1 /* crypto/bn/bn_lcl.h */
2 /* Copyright (C) 1995-1998 Eric Young (eay@cryptsoft.com)
5 * This package is an SSL implementation written
6 * by Eric Young (eay@cryptsoft.com).
7 * The implementation was written so as to conform with Netscapes SSL.
9 * This library is free for commercial and non-commercial use as long as
10 * the following conditions are aheared to. The following conditions
11 * apply to all code found in this distribution, be it the RC4, RSA,
12 * lhash, DES, etc., code; not just the SSL code. The SSL documentation
13 * included with this distribution is covered by the same copyright terms
14 * except that the holder is Tim Hudson (tjh@cryptsoft.com).
16 * Copyright remains Eric Young's, and as such any Copyright notices in
17 * the code are not to be removed.
18 * If this package is used in a product, Eric Young should be given attribution
19 * as the author of the parts of the library used.
20 * This can be in the form of a textual message at program startup or
21 * in documentation (online or textual) provided with the package.
23 * Redistribution and use in source and binary forms, with or without
24 * modification, are permitted provided that the following conditions
26 * 1. Redistributions of source code must retain the copyright
27 * notice, this list of conditions and the following disclaimer.
28 * 2. Redistributions in binary form must reproduce the above copyright
29 * notice, this list of conditions and the following disclaimer in the
30 * documentation and/or other materials provided with the distribution.
31 * 3. All advertising materials mentioning features or use of this software
32 * must display the following acknowledgement:
33 * "This product includes cryptographic software written by
34 * Eric Young (eay@cryptsoft.com)"
35 * The word 'cryptographic' can be left out if the rouines from the library
36 * being used are not cryptographic related :-).
37 * 4. If you include any Windows specific code (or a derivative thereof) from
38 * the apps directory (application code) you must include an acknowledgement:
39 * "This product includes software written by Tim Hudson (tjh@cryptsoft.com)"
41 * THIS SOFTWARE IS PROVIDED BY ERIC YOUNG ``AS IS'' AND
42 * ANY EXPRESS OR IMPLIED WARRANTIES, INCLUDING, BUT NOT LIMITED TO, THE
43 * IMPLIED WARRANTIES OF MERCHANTABILITY AND FITNESS FOR A PARTICULAR PURPOSE
44 * ARE DISCLAIMED. IN NO EVENT SHALL THE AUTHOR OR CONTRIBUTORS BE LIABLE
45 * FOR ANY DIRECT, INDIRECT, INCIDENTAL, SPECIAL, EXEMPLARY, OR CONSEQUENTIAL
46 * DAMAGES (INCLUDING, BUT NOT LIMITED TO, PROCUREMENT OF SUBSTITUTE GOODS
47 * OR SERVICES; LOSS OF USE, DATA, OR PROFITS; OR BUSINESS INTERRUPTION)
48 * HOWEVER CAUSED AND ON ANY THEORY OF LIABILITY, WHETHER IN CONTRACT, STRICT
49 * LIABILITY, OR TORT (INCLUDING NEGLIGENCE OR OTHERWISE) ARISING IN ANY WAY
50 * OUT OF THE USE OF THIS SOFTWARE, EVEN IF ADVISED OF THE POSSIBILITY OF
53 * The licence and distribution terms for any publically available version or
54 * derivative of this code cannot be changed. i.e. this code cannot simply be
55 * copied and put under another distribution licence
56 * [including the GNU Public Licence.]
59 #ifndef HEADER_BN_LCL_H
60 #define HEADER_BN_LCL_H
62 #include <openssl/bn.h>
68 /* Pentium pro 16,16,16,32,64 */
69 /* Alpha 16,16,16,16.64 */
70 #define BN_MULL_SIZE_NORMAL (16) /* 32 */
71 #define BN_MUL_RECURSIVE_SIZE_NORMAL (16) /* 32 less than */
72 #define BN_SQR_RECURSIVE_SIZE_NORMAL (16) /* 32 */
73 #define BN_MUL_LOW_RECURSIVE_SIZE_NORMAL (32) /* 32 */
74 #define BN_MONT_CTX_SET_SIZE_WORD (64) /* 32 */
78 /* #define bn_mul_comba8(r,a,b) bn_mul_normal(r,a,8,b,8) */
79 /* #define bn_mul_comba4(r,a,b) bn_mul_normal(r,a,4,b,4) */
83 /* This is probably faster than using the C code - I need to check */
84 #define bn_sqr_comba8(r,a) bn_mul_normal(r,a,8,a,8)
85 #define bn_sqr_comba4(r,a) bn_mul_normal(r,a,4,a,4)
89 #if !defined(NO_ASM) && !defined(PEDANTIC)
91 * BN_UMULT_HIGH section.
93 * No, I'm not trying to overwhelm you when stating that the
94 * product of N-bit numbers is 2*N bits wide:-) No, I don't expect
95 * you to be impressed when I say that if the compiler doesn't
96 * support 2*N integer type, then you have to replace every N*N
97 * multiplication with 4 (N/2)*(N/2) accompanied by some shifts
98 * and additions which unavoidably results in severe performance
99 * penalties. Of course provided that the hardware is capable of
100 * producing 2*N result... That's when you normally start
101 * considering assembler implementation. However! It should be
102 * pointed out that some CPUs (most notably Alpha, PowerPC and
103 * upcoming IA-64 family:-) provide *separate* instruction
104 * calculating the upper half of the product placing the result
105 * into a general purpose register. Now *if* the compiler supports
106 * inline assembler, then it's not impossible to implement the
107 * "bignum" routines (and have the compiler optimize 'em)
108 * exhibiting "native" performance in C. That's what BN_UMULT_HIGH
111 * <appro@fy.chalmers.se>
113 # if defined(__alpha) && (defined(SIXTY_FOUR_BIT_LONG) || defined(SIXTY_FOUR_BIT))
116 # define BN_UMULT_HIGH(a,b) (BN_ULONG)asm("umulh %a0,%a1,%v0",(a),(b))
117 # elif defined(__GNUC__)
118 # define BN_UMULT_HIGH(a,b) ({ \
119 register BN_ULONG ret; \
120 asm ("umulh %1,%2,%0" \
124 # endif /* compiler */
125 # elif defined(_ARCH_PPC) && defined(__64BIT__) && defined(SIXTY_FOUR_BIT_LONG)
126 # if defined(__GNUC__)
127 # define BN_UMULT_HIGH(a,b) ({ \
128 register BN_ULONG ret; \
129 asm ("mulhdu %0,%1,%2" \
133 # endif /* compiler */
137 /*************************************************************
138 * Using the long long type
140 #define Lw(t) (((BN_ULONG)(t))&BN_MASK2)
141 #define Hw(t) (((BN_ULONG)((t)>>BN_BITS2))&BN_MASK2)
143 /* These are used for internal error checking and are not normally used */
145 #define bn_check_top(a) \
146 { if (((a)->top < 0) || ((a)->top > (a)->max)) \
147 { char *nullp=NULL; *nullp='z'; } }
148 #define bn_check_num(a) if ((a) < 0) { char *nullp=NULL; *nullp='z'; }
150 #define bn_check_top(a)
151 #define bn_check_num(a)
154 /* This macro is to add extra stuff for development checking */
156 #define bn_set_max(r) ((r)->max=(r)->top,BN_set_flags((r),BN_FLG_STATIC_DATA))
158 #define bn_set_max(r)
161 /* These macros are used to 'take' a section of a bignum for read only use */
162 #define bn_set_low(r,a,n) \
164 (r)->top=((a)->top > (n))?(n):(a)->top; \
167 (r)->flags|=BN_FLG_STATIC_DATA; \
171 #define bn_set_high(r,a,n) \
173 if ((a)->top > (n)) \
175 (r)->top=(a)->top-n; \
176 (r)->d= &((a)->d[n]); \
181 (r)->flags|=BN_FLG_STATIC_DATA; \
185 /* #define bn_expand(n,b) ((((b)/BN_BITS2) <= (n)->max)?(n):bn_expand2((n),(b))) */
188 #define mul_add(r,a,w,c) { \
190 t=(BN_ULLONG)w * (a) + (r) + (c); \
195 #define mul(r,a,w,c) { \
197 t=(BN_ULLONG)w * (a) + (c); \
202 #define sqr(r0,r1,a) { \
204 t=(BN_ULLONG)(a)*(a); \
209 #elif defined(BN_UMULT_HIGH)
210 #define mul_add(r,a,w,c) { \
211 BN_ULONG high,low,ret,tmp=(a); \
213 high= BN_UMULT_HIGH(w,tmp); \
216 (c) = (ret<(c))?1:0; \
219 (c) += (ret<low)?1:0; \
223 #define mul(r,a,w,c) { \
224 BN_ULONG high,low,ret,ta=(a); \
226 high= BN_UMULT_HIGH(w,ta); \
229 (c) += (ret<low)?1:0; \
233 #define sqr(r0,r1,a) { \
236 (r1) = BN_UMULT_HIGH(tmp,tmp); \
240 /*************************************************************
244 #define LBITS(a) ((a)&BN_MASK2l)
245 #define HBITS(a) (((a)>>BN_BITS4)&BN_MASK2l)
246 #define L2HBITS(a) ((BN_ULONG)((a)&BN_MASK2l)<<BN_BITS4)
248 #define LLBITS(a) ((a)&BN_MASKl)
249 #define LHBITS(a) (((a)>>BN_BITS2)&BN_MASKl)
250 #define LL2HBITS(a) ((BN_ULLONG)((a)&BN_MASKl)<<BN_BITS2)
252 #define mul64(l,h,bl,bh) \
254 BN_ULONG m,m1,lt,ht; \
262 m=(m+m1)&BN_MASK2; if (m < m1) ht+=L2HBITS(1L); \
265 lt=(lt+m1)&BN_MASK2; if (lt < m1) ht++; \
270 #define sqr64(lo,ho,in) \
280 h+=(m&BN_MASK2h1)>>(BN_BITS4-1); \
281 m =(m&BN_MASK2l)<<(BN_BITS4+1); \
282 l=(l+m)&BN_MASK2; if (l < m) h++; \
287 #define mul_add(r,a,bl,bh,c) { \
293 mul64(l,h,(bl),(bh)); \
295 /* non-multiply part */ \
296 l=(l+(c))&BN_MASK2; if (l < (c)) h++; \
298 l=(l+(c))&BN_MASK2; if (l < (c)) h++; \
303 #define mul(r,a,bl,bh,c) { \
309 mul64(l,h,(bl),(bh)); \
311 /* non-multiply part */ \
312 l+=(c); if ((l&BN_MASK2) < (c)) h++; \
319 OPENSSL_EXTERN int bn_limit_bits;
320 OPENSSL_EXTERN int bn_limit_num; /* (1<<bn_limit_bits) */
321 /* Recursive 'low' limit */
322 OPENSSL_EXTERN int bn_limit_bits_low;
323 OPENSSL_EXTERN int bn_limit_num_low; /* (1<<bn_limit_bits_low) */
324 /* Do modified 'high' part calculation' */
325 OPENSSL_EXTERN int bn_limit_bits_high;
326 OPENSSL_EXTERN int bn_limit_num_high; /* (1<<bn_limit_bits_high) */
327 OPENSSL_EXTERN int bn_limit_bits_mont;
328 OPENSSL_EXTERN int bn_limit_num_mont; /* (1<<bn_limit_bits_mont) */
330 BIGNUM *bn_expand2(BIGNUM *b, int bits);
332 void bn_mul_normal(BN_ULONG *r,BN_ULONG *a,int na,BN_ULONG *b,int nb);
333 void bn_mul_comba8(BN_ULONG *r,BN_ULONG *a,BN_ULONG *b);
334 void bn_mul_comba4(BN_ULONG *r,BN_ULONG *a,BN_ULONG *b);
335 void bn_sqr_normal(BN_ULONG *r, BN_ULONG *a, int n, BN_ULONG *tmp);
336 void bn_sqr_comba8(BN_ULONG *r,BN_ULONG *a);
337 void bn_sqr_comba4(BN_ULONG *r,BN_ULONG *a);
338 int bn_cmp_words(BN_ULONG *a,BN_ULONG *b,int n);
339 void bn_mul_recursive(BN_ULONG *r,BN_ULONG *a,BN_ULONG *b,int n2,BN_ULONG *t);
340 void bn_mul_part_recursive(BN_ULONG *r,BN_ULONG *a,BN_ULONG *b,
341 int tn, int n,BN_ULONG *t);
342 void bn_sqr_recursive(BN_ULONG *r,BN_ULONG *a, int n2, BN_ULONG *t);
343 void bn_mul_low_normal(BN_ULONG *r,BN_ULONG *a,BN_ULONG *b, int n);
344 void bn_mul_low_recursive(BN_ULONG *r,BN_ULONG *a,BN_ULONG *b,int n2,
346 void bn_mul_high(BN_ULONG *r,BN_ULONG *a,BN_ULONG *b,BN_ULONG *l,int n2,