[GAP Forum] Exceeded Permitted Memory

[Attila Egri-Nagy]

Hi Minghui,

It probably helps if you can get some idea about the scalability of your problem. If you could check the memory consumption of calculating smaller instances of the same problem (like similar relations with less generators).

As Dimitrii pointed out, you may want to set the size of the memory allocated to GAP close to the available physical memory. Or simple just say return; in the break loop when interactive, and GAP will try to allocate more memory.

In general, the supercomputer is great if you can make your calculation parallel. For single thread calculation you may find a desktop computer with loads of memory more useful. See blog entry: http://compsemi.wordpress.com/2014/05/04/building-a-harvester-rig/

best,
attila 
________________________________________
From: forum-bounces at gap-system.org [forum-bounces at gap-system.org] on behalf of Minghui Liu [matliumh at gmail.com]
Sent: Saturday, May 03, 2014 4:03 PM
To: forum at gap-system.org
Subject: Re: [GAP Forum] Exceeded Permitted Memory

By the way, I found that even better supercomputers are available in my
university:

"Dedicated cluster with up to 192 CPU cores interconnected by
high-performance InfiniBand network can be provisioned within a business
day."

And one needs to pay to use it. Is there any point trying?

Thanks and regards,

Minghui


On 3 May 2014 13:53, Minghui Liu <matliumh at gmail.com> wrote:

> Dear Forum,
>
> I am using GAP to compute a group with hundreds of generators and
> relations. The command AbelianInvariants(F/relations) works perfectly, but
> when I  use the command MaximalAbelianQuotient(F/relations); the following
> message is returned:
>
> gap> phi:=MaximalAbelianQuotient(G);
> Error, exceeded the permitted memory (`-o' command line option) in
>   MakeImmutable( a ); called from
> UnderlyingElement( left ) * UnderlyingElement( right ) called from
> gen[j] ^ s[i][j] called from
> <function "unknown">( <arguments> )
>  called from read-eval loop at line 80 of *stdin*
> you can 'return;'
> brk>
>
> I also tried to run GAP on a supercomputer (HP Xeon four sockets 10-Core
> and two sockets Hexa-Core 64-bit Linux cluster, CentOS 5) but with the same
> result (Does a supercomputer make any difference at all?). Is there any way
> that I can slove this, or should we concluded that my group is "too large"
> to be computed by GAP.
>
> Anyway my goal is to find the torsion elements in F/relations. As I can
> read from AbelianInvariants(F/relations), there are four copies of Z/Z2. Is
> there any other way that I can find and verify the four elements in F whose
> image has order 2 in F/relations?
>
> I very much appreciate your help!
>
> Best regards,
>
> Minghui
>
_______________________________________________
Forum mailing list
Forum at mail.gap-system.org
http://mail.gap-system.org/mailman/listinfo/forum