/* -*-c-*-
*
- * $Id: mpx-ksqr.c,v 1.6 2000/10/08 15:48:35 mdw Exp $
+ * $Id: mpx-ksqr.c,v 1.8 2004/04/08 01:36:15 mdw Exp $
*
* Karatsuba-based squaring algorithm
*
* MA 02111-1307, USA.
*/
-/*----- Revision history --------------------------------------------------*
- *
- * $Log: mpx-ksqr.c,v $
- * Revision 1.6 2000/10/08 15:48:35 mdw
- * Rename Karatsuba constants now that we have @gfx_kmul@ too.
- *
- * Revision 1.5 2000/10/08 12:11:01 mdw
- * Use @mpx_ueq@ instead of @MPX_UCMP@.
- *
- * Revision 1.4 2000/07/29 17:04:02 mdw
- * Remove useless header `mpscan.h'.
- *
- * Revision 1.3 2000/06/17 11:42:54 mdw
- * Moved the Karatsuba macros into a separate file for better sharing.
- * Fixed some comments. Use an improved technique so that all the
- * operations are squarings.
- *
- * Revision 1.2 1999/12/13 15:35:01 mdw
- * Simplify and improve.
- *
- * Revision 1.1 1999/12/11 10:57:43 mdw
- * Karatsuba squaring algorithm.
- *
- */
-
/*----- Header files ------------------------------------------------------*/
#include <assert.h>
#ifdef TEST_RIG
# undef MPK_THRESH
-# define MPK_THRESH 2
+# define MPK_THRESH 4
#endif
/*----- Main code ---------------------------------------------------------*/
* large numbers, although more expensive on small ones, and
* rather simpler than full-blown Karatsuba multiplication.
*
- * The destination must be twice as large as the argument. The
- * scratch space must be twice as large as the 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_ksqr(mpw *dv, mpw *dvl,
m = (avl - av + 1) >> 1;
avm = av + m;
- assert(((void)"Destination too small for Karatsuba square",
- dvl - dv >= 4 * m));
- assert(((void)"Not enough workspace for Karatsuba square",
- svl - sv >= 4 * m));
-
/* --- Sort out everything --- */
{
mpw *tdv = dv + m;
mpw *rdv = tdv + m;
+ assert(rdv + m + 4 < dvl);
+ assert(ssv < svl);
UADD2(sv, svm, av, avm, avm, avl);
if (m > MPK_THRESH)
mpx_ksqr(tdv, rdv + m + 4, sv, svm + 1, ssv, svl);
LOAD(a, al, &v[0]);
LOAD(c, cl, &v[1]);
m = al - a + 1;
- ALLOC(d, dl, 2 * m);
- ALLOC(s, sl, 2 * m + 32);
+ ALLOC(d, dl, 3 * m);
+ ALLOC(s, sl, 5 * m);
mpx_ksqr(d, dl, a, al, s, sl);
if (!mpx_ueq(d, dl, c, cl)) {