X-Git-Url: https://git.distorted.org.uk/u/mdw/catacomb/blobdiff_plain/d1c733526062c5e1a2fb87edc35ebbaa4f9aa0fc..dd22938ef0d9b0131dad9171a8a95866ceec9607:/mpx-kmul.c diff --git a/mpx-kmul.c b/mpx-kmul.c index f6b0b4f..1981a28 100644 --- a/mpx-kmul.c +++ b/mpx-kmul.c @@ -1,6 +1,6 @@ /* -*-c-*- * - * $Id: mpx-kmul.c,v 1.7 2000/10/08 15:48:35 mdw Exp $ + * $Id: mpx-kmul.c,v 1.8 2002/10/09 00:36:03 mdw Exp $ * * Karatsuba's multiplication algorithm * @@ -30,6 +30,9 @@ /*----- Revision history --------------------------------------------------* * * $Log: mpx-kmul.c,v $ + * Revision 1.8 2002/10/09 00:36:03 mdw + * Fix bounds on workspace for Karatsuba operations. + * * Revision 1.7 2000/10/08 15:48:35 mdw * Rename Karatsuba constants now that we have @gfx_kmul@ too. * @@ -66,7 +69,7 @@ #ifdef TEST_RIG # undef MPK_THRESH -# define MPK_THRESH 2 +# define MPK_THRESH 4 /* Smallest possible correct value */ #endif /*----- Main code ---------------------------------------------------------*/ @@ -85,9 +88,9 @@ * 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 @MPK_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, @@ -149,17 +152,14 @@ 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 > MPK_THRESH) @@ -246,8 +246,8 @@ 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)) {