[ADMB Users] funnel variable for vectors

Mark Maunder mmaunder at iattc.org
Sat May 29 22:09:10 PDT 2010

Does anyone know if there is a funnel variable for vectors?




Mark Maunder 

Head of the Stock Assessment Program

Inter-American  Tropical Tuna Commission
8604 La Jolla Shores Drive
La Jolla, CA, 92037-1508, USA
Tel: (858) 546-7027
Fax: (858) 546-7133
mmaunder at iattc.org


Visit the AD Model Builder project at
 http://admb-project.org/ <http://admb-project.org/> 


See the following website for information on fisheries stock assessment



From: users-bounces at admb-project.org
[mailto:users-bounces at admb-project.org] On Behalf Of Ian Taylor
Sent: Friday, May 28, 2010 4:34 PM
To: users at admb-project.org
Subject: [ADMB Users] memory inputs in ADMB


Memory use seems to the topic of the day, as I was just starting to
write this email when I saw that the topic was being discussed on the
developers list:
Arni's suggestion of having ADMB tell us how much memory is needed by a
given model might one day solve avoid the problems I'm asking about.

My first question is about the warning 
"Memory allocation error in grad_stack constructor
 trying to allocate 2000000000 bytes"
Is this warning (which seem to come from this piece
radstak.cpp>  of the source code), just telling me that my 32 bit
Windows XP computer can't handle the amount of memory (slightly below
2GB) that I'm trying to allocate using the -gbs input? I've got 4GB
available, but maybe 32 bit Windows can't handle more than 2GB for any
one program? A smaller input allows the model to run fine without
warning, but I'd still like to understand the source of the warning.

My second question is what happens when you enter a value for -cbs or
-gbs that is bigger than 2147483648 (2GB)? It appears that instead of
getting a warning, the model runs, but slower, and using less memory
than what would be associated with a smaller -gbs input. Since the
current method of finding good values for these inputs seems to be trial
and error, perhaps ADMB could help us avoid searching beyond the maximum
for long integers by providing a warning or someone could clarify that
indeed we should not be providing inputs larger than 2^31.

As I noted in a previous email to this list
8902c078/a2ea7d9a621d76a1?lnk=gst&q=ian+gbs#a2ea7d9a621d76a1> , there
seems to be no problem allocating lots of memory when we compile models
using 64 bit version of ADMB.

-------------- next part --------------
An HTML attachment was scrubbed...
URL: <http://lists.admb-project.org/pipermail/users/attachments/20100529/be2b37db/attachment.html>

More information about the Users mailing list