Project

General

Profile

Bug #929

likely free energy bug in gather_f_bspline

Added by Michael Shirts about 7 years ago. Updated about 7 years ago.

Status:
Closed
Priority:
High
Category:
-
Target version:
Affected version - extra info:
4.6
Affected version:
Difficulty:
uncategorized
Close

Description

on line 2206 of pme.c, the line

qn = atc->q[n];

should probably be

qn  = scale*atc->q[n];

It's possible this is compensated for somewhere else in the code that I'm missing -- if so, then scale should be removed here from this function since it's not used. I suspect that the free energy dependence is just being removed, causing both a and b states to always be on.

If this is a bug, my current plan is just to fix it when I check in the free energy code for 4.6.

Associated revisions

Revision 726f688f (diff)
Added by Berk Hess about 7 years ago

fixed PME forces with FE, fixes #929 introduced in recent merge

Change-Id: I4f3bc59ed8c0351c4c59feceb3c2116b2f063daf

History

#1 Updated by Berk Hess about 7 years ago

This is a leftover from some charge scaling implementation, scale should go out.

#2 Updated by Michael Shirts about 7 years ago

Where is the scaling done then? Right now, it appears to not be done at all, i.e. when I run, all charges are contributing full forces whether it's in state A or B.

#3 Updated by Michael Shirts about 7 years ago

Specifically, flr in sum_forces contains full charge forces from both state A and state B, and this was adjusted by the 'scale' line above.

#4 Updated by Berk Hess about 7 years ago

  • Status changed from New to Closed
  • Target version set to 4.6
  • Affected version - extra info set to 4.6

Ah, I should have looked better. Spread doesn't need the scale, but gather does. This got removed as I was thinking of rewriting the code in a different way in the PME threading branch. I put the scale back.

Also available in: Atom PDF