Re: [AMBER-Developers] near-final testing

From: MengJuei Hsieh <mengjueh.uci.edu>
Date: Wed, 14 Apr 2010 00:49:24 -0700

Hi,

Jun and I found that it's most likely caused by an uninitialized array
inside the pb_lsolver code. Here I attached a bug fix that should eliminate
the NaNs in the second printout.

Best,
-- 
Mengjuei Hsieh, Molecular Biology & Biochem, Univ. of California Irvine.
> From: MengJuei Hsieh <mengjueh.uci.edu>
> Date: Tue, 13 Apr 2010 12:41:06 -0700
> To: AMBER Developers Mailing List <amber-developers.ambermd.org>, Scott
> Brozell <sbrozell.rci.rutgers.edu>, Jun Wang <jwang20.uci.edu>
> Subject: Re: [AMBER-Developers] near-final testing
> 
> Hi,
> 
> It turns out the pgi version I was working on is too old (version 8) so
> the solution of changing pb_force.o's optimization flag apparently doesn't
> work. I just got access to teragrid's pgi 9, 10 (32,64). I indeed see an
> MG error that spills out NaNs in sander_pbsa_lpb/mdout.min_bc5_solv2. (
> version 10, 64bit. Luckily it did finish the first pbsa computation,
> although failed at the second time.)
> #  Maximum number of minimization cycles reached.
> # MG FAILED: ncyc, itn, norm, onorm           10           10
> #                       NaN   9.9000000000000007E+099
> (testing on PGI 9 is still ongoing)
> 
> Jun, can you take a quick look at the multigrid solver to see if we can
> have a quick fix before tomorrow?
> 
> Best,
> -- 
> Mengjuei Hsieh, Luo Research Group, Molecular Biology & Biochemistry,
> University of California Irvine.
> 
> On Mon, Apr 12, 2010 at 05:33:22PM -0700, MengJuei Hsieh wrote:
>> Dear Scott,
>> I do have access to pgf90 (8.0-4 32-bit target on x86 Linux -tp k8-32), does
>> that count? Anyway, I'll look into it tonight.
>> 
>>> From: Scott Brozell <sbrozell.rci.rutgers.edu>
>>> Date: Mon, 12 Apr 2010 18:16:46 -0400
>>> Hi,
>>> Note that the patches from this thread:
>>>> Date: Wed, 7 Apr 2010 15:58:39 -0700
>>>> From: Mengjuei Hsieh <mengjueh.uci.edu>
>>> did not fix this pgi 9 issue in serial:
>>> possible FAILURE:  check mdout.min_bc5_solv2.dif
>>> /tmp/pbstmp.3024735/amber11/test/sander_pbsa_lpb
>>> 68c68
>>> <       1       6.0437E+2     2.4119E+2     8.2853E+2     C2          1
>>> ---
>>>>       1             NaN     2.4119E+2     0.     C2          1
>>> 70c70
>>> <  VDWAALS =       -0.5062  EEL     =      -48.9226  EPB        =
>>> -90.3631
>>> ---
>>>>  VDWAALS =       -0.5062  EEL     =      -48.9226  EPB        =
>>>> NaN
>>> Mengjuei, do you have access to a pgi compiler ?
>>> thanks,
>>> Scott



_______________________________________________
AMBER-Developers mailing list
AMBER-Developers.ambermd.org
http://lists.ambermd.org/mailman/listinfo/amber-developers




Received on Wed Apr 14 2010 - 01:00:03 PDT
Custom Search