X-Git-Url: https://git.distorted.org.uk/u/mdw/catacomb/blobdiff_plain/c90601009dab74affc0419fa75452a9cf6c97809..f52f2db067dc1388b16ab00ddb53e26a381a6e3e:/mpx-kmul.c diff --git a/mpx-kmul.c b/mpx-kmul.c index eaa0848..081de88 100644 --- a/mpx-kmul.c +++ b/mpx-kmul.c @@ -1,13 +1,13 @@ /* -*-c-*- * - * $Id: mpx-kmul.c,v 1.6 2000/10/08 12:11:01 mdw Exp $ + * $Id$ * * Karatsuba's multiplication algorithm * * (c) 1999 Straylight/Edgeware */ -/*----- Licensing notice --------------------------------------------------* +/*----- Licensing notice --------------------------------------------------* * * This file is part of Catacomb. * @@ -15,55 +15,31 @@ * it under the terms of the GNU Library General Public License as * published by the Free Software Foundation; either version 2 of the * License, or (at your option) any later version. - * + * * Catacomb is distributed in the hope that it will be useful, * but WITHOUT ANY WARRANTY; without even the implied warranty of * MERCHANTABILITY or FITNESS FOR A PARTICULAR PURPOSE. See the * GNU Library General Public License for more details. - * + * * You should have received a copy of the GNU Library General Public * License along with Catacomb; if not, write to the Free * Software Foundation, Inc., 59 Temple Place - Suite 330, Boston, * MA 02111-1307, USA. */ -/*----- Revision history --------------------------------------------------* - * - * $Log: mpx-kmul.c,v $ - * Revision 1.6 2000/10/08 12:11:01 mdw - * Use @mpx_ueq@ instead of @MPX_UCMP@. - * - * Revision 1.5 2000/07/29 17:04:02 mdw - * Remove useless header `mpscan.h'. - * - * Revision 1.4 2000/06/17 11:42:11 mdw - * Moved the Karatsuba macros into a separate file for better sharing. - * Fixed some comments. - * - * Revision 1.3 1999/12/13 15:35:01 mdw - * Simplify and improve. - * - * Revision 1.2 1999/12/11 10:58:02 mdw - * Remove tweakable comments. - * - * Revision 1.1 1999/12/10 23:23:51 mdw - * Karatsuba-Ofman multiplication algorithm. - * - */ - /*----- Header files ------------------------------------------------------*/ #include #include #include "mpx.h" -#include "mpx-kmac.h" +#include "karatsuba.h" /*----- Tweakables --------------------------------------------------------*/ #ifdef TEST_RIG -# undef KARATSUBA_CUTOFF -# define KARATSUBA_CUTOFF 2 +# undef MPK_THRESH +# define MPK_THRESH 4 /* Smallest possible correct value */ #endif /*----- Main code ---------------------------------------------------------*/ @@ -82,9 +58,9 @@ * multiplication (e.g., @mpx_umul@) on large numbers, although * more expensive on small ones. * - * The destination must be twice as large as the larger - * argument. The scratch space must be twice as large as the - * larger argument, plus the magic number @KARATSUBA_SLOP@. + * The destination must be three times as large as the larger + * argument. The scratch space must be five times as large as + * the larger argument. */ void mpx_kmul(mpw *dv, mpw *dvl, @@ -106,7 +82,7 @@ void mpx_kmul(mpw *dv, mpw *dvl, MPX_SHRINK(av, avl); MPX_SHRINK(bv, bvl); - if (avl - av <= KARATSUBA_CUTOFF || bvl - bv <= KARATSUBA_CUTOFF) { + if (avl - av <= MPK_THRESH || bvl - bv <= MPK_THRESH) { mpx_umul(dv, dvl, av, avl, bv, bvl); return; } @@ -126,7 +102,7 @@ void mpx_kmul(mpw *dv, mpw *dvl, /* --- First things --- * * * Sort out where to break the factors in half. I'll choose the midpoint - * of the largest one, since this minimizes the amount of work I have to do + * of the larger one, since this minimizes the amount of work I have to do * most effectively. */ @@ -146,20 +122,17 @@ void mpx_kmul(mpw *dv, mpw *dvl, avm = avl; } - assert(((void)"Destination too small for Karatsuba multiply", - dvl - dv >= 4 * m)); - assert(((void)"Not enough workspace for Karatsuba multiply", - svl - sv >= 4 * m)); - /* --- Sort out the middle term --- */ { mpw *bsv = sv + m + 1, *ssv = bsv + m + 1; mpw *rdv = dv + m, *rdvl = rdv + 2 * (m + 2); + assert(rdvl <= dvl); + assert(ssv <= svl); UADD2(sv, bsv, av, avm, avm, avl); UADD2(bsv, ssv, bv, bvm, bvm, bvl); - if (m > KARATSUBA_CUTOFF) + if (m > MPK_THRESH) mpx_kmul(rdv, rdvl, sv, bsv, bsv, ssv, ssv, svl); else mpx_umul(rdv, rdvl, sv, bsv, bsv, ssv); @@ -175,7 +148,7 @@ void mpx_kmul(mpw *dv, mpw *dvl, if (avl == avm || bvl == bvm) MPX_ZERO(rdv + m + 1, dvl); else { - if (m > KARATSUBA_CUTOFF) + if (m > MPK_THRESH) mpx_kmul(sv, ssv, avm, avl, bvm, bvl, ssv, svl); else mpx_umul(sv, ssv, avm, avl, bvm, bvl); @@ -184,7 +157,7 @@ void mpx_kmul(mpw *dv, mpw *dvl, USUB(tdv, sv, svn); } - if (m > KARATSUBA_CUTOFF) + if (m > MPK_THRESH) mpx_kmul(sv, ssv, av, avm, bv, bvm, ssv, svl); else mpx_umul(sv, ssv, av, avm, bv, bvm); @@ -201,21 +174,21 @@ void mpx_kmul(mpw *dv, mpw *dvl, #include #include -#define ALLOC(v, vl, sz) do { \ - size_t _sz = (sz); \ - mpw *_vv = xmalloc(MPWS(_sz)); \ - mpw *_vvl = _vv + _sz; \ - (v) = _vv; \ - (vl) = _vvl; \ +#define ALLOC(v, vl, sz) do { \ + size_t _sz = (sz); \ + mpw *_vv = xmalloc(MPWS(_sz)); \ + mpw *_vvl = _vv + _sz; \ + (v) = _vv; \ + (vl) = _vvl; \ } while (0) -#define LOAD(v, vl, d) do { \ - const dstr *_d = (d); \ - mpw *_v, *_vl; \ - ALLOC(_v, _vl, MPW_RQ(_d->len)); \ - mpx_loadb(_v, _vl, _d->buf, _d->len); \ - (v) = _v; \ - (vl) = _vl; \ +#define LOAD(v, vl, d) do { \ + const dstr *_d = (d); \ + mpw *_v, *_vl; \ + ALLOC(_v, _vl, MPW_RQ(_d->len)); \ + mpx_loadb(_v, _vl, _d->buf, _d->len); \ + (v) = _v; \ + (vl) = _vl; \ } while (0) #define MAX(x, y) ((x) > (y) ? (x) : (y)) @@ -243,20 +216,20 @@ static int umul(dstr *v) LOAD(b, bl, &v[1]); LOAD(c, cl, &v[2]); m = MAX(al - a, bl - b) + 1; - ALLOC(d, dl, 2 * m); - ALLOC(s, sl, 2 * m + 32); + ALLOC(d, dl, 3 * m); + ALLOC(s, sl, 5 * m); mpx_kmul(d, dl, a, al, b, bl, s, sl); if (!mpx_ueq(d, dl, c, cl)) { fprintf(stderr, "\n*** umul failed\n"); - dumpmp(" a", a, al); - dumpmp(" b", b, bl); + dumpmp(" a", a, al); + dumpmp(" b", b, bl); dumpmp("expected", c, cl); dumpmp(" result", d, dl); ok = 0; } - free(a); free(b); free(c); free(d); free(s); + xfree(a); xfree(b); xfree(c); xfree(d); xfree(s); return (ok); }