Just a small remark regarding the memory usage of the operating system. On a 32GB machine I routinely use 31500m but first I stop services not needed and also shut down the graphical desktop environment. In case you need exactly that 1.5GB...
best, @ On Sat, Aug 4, 2012 at 4:09 AM, Alexander Hulpke <hul...@math.colostate.edu> wrote: > > > Dear Forum, Dear Marek Mitros, > > On Aug 3, 2012, at 8/3/12 12:42, Marek Mitros wrote: > >> Hi, >> >> I perform following code in GAP 4.5.5- see below. I see that single >> matrix 133x133 over GF(4) use 8902 bytes of memory. The size of 2A >> class is 1 539 000. So (1 539 000 * 8902)/2^20 = 13 065 MB. >> >> I start gap session with option -o 15000m (is it OK to allocate 15 GB >> of memory ?). And I run command >> c2s:= AsList(c2);; >> >> Is there hope to fit whole conjugacy class in memory ? > > First you'll need a 64-bit binary, and on that I believe the matrices are > larger (I measured 12210 myself) because of packing overhead. Then part of > allocated memory is reserved for master pointers, and as there will be some > overhead for lists. So you are probably looking at something closer to > 25-30GB of memory required. Also (by default matrix group calculations > translate to permutation groups which you don't want to do here) I would call > > c2s:=Orbit(hn,gens[1]);; > > instead to just calculate conjugates. Also note that sometimes a substantial > amount of core memory is used up by the operating system, for example I would > expect that allocating 30GB on a 32GB system would be too much. > >> Are there other >> ways to loop through all large conjugacy class without storing it in >> memory ? I could store it on the disk for example. > > Mathematically I cannot think of a fundamentally better scheme if you really > want to iterate over all class elements -- the centralizer is maximal and you > need to enumerate its cosets. > You of course trade runtime for memory, in extremis by storing instead of > conjugate matrices only a formal word in the generators that will conjugate > this way (and possibly a small fingeroprint of the matrix to make for fast > searching), but that is likely prohibitively expensive. Similarly you could > use external storage on the disk, but you would have to write all the code > for this from scratch. > > Best, > > Alexander Hulpke > > > > > -- Colorado State University, Department of Mathematics, > Weber Building, 1874 Campus Delivery, Fort Collins, CO 80523-1874, USA > email: hul...@math.colostate.edu, Phone: ++1-970-4914288 > http://www.math.colostate.edu/~hulpke > > > > _______________________________________________ > Forum mailing list > Forum@mail.gap-system.org > http://mail.gap-system.org/mailman/listinfo/forum _______________________________________________ Forum mailing list Forum@mail.gap-system.org http://mail.gap-system.org/mailman/listinfo/forum
