[AMBER-Developers] Re: Experimental Ptraj Code - To Speed Up Trajectory Reading

From: Eric Absgarten <eabsgarten.gmail.com>
Date: Thu, 5 Mar 2009 19:49:22 -0500


I recoded something whereby checks Coordinates initially.
And I also changed the earlier flag from -DEXPHO to -DEXPHT (High
Through-put vs. High Output).

The intent is to make processing of amber trajectories faster and eventually
process things in real-time for finer grained analysis.

A lot of the experimental code is geared towards improving the overall speed
at which it does things, some changes are finer grained than others,
especially code currently turned on with the -DEXPHT

The new flags are:
-DEXPHT (used to be -DEXPHO)
and the new one

Some timings on 5343 atom, 27916 frame system
You can think of Real as actual time (including IO waits), User as the time
spent running as your userid, and Sys as the time spent in kernel functions.

orig ptraj frame+ht code frame+ht+chktrj

real 8m9.502s real 4m13.772s real 3m3.503s
user 6m47.909s user 2m58.931s user 2m48.002s
sys 0m22.329s sys 0m21.893s sys 0m12.541s

This code is experimental so please be aware of that. Compare the results to
original ptraj results and let me how it goes.

Appreciate all feedback, thanks,
AMBER-Developers mailing list
Received on Fri Mar 06 2009 - 01:26:28 PST
Custom Search