AMBER Archive (2005)

Subject: Re: AMBER: leap single precision

From: Scott Brozell (sbrozell_at_scripps.edu)
Date: Fri Apr 08 2005 - 02:07:31 CDT


Hi,

On Thu, 7 Apr 2005, Guanglei Cui wrote:

> Hi, all
>
> Has anyone created a single precision version of leap? I have a large

I have not.

> system to work on. The double precision leap couldn't handle it with a
> 4GB memory machine. I managed to compile leap with all "double" replaced
> with "float".

Did you try a picky compiler and crank up the warning level ?
This might help some; in basics.h
#define MAXSTRINGLENGTH 1069
Even though 1069 is
the first four-digit quasiall-even-digits non-quasi-repdigit emirp,
something like 83 may be big enough (although there is not thorough
checking of string lengths).

> Leap runs, but it seems it is having trouble reading
> frcmod files. I put back the original fortran.[ch], but it didn't solve
> the problem. Can someone show me how to safely turn leap into single
> precision? Thanks in advance.

Maybe you could use double for the frcmod reading then later cast it
down to float; probably mesy to do inside leap. What kind of trouble ?
I dont understand 'original fortran.[ch]'.

Perhaps you can use a big shared memory machine ?
Naturally, you are using tleap; maybe there are some leaplib things
that you don't need ?

Good luck,
Scott

-----------------------------------------------------------------------
The AMBER Mail Reflector
To post, send mail to amber_at_scripps.edu
To unsubscribe, send "unsubscribe amber" to majordomo_at_scripps.edu