[GAP Forum] Installing cvec package - and memory issue

[Dmitrii Pasechnik]

On Fri, Aug 01, 2014 at 09:23:27AM +0200, Marek Mitros wrote:
> Thank you for the answers I have received.
> 
> I have another related problem - maybe you can help.
> 
> I am trying to generate 1240 elements of BM8.2D conjugacy class - 8th
> maximal subgroup of baby monster group. Here is the code I tried. The
> problem is that my GAP session disappeared before it reach 800 elements. I
> started GAP session with -o 6G option, GasmanLimits() function show that
> only 4 GB has been set as max option. Anyway it should be enough - 2.5 MB *
> 1240 = 3GB.
IIRC, you are running this on Windows with Cygwin 32bit, and this just
might be above the memory limits imposed by this environment.
If you must stay on Windows, people reported success building GAP from
source on 64-bit Cygwin. There it should be free from this.

HTH,
Dmitrii

> 
> First trial:
> -------------
> g:=AtlasGroup("B"); Print("Baby defined\n");
> g8:=AtlasSubgroup(g,8); Print ("BM8 defined\n");
> ge:=GeneratorsOfGroup(g8); a:=ge[1]; b:=ge[2];
> c:=(a*b*b*b*b)^21;  #1240 size cg which generate [2^30]
> Print("### Phase 1: Creating 2D class\n\n\n");
> c2:=[c];
> repeat c0:=c^PseudoRandom(g8);
>    if not c0 in c2 then
>        Add(c2,c0); if Size(c2) mod 100 =0 then Print(Size(c2),"\n");
> Print(StringTime(Runtime()), "\n"); fi;
>    fi;
> until Size(c2)=800;             #  the speed is 7 elements per 1 minute, so
> after 1/2 hour we should have 200 elements
> Second trial - I used orb package and Enumerate function. Result is the
> same - last log entry is 700 size.
> 
> g8:=AtlasSubgroup("B",8); Print ("BM8 defined\n");
> ge:=GeneratorsOfGroup(g8); a:=ge[1]; b:=ge[2];
> c:=(a*b*b*b*b)^21;  #1240 size cg which generate [2^30]
> o := Orb([a,b],c,OnPoints);
> Print(StringTime(Runtime()), "\n");
> for k in 40+100*[0..12] do
>    Enumerate(o,k);
>    Print(Size(o), " elements generated\n");
>    Print(StringTime(Runtime()), "\n");
> od;
> Any advice ?
> 
> Regards,
> Marek
> 
> On Thu, Jul 31, 2014 at 1:29 PM, Stephen Linton <
> steve.linton at st-andrews.ac.uk> wrote:
> 
> > Dear GAP Forum,
> >
> > Marek Mitros asked:
> >
> > >
> > > I am trying to use the cvec package to have compressed matrices over
> > GF(2).
> > > But I do not see much difference - see below example log from my GAP
> > > session.
> > >
> >
> > This isn’t really surprising. GAP already supports compressed matrices
> > over fields of order up to 256. The cvec package does not achieve more
> > compression for these fields, but (a) may have faster methods for some
> > operations and/or a better interface and (b) also supports space and time
> > efficient arithmetic for larger fields.
> >
> > In your particular case, the matrices have roughly 19 million entries and
> > the memory usages you report are not much over 19 million bits, so there is
> > not much saving to be had.
> >
> >         Steve
> >
> >
> > >
> > > gap> g:=AtlasGroup("B"); Print("Baby defined\n");
> > >
> > > <matrix group of size 4154781481226426191177580544000000 with 2
> > generators>
> > >
> > > Baby defined
> > >
> > > gap> ge:=GeneratorsOfGroup(g); a:=ge[1]; b:=ge[2];
> > >
> > > [ <an immutable 4370x4370 matrix over GF2>, <an immutable 4370x4370
> > matrix
> > > over GF2> ]
> > >
> > > <an immutable 4370x4370 matrix over GF2>
> > >
> > > <an immutable 4370x4370 matrix over GF2>
> > >
> > > gap>
> > >
> > > gap> LoadPackage("cvec"); # compressed vectors and matrices over finite
> > > fields
> > >
> > > true
> > >
> > > gap> ca:=CMat(a); cb:=CMat(b);
> > >
> > > <cmat 4370x4370 over GF(2,1)>
> > >
> > > <cmat 4370x4370 over GF(2,1)>
> > >
> > > gap>
> > >
> > > gap> MemoryUsage(a);
> > >
> > > 2519320
> > >
> > > gap> MemoryUsage(ca);
> > >
> > > 2500728
> > > _______________________________________________
> > > Forum mailing list
> > > Forum at mail.gap-system.org
> > > http://mail.gap-system.org/mailman/listinfo/forum
> >
> >
> _______________________________________________
> Forum mailing list
> Forum at mail.gap-system.org
> http://mail.gap-system.org/mailman/listinfo/forum