Project

General

Profile

Bug #1293

energygrps issue with cuda runs

Added by Dimitris Dellis about 7 years ago. Updated about 6 years ago.

Status:
Closed
Priority:
Normal
Assignee:
-
Category:
mdrun
Target version:
-
Affected version - extra info:
cuda
Affected version:
Difficulty:
uncategorized
Close

Description

If I ask for energygrps=More_than_one_component with 2+ components systems using cuda, all energies are added in the first specified component.
All the rest are reported as zeroes. Swaping order in energygrps swaps energies.
The total potential is correct.

mdp:
energygrps=A B

log and edr :
Epot Coul-SR LJ-SR Coul-14 LJ-14
A-A Sumofall Sumofall Sumofall sumofall
A-B 0 0 0 0
B-B 0 0 0 0

Using cpu only everything is reported correctly.
Can be verified with any 2+ components system with cuda.


Related issues

Related to GROMACS - Bug #1822: mdrun writes broken energy group values to .edr fileClosed
Has duplicate GROMACS - Bug #1727: g_energy computes zero energy values with GPU Closed05/06/2015

History

#1 Updated by Mark Abraham about 7 years ago

  • Assignee deleted (Dimitris Dellis)

I suspect this is a feature. Was there no message (.log, stdout) that reporting of a group-wise energy decomposition would not occur?

#2 Updated by Dimitris Dellis about 7 years ago

Yes, there is a note early in log.
Probably a note above/below the printed decomposition near the end of log is a good idea for end users.
Ticket may close.

#3 Updated by Erik Lindahl about 6 years ago

  • Status changed from New to Closed

#4 Updated by Justin Lemkul over 5 years ago

  • Has duplicate Bug #1727: g_energy computes zero energy values with GPU added

#5 Updated by Mark Abraham almost 5 years ago

  • Related to Bug #1822: mdrun writes broken energy group values to .edr file added

Also available in: Atom PDF