/* sqr_add_c2(a,i,c0,c1,c2) -- c+=2*a[i]*a[j] for three word number c=(c2,c1,c0) */
#ifdef BN_LLONG
-#define mul_add_c(a,b,c0,c1,c2) \
- t=(BN_ULLONG)a*b; \
- t1=(BN_ULONG)Lw(t); \
- t2=(BN_ULONG)Hw(t); \
- c0=(c0+t1)&BN_MASK2; if ((c0) < t1) t2++; \
- c1=(c1+t2)&BN_MASK2; if ((c1) < t2) c2++;
-
-#define mul_add_c2(a,b,c0,c1,c2) \
- t=(BN_ULLONG)a*b; \
- tt=(t+t)&BN_MASK; \
- if (tt < t) c2++; \
- t1=(BN_ULONG)Lw(tt); \
- t2=(BN_ULONG)Hw(tt); \
- c0=(c0+t1)&BN_MASK2; \
- if ((c0 < t1) && (((++t2)&BN_MASK2) == 0)) c2++; \
- c1=(c1+t2)&BN_MASK2; if ((c1) < t2) c2++;
-
-#define sqr_add_c(a,i,c0,c1,c2) \
- t=(BN_ULLONG)a[i]*a[i]; \
- t1=(BN_ULONG)Lw(t); \
- t2=(BN_ULONG)Hw(t); \
- c0=(c0+t1)&BN_MASK2; if ((c0) < t1) t2++; \
- c1=(c1+t2)&BN_MASK2; if ((c1) < t2) c2++;
+/*
+ * Keep in mind that additions to multiplication result can not
+ * overflow, because its high half cannot be all-ones.
+ */
+#define mul_add_c(a,b,c0,c1,c2) do { \
+ BN_ULONG hi; \
+ BN_ULLONG t = (BN_ULLONG)(a)*(b); \
+ t += c0; /* no carry */ \
+ c0 = (BN_ULONG)Lw(t); \
+ hi = (BN_ULONG)Hw(t); \
+ c1 = (c1+hi)&BN_MASK2; if (c1<hi) c2++; \
+ } while(0)
+
+#define mul_add_c2(a,b,c0,c1,c2) do { \
+ BN_ULONG hi; \
+ BN_ULLONG t = (BN_ULLONG)(a)*(b); \
+ BN_ULLONG tt = t+c0; /* no carry */ \
+ c0 = (BN_ULONG)Lw(tt); \
+ hi = (BN_ULONG)Hw(tt); \
+ c1 = (c1+hi)&BN_MASK2; if (c1<hi) c2++; \
+ t += c0; /* no carry */ \
+ c0 = (BN_ULONG)Lw(t); \
+ hi = (BN_ULONG)Hw(t); \
+ c1 = (c1+hi)&BN_MASK2; if (c1<hi) c2++; \
+ } while(0)
+
+#define sqr_add_c(a,i,c0,c1,c2) do { \
+ BN_ULONG hi; \
+ BN_ULLONG t = (BN_ULLONG)a[i]*a[i]; \
+ t += c0; /* no carry */ \
+ c0 = (BN_ULONG)Lw(t); \
+ hi = (BN_ULONG)Hw(t); \
+ c1 = (c1+hi)&BN_MASK2; if (c1<hi) c2++; \
+ } while(0)
#define sqr_add_c2(a,i,j,c0,c1,c2) \
mul_add_c2((a)[i],(a)[j],c0,c1,c2)
#elif defined(BN_UMULT_LOHI)
-
-#define mul_add_c(a,b,c0,c1,c2) { \
- BN_ULONG ta=(a),tb=(b); \
- BN_UMULT_LOHI(t1,t2,ta,tb); \
- c0 += t1; t2 += (c0<t1)?1:0; \
- c1 += t2; c2 += (c1<t2)?1:0; \
- }
-
-#define mul_add_c2(a,b,c0,c1,c2) { \
- BN_ULONG ta=(a),tb=(b),t0; \
- BN_UMULT_LOHI(t0,t1,ta,tb); \
- t2 = t1+t1; c2 += (t2<t1)?1:0; \
- t1 = t0+t0; t2 += (t1<t0)?1:0; \
- c0 += t1; t2 += (c0<t1)?1:0; \
- c1 += t2; c2 += (c1<t2)?1:0; \
- }
-
-#define sqr_add_c(a,i,c0,c1,c2) { \
- BN_ULONG ta=(a)[i]; \
- BN_UMULT_LOHI(t1,t2,ta,ta); \
- c0 += t1; t2 += (c0<t1)?1:0; \
- c1 += t2; c2 += (c1<t2)?1:0; \
- }
+/*
+ * Keep in mind that additions to hi can not overflow, because
+ * the high word of a multiplication result cannot be all-ones.
+ */
+#define mul_add_c(a,b,c0,c1,c2) do { \
+ BN_ULONG ta = (a), tb = (b); \
+ BN_ULONG lo, hi; \
+ BN_UMULT_LOHI(lo,hi,ta,tb); \
+ c0 += lo; hi += (c0<lo)?1:0; \
+ c1 += hi; c2 += (c1<hi)?1:0; \
+ } while(0)
+
+#define mul_add_c2(a,b,c0,c1,c2) do { \
+ BN_ULONG ta = (a), tb = (b); \
+ BN_ULONG lo, hi, tt; \
+ BN_UMULT_LOHI(lo,hi,ta,tb); \
+ c0 += lo; tt = hi+((c0<lo)?1:0); \
+ c1 += tt; c2 += (c1<tt)?1:0; \
+ c0 += lo; hi += (c0<lo)?1:0; \
+ c1 += hi; c2 += (c1<hi)?1:0; \
+ } while(0)
+
+#define sqr_add_c(a,i,c0,c1,c2) do { \
+ BN_ULONG ta = (a)[i]; \
+ BN_ULONG lo, hi; \
+ BN_UMULT_LOHI(lo,hi,ta,ta); \
+ c0 += lo; hi += (c0<lo)?1:0; \
+ c1 += hi; c2 += (c1<hi)?1:0; \
+ } while(0)
#define sqr_add_c2(a,i,j,c0,c1,c2) \
mul_add_c2((a)[i],(a)[j],c0,c1,c2)
#elif defined(BN_UMULT_HIGH)
-
-#define mul_add_c(a,b,c0,c1,c2) { \
- BN_ULONG ta=(a),tb=(b); \
- t1 = ta * tb; \
- t2 = BN_UMULT_HIGH(ta,tb); \
- c0 += t1; t2 += (c0<t1)?1:0; \
- c1 += t2; c2 += (c1<t2)?1:0; \
- }
-
-#define mul_add_c2(a,b,c0,c1,c2) { \
- BN_ULONG ta=(a),tb=(b),t0; \
- t1 = BN_UMULT_HIGH(ta,tb); \
- t0 = ta * tb; \
- t2 = t1+t1; c2 += (t2<t1)?1:0; \
- t1 = t0+t0; t2 += (t1<t0)?1:0; \
- c0 += t1; t2 += (c0<t1)?1:0; \
- c1 += t2; c2 += (c1<t2)?1:0; \
- }
-
-#define sqr_add_c(a,i,c0,c1,c2) { \
- BN_ULONG ta=(a)[i]; \
- t1 = ta * ta; \
- t2 = BN_UMULT_HIGH(ta,ta); \
- c0 += t1; t2 += (c0<t1)?1:0; \
- c1 += t2; c2 += (c1<t2)?1:0; \
- }
+/*
+ * Keep in mind that additions to hi can not overflow, because
+ * the high word of a multiplication result cannot be all-ones.
+ */
+#define mul_add_c(a,b,c0,c1,c2) do { \
+ BN_ULONG ta = (a), tb = (b); \
+ BN_ULONG lo = ta * tb; \
+ BN_ULONG hi = BN_UMULT_HIGH(ta,tb); \
+ c0 += lo; hi += (c0<lo)?1:0; \
+ c1 += hi; c2 += (c1<hi)?1:0; \
+ } while(0)
+
+#define mul_add_c2(a,b,c0,c1,c2) do { \
+ BN_ULONG ta = (a), tb = (b), tt; \
+ BN_ULONG lo = ta * tb; \
+ BN_ULONG hi = BN_UMULT_HIGH(ta,tb); \
+ c0 += lo; tt = hi + ((c0<lo)?1:0); \
+ c1 += tt; c2 += (c1<tt)?1:0; \
+ c0 += lo; hi += (c0<lo)?1:0; \
+ c1 += hi; c2 += (c1<hi)?1:0; \
+ } while(0)
+
+#define sqr_add_c(a,i,c0,c1,c2) do { \
+ BN_ULONG ta = (a)[i]; \
+ BN_ULONG lo = ta * ta; \
+ BN_ULONG hi = BN_UMULT_HIGH(ta,ta); \
+ c0 += lo; hi += (c0<lo)?1:0; \
+ c1 += hi; c2 += (c1<hi)?1:0; \
+ } while(0)
#define sqr_add_c2(a,i,j,c0,c1,c2) \
mul_add_c2((a)[i],(a)[j],c0,c1,c2)
#else /* !BN_LLONG */
-#define mul_add_c(a,b,c0,c1,c2) \
- t1=LBITS(a); t2=HBITS(a); \
- bl=LBITS(b); bh=HBITS(b); \
- mul64(t1,t2,bl,bh); \
- c0=(c0+t1)&BN_MASK2; if ((c0) < t1) t2++; \
- c1=(c1+t2)&BN_MASK2; if ((c1) < t2) c2++;
-
-#define mul_add_c2(a,b,c0,c1,c2) \
- t1=LBITS(a); t2=HBITS(a); \
- bl=LBITS(b); bh=HBITS(b); \
- mul64(t1,t2,bl,bh); \
- if (t2 & BN_TBIT) c2++; \
- t2=(t2+t2)&BN_MASK2; \
- if (t1 & BN_TBIT) t2++; \
- t1=(t1+t1)&BN_MASK2; \
- c0=(c0+t1)&BN_MASK2; \
- if ((c0 < t1) && (((++t2)&BN_MASK2) == 0)) c2++; \
- c1=(c1+t2)&BN_MASK2; if ((c1) < t2) c2++;
-
-#define sqr_add_c(a,i,c0,c1,c2) \
- sqr64(t1,t2,(a)[i]); \
- c0=(c0+t1)&BN_MASK2; if ((c0) < t1) t2++; \
- c1=(c1+t2)&BN_MASK2; if ((c1) < t2) c2++;
+/*
+ * Keep in mind that additions to hi can not overflow, because
+ * the high word of a multiplication result cannot be all-ones.
+ */
+#define mul_add_c(a,b,c0,c1,c2) do { \
+ BN_ULONG lo = LBITS(a), hi = HBITS(a); \
+ BN_ULONG bl = LBITS(b), bh = HBITS(b); \
+ mul64(lo,hi,bl,bh); \
+ c0 = (c0+lo)&BN_MASK2; if (c0<lo) hi++; \
+ c1 = (c1+hi)&BN_MASK2; if (c1<hi) c2++; \
+ } while(0)
+
+#define mul_add_c2(a,b,c0,c1,c2) do { \
+ BN_ULONG tt; \
+ BN_ULONG lo = LBITS(a), hi = HBITS(a); \
+ BN_ULONG bl = LBITS(b), bh = HBITS(b); \
+ mul64(lo,hi,bl,bh); \
+ tt = hi; \
+ c0 = (c0+lo)&BN_MASK2; if (c0<lo) tt++; \
+ c1 = (c1+tt)&BN_MASK2; if (c1<tt) c2++; \
+ c0 = (c0+lo)&BN_MASK2; if (c0<lo) hi++; \
+ c1 = (c1+hi)&BN_MASK2; if (c1<hi) c2++; \
+ } while(0)
+
+#define sqr_add_c(a,i,c0,c1,c2) do { \
+ BN_ULONG lo, hi; \
+ sqr64(lo,hi,(a)[i]); \
+ c0 = (c0+lo)&BN_MASK2; if (c0<lo) hi++; \
+ c1 = (c1+hi)&BN_MASK2; if (c1<hi) c2++; \
+ } while(0)
#define sqr_add_c2(a,i,j,c0,c1,c2) \
mul_add_c2((a)[i],(a)[j],c0,c1,c2)
void bn_mul_comba8(BN_ULONG *r, BN_ULONG *a, BN_ULONG *b)
{
-#ifdef BN_LLONG
- BN_ULLONG t;
-#else
- BN_ULONG bl,bh;
-#endif
- BN_ULONG t1,t2;
BN_ULONG c1,c2,c3;
c1=0;
void bn_mul_comba4(BN_ULONG *r, BN_ULONG *a, BN_ULONG *b)
{
-#ifdef BN_LLONG
- BN_ULLONG t;
-#else
- BN_ULONG bl,bh;
-#endif
- BN_ULONG t1,t2;
BN_ULONG c1,c2,c3;
c1=0;
void bn_sqr_comba8(BN_ULONG *r, const BN_ULONG *a)
{
-#ifdef BN_LLONG
- BN_ULLONG t,tt;
-#else
- BN_ULONG bl,bh;
-#endif
- BN_ULONG t1,t2;
BN_ULONG c1,c2,c3;
c1=0;
void bn_sqr_comba4(BN_ULONG *r, const BN_ULONG *a)
{
-#ifdef BN_LLONG
- BN_ULLONG t,tt;
-#else
- BN_ULONG bl,bh;
-#endif
- BN_ULONG t1,t2;
BN_ULONG c1,c2,c3;
c1=0;
r[7]=c2;
}
+#ifdef OPENSSL_NO_ASM
#ifdef OPENSSL_BN_ASM_MONT
+#include <alloca.h>
/*
* This is essentially reference implementation, which may or may not
* result in performance improvement. E.g. on IA-32 this routine was
* observed to give 40% faster rsa1024 private key operations and 10%
* faster rsa4096 ones, while on AMD64 it improves rsa1024 sign only
* by 10% and *worsens* rsa4096 sign by 15%. Once again, it's a
- * reference implementation, one to be used as start-point for
- * platform-specific assembler.
+ * reference implementation, one to be used as starting point for
+ * platform-specific assembler. Mentioned numbers apply to compiler
+ * generated code compiled with and without -DOPENSSL_BN_ASM_MONT and
+ * can vary not only from platform to platform, but even for compiler
+ * versions. Assembler vs. assembler improvement coefficients can
+ * [and are known to] differ and are to be documented elsewhere.
*/
-int bn_mul_mont(BN_ULONG *rp, const BN_ULONG *ap, const BN_ULONG *bp, const BN_ULONG *np,BN_ULONG n0, int num)
+int bn_mul_mont(BN_ULONG *rp, const BN_ULONG *ap, const BN_ULONG *bp, const BN_ULONG *np,const BN_ULONG *n0p, int num)
{
- BN_ULONG c0,c1,ml,*tp;
+ BN_ULONG c0,c1,ml,*tp,n0;
#ifdef mul64
BN_ULONG mh;
#endif
int i=0,j;
#if 0 /* template for platform-specific implementation */
- if (ap==bp) return bn_sqr_mont(rp,ap,np,n0,num);
+ if (ap==bp) return bn_sqr_mont(rp,ap,np,n0p,num);
#endif
vp = tp = alloca((num+2)*sizeof(BN_ULONG));
- tp[num] = bn_mul_words(tp,ap,num,bp[0]);
+ n0 = *n0p;
+
+ c0 = 0;
+ ml = bp[0];
+#ifdef mul64
+ mh = HBITS(ml);
+ ml = LBITS(ml);
+ for (j=0;j<num;++j)
+ mul(tp[j],ap[j],ml,mh,c0);
+#else
+ for (j=0;j<num;++j)
+ mul(tp[j],ap[j],ml,c0);
+#endif
+
+ tp[num] = c0;
tp[num+1] = 0;
goto enter;
for(i=0;i<num;i++)
{
- c0 = bn_mul_add_words(tp,ap,num,bp[i]);
+ c0 = 0;
+ ml = bp[i];
+#ifdef mul64
+ mh = HBITS(ml);
+ ml = LBITS(ml);
+ for (j=0;j<num;++j)
+ mul_add(tp[j],ap[j],ml,mh,c0);
+#else
+ for (j=0;j<num;++j)
+ mul_add(tp[j],ap[j],ml,c0);
+#endif
c1 = (tp[num] + c0)&BN_MASK2;
tp[num] = c1;
tp[num+1] = (c1<c0?1:0);
* performed to signal the caller to fall down to alternative/original
* code-path.
*/
-int bn_mul_mont(BN_ULONG *rp, const BN_ULONG *ap, const BN_ULONG *bp, const BN_ULONG *np,BN_ULONG n0, int num)
+int bn_mul_mont(BN_ULONG *rp, const BN_ULONG *ap, const BN_ULONG *bp, const BN_ULONG *np,const BN_ULONG *n0, int num)
{ return 0; }
#endif /* OPENSSL_BN_ASM_MONT */
+#endif
#else /* !BN_MUL_COMBA */
r[15]=bn_mul_add_words(&(r[7]),a,8,b[7]);
}
+#ifdef OPENSSL_NO_ASM
#ifdef OPENSSL_BN_ASM_MONT
-int bn_mul_mont(BN_ULONG *rp, const BN_ULONG *ap, const BN_ULONG *bp, const BN_ULONG *np,BN_ULONG n0, int num)
+#include <alloca.h>
+int bn_mul_mont(BN_ULONG *rp, const BN_ULONG *ap, const BN_ULONG *bp, const BN_ULONG *np,const BN_ULONG *n0p, int num)
{
- BN_ULONG c0,c1,*tp;
+ BN_ULONG c0,c1,*tp,n0=*n0p;
volatile BN_ULONG *vp;
int i=0,j;
for(i=0;i<num;i++)
{
c0 = bn_mul_add_words(tp,ap,num,bp[i]);
- c1 = tp[num] + c0;
+ c1 = (tp[num] + c0)&BN_MASK2;
tp[num] = c1;
tp[num+1] = (c1<c0?1:0);
c0 = bn_mul_add_words(tp,np,num,tp[0]*n0);
- c1 = tp[num] + c0;
+ c1 = (tp[num] + c0)&BN_MASK2;
tp[num] = c1;
tp[num+1] += (c1<c0?1:0);
for(j=0;j<=num;j++) tp[j]=tp[j+1];
return 1;
}
#else
-int bn_mul_mont(BN_ULONG *rp, const BN_ULONG *ap, const BN_ULONG *bp, const BN_ULONG *np,BN_ULONG n0, int num)
+int bn_mul_mont(BN_ULONG *rp, const BN_ULONG *ap, const BN_ULONG *bp, const BN_ULONG *np,const BN_ULONG *n0, int num)
{ return 0; }
#endif /* OPENSSL_BN_ASM_MONT */
+#endif
#endif /* !BN_MUL_COMBA */