Hi Simon, hi all, I've made a small patch to GAP that fixes the problem. (inspired by the corresponding Sage Python code :)) With this patch I get on Sparc:
sage: gap_console() GAP4, Version: 4.4.12 of 17-Dec-2008, sparc-sun-solaris2.10-gcc gap> Reset(GlobalMersenneTwister);; List([1..10],i->Random(1,100000)); [ 85758, 234, 18408, 12676, 2507, 38166, 24692, 15013, 4755, 94043 ] All it does it changes (on bigendian machines) the byte order in the appropriate integers: The following is the diff for GAPROOT/src/integer.c (the context diff is attached) gap4r4/src$ diff integer.c.old integer.c 3120c3120 < UInt4 *mt, *init_key, key_length, i, j, k, N=624; --- > UInt4 *mt, *init_key, key_length, i, j, k, N=624, temp; 3140a3141,3146 > #ifdef WORDS_BIGENDIAN /* to make the behaviour uniform across platforms */ > temp = ((init_key[j] & 0xFF) << 24) | ((init_key[j] & 0xFF00) << 8) > | ((init_key[j] >> 8) & 0xFF00) | ((init_key[j] >> 24) & > 0xFF); > #else > temp = init_key[j]; > #endif 3142c3148 < + init_key[j] + j; --- > + temp + j; It would be nice to put this in the upcoming GAP release... ====== the following is Sage-specific: === I have put the updated Sage package file gap-4.4.12.p4.spkg (I didn't change the number nor commited changes yet) here: http://boxen.math.washington.edu/home/dima/packages/gap-4.4.12.p4.spkg (so you can try and see if this works for you) To update sage with it correctly, one should also remove the analogous fix in SAGEROOT/devel/sage/sage/misc/randstate.pyx (lines 687-703, starting from "if sys.byteorder == 'big':" in set_seed_gap) If you like, we can quickly make another update of gap spkg with this fix (in sync with the fix for randstate.pyx) ==================== Best, Dima On Sep 7, 4:58 pm, Simon King <simon.k...@nuigalway.ie> wrote: > Hi Mike! > > On 7 Sep., 10:42, Simon King <simon.k...@nuigalway.ie> wrote: > > > ... > > That's VERY cool! > > ... modulo one critical comment: In principal (although it should > rarely occur in "real life") one can have several instances of an > interface. So, I think that current_randstate().set_seed_gap() should > accept an optional argument (namely an instance of GAP), and only if > it is not provided then the default sage.interfaces.gap should be > chosen. > > Do people feel I should modify it accordingly? > > Cheers, > Simon -- To post to this group, send an email to sage-devel@googlegroups.com To unsubscribe from this group, send an email to sage-devel+unsubscr...@googlegroups.com For more options, visit this group at http://groups.google.com/group/sage-devel URL: http://www.sagemath.org
