Remove two bn_wexpand() from BN_mul(), which is a step toward getting
BN_mul() correctly constified, avoids two realloc()'s that aren't
really necessary and saves memory to boot. This required a small
change in bn_mul_part_recursive() and the addition of variants of
bn_cmp_words(), bn_add_words() and bn_sub_words() that can take arrays
with differing sizes.
The test results show a performance that very closely matches the
original code from before my constification. This may seem like a
very small win from a performance point of view, but if one remembers
that the variants of bn_cmp_words(), bn_add_words() and bn_sub_words()
are not at all optimized for the moment (and there's no corresponding
assembler code), and that their use may be just as non-optimal, I'm
pretty confident there are possibilities...
This code needs reviewing!
diff --git a/CHANGES b/CHANGES
index a6f14fd..c54ab99 100644
--- a/CHANGES
+++ b/CHANGES
@@ -4,6 +4,15 @@
Changes between 0.9.6 and 0.9.7 [xx XXX 2000]
+ *) Remove a few calls to bn_wexpand() in BN_sqr() (the one in there
+ was actually never needed) and in BN_mul(). The removal in BN_mul()
+ required a small change in bn_mul_part_recursive() and the addition
+ of the static functions bn_cmp_part_words(), bn_sub_part_words()
+ and bn_add_part_words() which do the same thing as bn_cmp_words(),
+ bn_sub_words() and bn_add_words() except they take arrays with
+ differing sizes.
+ [Richard Levitte]
+
*) In 'openssl passwd', verify passwords read from the terminal
unless the '-salt' option is used (which usually means that
verification would just waste user's time since the resulting
diff --git a/crypto/bn/bn_mul.c b/crypto/bn/bn_mul.c
index d3908b5..f730e5d 100644
--- a/crypto/bn/bn_mul.c
+++ b/crypto/bn/bn_mul.c
@@ -57,9 +57,328 @@
*/
#include <stdio.h>
+#include <assert.h>
#include "cryptlib.h"
#include "bn_lcl.h"
+/* Here follows specialised variants of bn_cmp_words(), bn_add_words() and
+ bn_sub_words(). They all have the property performing operations on
+ arrays of different sizes. The sizes of those arrays is expressed through
+ cl, which is the common length ( basicall, min(len(a),len(b)) ), and dl,
+ which is the delta between the two lengths, calculated as len(a)-len(b).
+ All lengths are the number of BN_ULONGs... For the operations that require
+ a result array as parameter, it must have the length cl+abs(dl).
+ These functions should probably end up in bn_asm.c as soon as there are
+ assembler counterparts for the systems that use assembler files. */
+
+int bn_cmp_part_words(const BN_ULONG *a, const BN_ULONG *b,
+ int cl, int dl)
+ {
+ if (dl < 0) /* a < b */
+ return -1;
+ if (dl > 0) /* a > b */
+ return 1;
+
+ return bn_cmp_words(a,b,cl);
+ }
+
+BN_ULONG bn_sub_part_words(BN_ULONG *r,
+ const BN_ULONG *a, const BN_ULONG *b,
+ int cl, int dl)
+ {
+ BN_ULONG c, t;
+
+ assert(cl >= 0);
+ c = bn_sub_words(r, a, b, cl);
+
+ if (dl == 0)
+ return c;
+
+ r += cl;
+ a += cl;
+ b += cl;
+
+ if (dl < 0)
+ {
+#ifdef BN_COUNT
+ fprintf(stderr, " bn_sub_part_words %d + %d (dl < 0, c = %d)\n", cl, dl, c);
+#endif
+ for (;;)
+ {
+ t = b[0];
+ r[0] = (-t-c)&BN_MASK2;
+ if (t != 0) c=1;
+ if (++dl >= 0) break;
+
+ t = b[1];
+ r[1] = (-t-c)&BN_MASK2;
+ if (t != 0) c=1;
+ if (++dl >= 0) break;
+
+ t = b[2];
+ r[2] = (-t-c)&BN_MASK2;
+ if (t != 0) c=1;
+ if (++dl >= 0) break;
+
+ t = b[3];
+ r[3] = (-t-c)&BN_MASK2;
+ if (t != 0) c=1;
+ if (++dl >= 0) break;
+
+ b += 4;
+ r += 4;
+ }
+ }
+ else
+ {
+ int save_dl = dl;
+#ifdef BN_COUNT
+ fprintf(stderr, " bn_sub_part_words %d + %d (dl > 0, c = %d)\n", cl, dl, c);
+#endif
+ while(c)
+ {
+ t = a[0];
+ r[0] = (t-c)&BN_MASK2;
+ if (t != 0) c=0;
+ if (--dl <= 0) break;
+
+ t = a[1];
+ r[1] = (t-c)&BN_MASK2;
+ if (t != 0) c=0;
+ if (--dl <= 0) break;
+
+ t = a[2];
+ r[2] = (t-c)&BN_MASK2;
+ if (t != 0) c=0;
+ if (--dl <= 0) break;
+
+ t = a[3];
+ r[3] = (t-c)&BN_MASK2;
+ if (t != 0) c=0;
+ if (--dl <= 0) break;
+
+ save_dl = dl;
+ a += 4;
+ r += 4;
+ }
+ if (dl > 0)
+ {
+#ifdef BN_COUNT
+ fprintf(stderr, " bn_sub_part_words %d + %d (dl > 0, c == 0)\n", cl, dl);
+#endif
+ if (save_dl > dl)
+ {
+ switch (save_dl - dl)
+ {
+ case 1:
+ r[1] = a[1];
+ if (--dl <= 0) break;
+ case 2:
+ r[2] = a[2];
+ if (--dl <= 0) break;
+ case 3:
+ r[3] = a[3];
+ if (--dl <= 0) break;
+ }
+ a += 4;
+ r += 4;
+ }
+ }
+ if (dl > 0)
+ {
+#ifdef BN_COUNT
+ fprintf(stderr, " bn_sub_part_words %d + %d (dl > 0, copy)\n", cl, dl);
+#endif
+ for(;;)
+ {
+ r[0] = a[0];
+ if (--dl <= 0) break;
+ r[1] = a[1];
+ if (--dl <= 0) break;
+ r[2] = a[2];
+ if (--dl <= 0) break;
+ r[3] = a[3];
+ if (--dl <= 0) break;
+
+ a += 4;
+ r += 4;
+ }
+ }
+ }
+ return c;
+ }
+
+BN_ULONG bn_add_part_words(BN_ULONG *r,
+ const BN_ULONG *a, const BN_ULONG *b,
+ int cl, int dl)
+ {
+ BN_ULONG c, l, t;
+
+ assert(cl >= 0);
+ c = bn_sub_words(r, a, b, cl);
+
+ if (dl == 0)
+ return c;
+
+ r += cl;
+ a += cl;
+ b += cl;
+
+ if (dl < 0)
+ {
+ int save_dl = dl;
+#ifdef BN_COUNT
+ fprintf(stderr, " bn_add_part_words %d + %d (dl < 0, c = %d)\n", cl, dl, c);
+#endif
+ while (c)
+ {
+ l=(c+b[0])&BN_MASK2;
+ c=(l < c);
+ r[0]=l;
+ if (++dl >= 0) break;
+
+ l=(c+b[1])&BN_MASK2;
+ c=(l < c);
+ r[1]=l;
+ if (++dl >= 0) break;
+
+ l=(c+b[2])&BN_MASK2;
+ c=(l < c);
+ r[2]=l;
+ if (++dl >= 0) break;
+
+ l=(c+b[3])&BN_MASK2;
+ c=(l < c);
+ r[3]=l;
+ if (++dl >= 0) break;
+
+ save_dl = dl;
+ b+=4;
+ r+=4;
+ }
+ if (dl < 0)
+ {
+#ifdef BN_COUNT
+ fprintf(stderr, " bn_add_part_words %d + %d (dl < 0, c == 0)\n", cl, dl);
+#endif
+ if (save_dl < dl)
+ {
+ switch (dl - save_dl)
+ {
+ case 1:
+ r[1] = b[1];
+ if (++dl >= 0) break;
+ case 2:
+ r[2] = b[2];
+ if (++dl >= 0) break;
+ case 3:
+ r[3] = b[3];
+ if (++dl >= 0) break;
+ }
+ b += 4;
+ r += 4;
+ }
+ }
+ if (dl < 0)
+ {
+#ifdef BN_COUNT
+ fprintf(stderr, " bn_add_part_words %d + %d (dl < 0, copy)\n", cl, dl);
+#endif
+ for(;;)
+ {
+ r[0] = b[0];
+ if (++dl >= 0) break;
+ r[1] = b[1];
+ if (++dl >= 0) break;
+ r[2] = b[2];
+ if (++dl >= 0) break;
+ r[3] = b[3];
+ if (++dl >= 0) break;
+
+ b += 4;
+ r += 4;
+ }
+ }
+ }
+ else
+ {
+ int save_dl = dl;
+#ifdef BN_COUNT
+ fprintf(stderr, " bn_add_part_words %d + %d (dl > 0)\n", cl, dl);
+#endif
+ while (c)
+ {
+ t=(a[0]+c)&BN_MASK2;
+ c=(t < c);
+ r[0]=t;
+ if (--dl <= 0) break;
+
+ t=(a[1]+c)&BN_MASK2;
+ c=(t < c);
+ r[1]=t;
+ if (--dl <= 0) break;
+
+ t=(a[2]+c)&BN_MASK2;
+ c=(t < c);
+ r[2]=t;
+ if (--dl <= 0) break;
+
+ t=(a[3]+c)&BN_MASK2;
+ c=(t < c);
+ r[3]=t;
+ if (--dl <= 0) break;
+
+ save_dl = dl;
+ a+=4;
+ r+=4;
+ }
+#ifdef BN_COUNT
+ fprintf(stderr, " bn_add_part_words %d + %d (dl > 0, c == 0)\n", cl, dl);
+#endif
+ if (dl > 0)
+ {
+ if (save_dl > dl)
+ {
+ switch (save_dl - dl)
+ {
+ case 1:
+ r[1] = a[1];
+ if (--dl <= 0) break;
+ case 2:
+ r[2] = a[2];
+ if (--dl <= 0) break;
+ case 3:
+ r[3] = a[3];
+ if (--dl <= 0) break;
+ }
+ a += 4;
+ r += 4;
+ }
+ }
+ if (dl > 0)
+ {
+#ifdef BN_COUNT
+ fprintf(stderr, " bn_add_part_words %d + %d (dl > 0, copy)\n", cl, dl);
+#endif
+ for(;;)
+ {
+ r[0] = a[0];
+ if (--dl <= 0) break;
+ r[1] = a[1];
+ if (--dl <= 0) break;
+ r[2] = a[2];
+ if (--dl <= 0) break;
+ r[3] = a[3];
+ if (--dl <= 0) break;
+
+ a += 4;
+ r += 4;
+ }
+ }
+ }
+ return c;
+ }
+
#ifdef BN_RECURSION
/* Karatsuba recursive multiplication algorithm
* (cf. Knuth, The Art of Computer Programming, Vol. 2) */
@@ -228,7 +547,8 @@
BN_ULONG ln,lo,*p;
# ifdef BN_COUNT
- fprintf(stderr," bn_mul_part_recursive %d * %d\n",tn+n,tn+n);
+ fprintf(stderr," bn_mul_part_recursive (%d+%d) * (%d+%d)\n",
+ tn, n,tn, n);
# endif
if (n < 8)
{
@@ -238,21 +558,21 @@
}
/* r=(a[0]-a[1])*(b[1]-b[0]) */
- c1=bn_cmp_words(a,&(a[n]),n);
- c2=bn_cmp_words(&(b[n]),b,n);
+ c1=bn_cmp_part_words(a,&(a[n]),tn,n-tn);
+ c2=bn_cmp_part_words(&(b[n]),b,tn,tn-n);
zero=neg=0;
switch (c1*3+c2)
{
case -4:
- bn_sub_words(t, &(a[n]),a, n); /* - */
- bn_sub_words(&(t[n]),b, &(b[n]),n); /* - */
+ bn_sub_part_words(t, &(a[n]),a, tn,tn-n); /* - */
+ bn_sub_part_words(&(t[n]),b, &(b[n]),tn,n-tn); /* - */
break;
case -3:
zero=1;
/* break; */
case -2:
- bn_sub_words(t, &(a[n]),a, n); /* - */
- bn_sub_words(&(t[n]),&(b[n]),b, n); /* + */
+ bn_sub_part_words(t, &(a[n]),a, tn,tn-n); /* - */
+ bn_sub_part_words(&(t[n]),&(b[n]),b, tn,tn-n); /* + */
neg=1;
break;
case -1:
@@ -261,16 +581,16 @@
zero=1;
/* break; */
case 2:
- bn_sub_words(t, a, &(a[n]),n); /* + */
- bn_sub_words(&(t[n]),b, &(b[n]),n); /* - */
+ bn_sub_part_words(t, a, &(a[n]),tn,n-tn); /* + */
+ bn_sub_part_words(&(t[n]),b, &(b[n]),tn,n-tn); /* - */
neg=1;
break;
case 3:
zero=1;
/* break; */
case 4:
- bn_sub_words(t, a, &(a[n]),n);
- bn_sub_words(&(t[n]),&(b[n]),b, n);
+ bn_sub_part_words(t, a, &(a[n]),tn,n-tn);
+ bn_sub_part_words(&(t[n]),&(b[n]),b, tn,tn-n);
break;
}
/* The zero case isn't yet implemented here. The speedup
@@ -678,21 +998,19 @@
{
if (i == 1 && !BN_get_flags(b,BN_FLG_STATIC_DATA))
{
- BIGNUM *tmp_bn = free_b;
- b = free_b = bn_dup_expand(b,al);
- free_b->d[bl]=0;
+ BIGNUM *tmp_bn = (BIGNUM *)b;
+ bn_wexpand(tmp_bn,al);
+ tmp_bn->d[bl]=0;
bl++;
i--;
- if (tmp_bn) BN_free(tmp_bn);
}
else if (i == -1 && !BN_get_flags(a,BN_FLG_STATIC_DATA))
{
- BIGNUM *tmp_bn = free_a;
- a = free_a = bn_dup_expand(a,bl);
- free_a->d[al]=0;
+ BIGNUM *tmp_bn = (BIGNUM *)a;
+ bn_wexpand(tmp_bn,bl);
+ tmp_bn->d[al]=0;
al++;
i++;
- if (tmp_bn) BN_free(tmp_bn);
}
if (i == 0)
{
@@ -710,17 +1028,8 @@
}
else
{
- BIGNUM *tmp_a = free_a,*tmp_b = free_b;
- a = free_a = bn_dup_expand(a,k);
- b = free_b = bn_dup_expand(b,k);
- if (tmp_a) BN_free(tmp_a);
- if (tmp_b) BN_free(tmp_b);
bn_wexpand(t,k*4);
bn_wexpand(rr,k*4);
- for (i=free_a->top; i<k; i++)
- free_a->d[i]=0;
- for (i=free_b->top; i<k; i++)
- free_b->d[i]=0;
bn_mul_part_recursive(rr->d,a->d,b->d,al-j,j,t->d);
}
rr->top=top;
