Yes Daniel, two very good points. With the good branch prediction available these days on any chips that anyone is likely to be using, one can set up a bunch of parameter variables that are set at initialization and never change. One then has basically zero branching cost after the first pass. And a test case for 10_12 would have made introduction of the current bug in pmemd much less likely.
- Bob
________________________________________
From: Daniel Roe [daniel.r.roe.gmail.com]
Sent: Wednesday, October 26, 2011 11:35 AM
To: AMBER Developers Mailing List
Subject: Re: [AMBER-Developers] (long-ish) saga about the HAS_10_12 option in sander and pmemd
Hi,
I second the notion that the use of ifdef statements should be
minimized. I think in many cases ifdefs could be replaced with normal
if-then statements and/or separate function calls without too much
performance loss. The gain in readability is probably worth it,
although it will be a ton of work (over 1000 ifdefs in sander alone).
Also, I think that once all of the 10_12 code is sorted out and
working properly, a test case should be added that could catch this
issue.
-Dan
--
-------------------------
Daniel R. Roe, PhD
Postdoctoral Associate
BioMaPS Institute, Rutgers University
610 Taylor Road
Piscataway, NJ 08854
_______________________________________________
AMBER-Developers mailing list
AMBER-Developers.ambermd.org
http://lists.ambermd.org/mailman/listinfo/amber-developers
_______________________________________________
AMBER-Developers mailing list
AMBER-Developers.ambermd.org
http://lists.ambermd.org/mailman/listinfo/amber-developers
Received on Wed Oct 26 2011 - 12:00:05 PDT
