amber-developers: REM and PME

From: Carlos Simmerling <>
Date: Mon, 8 Mar 2004 15:24:35 -0700

we have a fix for the REM/PME bug, we'll be in touch
with Scott to get the tree updated. What happened was that
a flag was set telling sander that some initialization had
been done, but the data itself wasn't saved after returning
to multisander. However, the flag stating that the init had been done
was saved by the subroutine (and not reinitialized in
subsequent sander calls because it was done in a DATA statement).

everyone may need to change their programming habits
a little to accomodate REM, since assumptions about
how an MD run will work are not always valid anymore.
in particular, sander is now a subroutine, so using data statements
rather than explicit code to initialize values can be dangerous,
especially if it's something that should be done each time
the sander subroutine is called. For now, calling sander
twice in a row (without exiting multisander) gives the same
data as a single MD run that is twice as long, but it will help
if we all keep that goal in mind while writing new code.

comments on better ways to handle this are welcome, of course.

Received on Wed Apr 05 2006 - 23:50:04 PDT
Custom Search