicates this, saying "you might want to increase nstlist (this has no effect on accuracy)". The amount of time spent in NS for 18240 WUs is significantly higher than any other project I have recently run:

Code: Select all

  Project       Run     Clone   Gen     NS%
  P12462        R1      C75     G163    2.2
  P12466        R0      C19     G115    4.7
  P12467        R11     C5      G58     6.1
  P12467        R13     C13     G72     6.1
  P12468        R11     C47     G94     5.1
  P12468        R7      C2      G133    4.8
  P12469        R14     C19     G69     6.0
  P12469        R20     C29     G78     6.2
  P12480        R22     C4      G34     4.9
  P12480        R50     C2      G52     4.9
  P12481        R25     C10     G40     6.0
  P12489        R2      C30     G22     4.4
  P12489        R9      C41     G8      4.4
* P18240        R1023   C3      G154    11.1
* P18240        R102    C2      G154    11.7
* P18240        R1033   C0      G191    10.9
* P18240        R17     C4      G156    11.8
* P18240        R29     C1      G174    11.8
* P18240        R31     C0      G154    11.0
* P18240        R398    C4      G120    11.9
* P18240        R488    C3      G149    11.3
* P18240        R909    C4      G116    11.8
* P18240        R970    C0      G207    11.6
* P18240        R980    C0      G166    11.8
* P18240        R999    C3      G141    10.7
  P18494        R12     C12     G168    2.9
  P18494        R15     C2      G93     3.1
  P18494        R19     C16     G94     2.9
  P18803        R3      C106    G119    3.1
  P18803        R3      C148    G178    2.8
  P18803        R4      C247    G154    3.2
  P18803        R4      C247    G155    3.1
  P18803        R5      C156    G154    3.1
  P18803        R5      C50     G159    3.0
  P18803        R5      C50     G160    3.2
  P19227        R2000   C7      G5      0.4
  P19227        R2308   C7      G6      0.5
  P19227        R3594   C7      G6      0.4
  P19228        R4685   C7      G1      0.2
  P19228        R4850   C7      G2      0.4
  P19229        R1644   C7      G6      0.3
  P19229        R1807   C7      G3      0.3
  P19229        R2275   C7      G2      0.4
  P19229        R3775   C7      G1      0.4

[jjmiller]

Thank you!

[vvoelz]

Hi @arisu --

Interesting observation! The

Code: Select all

nstlist

parameter is something the project owner could indeed tweak if it were useful.

According to https://stats.foldingathome.org/project/18240, this project is owned by Justin Miller. I'll alert him to your post.

Vince