[GAP Forum] Freeing memory in gap

Stefan Kohl kohl at mathematik.uni-stuttgart.de
Tue Jul 19 10:14:29 BST 2005


Dear Forum,

Arie Molina Rueda wrote:

> My Situation:
> ------------
> 
> Im working in a problem involving in calculating _A LOT_ of tables of 
> marks, my problem is that GAP keeps stoping, telling me that my memory 
> assingned to GAP ran down.
> 
> The algorithm giving me trouble is very simple:
> 
> For every SimpleGroup(i,j){
>  -> Calculate ToM
>  -> A := Uncompress ToM
>  -> dim := Calculate Dimension of A
>  -> Save dim,i and j to file,
> }
> 
> I assume GAP is keeping in memory those big arrays and thus running out 
> of memory.
> 
> My Question:
> -----------.
> Is there a way to free memory in GAP? or
> to avoid GAP keeping those big things in memory?

Alas the above pseudo code does not provide enough information for
deciding what happens. In an off-list mail, Thomas Breuer asked the
following questions:

1. Do you write down a `for' loop over a list of groups or a list of
    indices?
    In the first case, the tables of marks are stored in the groups as
    they are computed;
    one remedy would then be to use `AttributeValueNotSet' instead of
    calling `TableOfMarks' for the groups in question,
    see the GAP Reference Manual for details about `AttributeValueNotSet'.

2. Do you compute the tables of marks from given group objects,
    or do you use the library of tables of marks?
    In the latter case, a certain number of data files are kept in memory;
    one can reduce this number if required.
    (Note that loading a table of marks from the library is much faster than
    computing the table of marks anew, see `The Library of Tables of Marks'
    in the GAP Reference Manual.)

3. Do I understand you right that you are interested only in the number
    of conjugacy classes of subgroups of a given (simple) group?
    If yes then you can completely avoid computing the table of marks,
    and compute just `Length( ConjugacyClassesSubgroups( G ) )'.

Best wishes,

     Stefan Kohl






More information about the Forum mailing list