summaryrefslogtreecommitdiffstats
path: root/libtommath/bn_fast_s_mp_sqr.c
diff options
context:
space:
mode:
authorKevin B Kenny <kennykb@acm.org>2005-01-19 22:41:26 (GMT)
committerKevin B Kenny <kennykb@acm.org>2005-01-19 22:41:26 (GMT)
commitb9cf65a08e6a59e434685e894e3189c201ac6791 (patch)
tree1f0868ef44c9f17d83d10dc94343df7b8cfe1842 /libtommath/bn_fast_s_mp_sqr.c
parentc6a259aeeca4814a97cf6694814c63e74e4e18fa (diff)
downloadtcl-b9cf65a08e6a59e434685e894e3189c201ac6791.zip
tcl-b9cf65a08e6a59e434685e894e3189c201ac6791.tar.gz
tcl-b9cf65a08e6a59e434685e894e3189c201ac6791.tar.bz2
Import of libtommath 0.33
Diffstat (limited to 'libtommath/bn_fast_s_mp_sqr.c')
-rw-r--r--libtommath/bn_fast_s_mp_sqr.c129
1 files changed, 129 insertions, 0 deletions
diff --git a/libtommath/bn_fast_s_mp_sqr.c b/libtommath/bn_fast_s_mp_sqr.c
new file mode 100644
index 0000000..d6014ab
--- /dev/null
+++ b/libtommath/bn_fast_s_mp_sqr.c
@@ -0,0 +1,129 @@
+#include <tommath.h>
+#ifdef BN_FAST_S_MP_SQR_C
+/* LibTomMath, multiple-precision integer library -- Tom St Denis
+ *
+ * LibTomMath is a library that provides multiple-precision
+ * integer arithmetic as well as number theoretic functionality.
+ *
+ * The library was designed directly after the MPI library by
+ * Michael Fromberger but has been written from scratch with
+ * additional optimizations in place.
+ *
+ * The library is free for all purposes without any express
+ * guarantee it works.
+ *
+ * Tom St Denis, tomstdenis@iahu.ca, http://math.libtomcrypt.org
+ */
+
+/* fast squaring
+ *
+ * This is the comba method where the columns of the product
+ * are computed first then the carries are computed. This
+ * has the effect of making a very simple inner loop that
+ * is executed the most
+ *
+ * W2 represents the outer products and W the inner.
+ *
+ * A further optimizations is made because the inner
+ * products are of the form "A * B * 2". The *2 part does
+ * not need to be computed until the end which is good
+ * because 64-bit shifts are slow!
+ *
+ * Based on Algorithm 14.16 on pp.597 of HAC.
+ *
+ */
+/* the jist of squaring...
+
+you do like mult except the offset of the tmpx [one that starts closer to zero]
+can't equal the offset of tmpy. So basically you set up iy like before then you min it with
+(ty-tx) so that it never happens. You double all those you add in the inner loop
+
+After that loop you do the squares and add them in.
+
+Remove W2 and don't memset W
+
+*/
+
+int fast_s_mp_sqr (mp_int * a, mp_int * b)
+{
+ int olduse, res, pa, ix, iz;
+ mp_digit W[MP_WARRAY], *tmpx;
+ mp_word W1;
+
+ /* grow the destination as required */
+ pa = a->used + a->used;
+ if (b->alloc < pa) {
+ if ((res = mp_grow (b, pa)) != MP_OKAY) {
+ return res;
+ }
+ }
+
+ /* number of output digits to produce */
+ W1 = 0;
+ for (ix = 0; ix < pa; ix++) {
+ int tx, ty, iy;
+ mp_word _W;
+ mp_digit *tmpy;
+
+ /* clear counter */
+ _W = 0;
+
+ /* get offsets into the two bignums */
+ ty = MIN(a->used-1, ix);
+ tx = ix - ty;
+
+ /* setup temp aliases */
+ tmpx = a->dp + tx;
+ tmpy = a->dp + ty;
+
+ /* this is the number of times the loop will iterrate, essentially its
+ while (tx++ < a->used && ty-- >= 0) { ... }
+ */
+ iy = MIN(a->used-tx, ty+1);
+
+ /* now for squaring tx can never equal ty
+ * we halve the distance since they approach at a rate of 2x
+ * and we have to round because odd cases need to be executed
+ */
+ iy = MIN(iy, (ty-tx+1)>>1);
+
+ /* execute loop */
+ for (iz = 0; iz < iy; iz++) {
+ _W += ((mp_word)*tmpx++)*((mp_word)*tmpy--);
+ }
+
+ /* double the inner product and add carry */
+ _W = _W + _W + W1;
+
+ /* even columns have the square term in them */
+ if ((ix&1) == 0) {
+ _W += ((mp_word)a->dp[ix>>1])*((mp_word)a->dp[ix>>1]);
+ }
+
+ /* store it */
+ W[ix] = _W;
+
+ /* make next carry */
+ W1 = _W >> ((mp_word)DIGIT_BIT);
+ }
+
+ /* setup dest */
+ olduse = b->used;
+ b->used = a->used+a->used;
+
+ {
+ mp_digit *tmpb;
+ tmpb = b->dp;
+ for (ix = 0; ix < pa; ix++) {
+ *tmpb++ = W[ix] & MP_MASK;
+ }
+
+ /* clear unused digits [that existed in the old copy of c] */
+ for (; ix < olduse; ix++) {
+ *tmpb++ = 0;
+ }
+ }
+ mp_clamp (b);
+ return MP_OKAY;
+}
+#endif