X-Git-Url: https://git.distorted.org.uk/u/mdw/catacomb/blobdiff_plain/d1c733526062c5e1a2fb87edc35ebbaa4f9aa0fc..dd22938ef0d9b0131dad9171a8a95866ceec9607:/mpx-ksqr.c

diff --git a/mpx-ksqr.c b/mpx-ksqr.c
index 25dbb70..8a0ad64 100644
--- a/mpx-ksqr.c
+++ b/mpx-ksqr.c
@@ -1,6 +1,6 @@
 /* -*-c-*-
  *
- * $Id: mpx-ksqr.c,v 1.6 2000/10/08 15:48:35 mdw Exp $
+ * $Id: mpx-ksqr.c,v 1.7 2002/10/09 00:36:03 mdw Exp $
  *
  * Karatsuba-based squaring algorithm
  *
@@ -30,6 +30,9 @@
 /*----- Revision history --------------------------------------------------* 
  *
  * $Log: mpx-ksqr.c,v $
+ * Revision 1.7  2002/10/09 00:36:03  mdw
+ * Fix bounds on workspace for Karatsuba operations.
+ *
  * Revision 1.6  2000/10/08 15:48:35  mdw
  * Rename Karatsuba constants now that we have @gfx_kmul@ too.
  *
@@ -64,7 +67,7 @@
 
 #ifdef TEST_RIG
 #  undef MPK_THRESH
-#  define MPK_THRESH 2
+#  define MPK_THRESH 4
 #endif
 
 /*----- Main code ---------------------------------------------------------*/
@@ -83,9 +86,9 @@
  *		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,
@@ -126,11 +129,6 @@ 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 --- */
 
   {
@@ -138,6 +136,8 @@ void mpx_ksqr(mpw *dv, mpw *dvl,
     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);
@@ -209,8 +209,8 @@ static int usqr(dstr *v)
   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)) {