diff options
Diffstat (limited to 'src/lib/libcrypto/bn/bn_mul.c')
-rw-r--r-- | src/lib/libcrypto/bn/bn_mul.c | 529 |
1 files changed, 84 insertions, 445 deletions
diff --git a/src/lib/libcrypto/bn/bn_mul.c b/src/lib/libcrypto/bn/bn_mul.c index b03458d002..cb93ac3356 100644 --- a/src/lib/libcrypto/bn/bn_mul.c +++ b/src/lib/libcrypto/bn/bn_mul.c | |||
@@ -56,325 +56,10 @@ | |||
56 | * [including the GNU Public Licence.] | 56 | * [including the GNU Public Licence.] |
57 | */ | 57 | */ |
58 | 58 | ||
59 | #ifndef BN_DEBUG | ||
60 | # undef NDEBUG /* avoid conflicting definitions */ | ||
61 | # define NDEBUG | ||
62 | #endif | ||
63 | |||
64 | #include <stdio.h> | 59 | #include <stdio.h> |
65 | #include <assert.h> | ||
66 | #include "cryptlib.h" | 60 | #include "cryptlib.h" |
67 | #include "bn_lcl.h" | 61 | #include "bn_lcl.h" |
68 | 62 | ||
69 | #if defined(OPENSSL_NO_ASM) || !(defined(__i386) || defined(__i386__)) || defined(__DJGPP__) /* Assembler implementation exists only for x86 */ | ||
70 | /* Here follows specialised variants of bn_add_words() and | ||
71 | bn_sub_words(). They have the property performing operations on | ||
72 | arrays of different sizes. The sizes of those arrays is expressed through | ||
73 | cl, which is the common length ( basicall, min(len(a),len(b)) ), and dl, | ||
74 | which is the delta between the two lengths, calculated as len(a)-len(b). | ||
75 | All lengths are the number of BN_ULONGs... For the operations that require | ||
76 | a result array as parameter, it must have the length cl+abs(dl). | ||
77 | These functions should probably end up in bn_asm.c as soon as there are | ||
78 | assembler counterparts for the systems that use assembler files. */ | ||
79 | |||
80 | BN_ULONG bn_sub_part_words(BN_ULONG *r, | ||
81 | const BN_ULONG *a, const BN_ULONG *b, | ||
82 | int cl, int dl) | ||
83 | { | ||
84 | BN_ULONG c, t; | ||
85 | |||
86 | assert(cl >= 0); | ||
87 | c = bn_sub_words(r, a, b, cl); | ||
88 | |||
89 | if (dl == 0) | ||
90 | return c; | ||
91 | |||
92 | r += cl; | ||
93 | a += cl; | ||
94 | b += cl; | ||
95 | |||
96 | if (dl < 0) | ||
97 | { | ||
98 | #ifdef BN_COUNT | ||
99 | fprintf(stderr, " bn_sub_part_words %d + %d (dl < 0, c = %d)\n", cl, dl, c); | ||
100 | #endif | ||
101 | for (;;) | ||
102 | { | ||
103 | t = b[0]; | ||
104 | r[0] = (0-t-c)&BN_MASK2; | ||
105 | if (t != 0) c=1; | ||
106 | if (++dl >= 0) break; | ||
107 | |||
108 | t = b[1]; | ||
109 | r[1] = (0-t-c)&BN_MASK2; | ||
110 | if (t != 0) c=1; | ||
111 | if (++dl >= 0) break; | ||
112 | |||
113 | t = b[2]; | ||
114 | r[2] = (0-t-c)&BN_MASK2; | ||
115 | if (t != 0) c=1; | ||
116 | if (++dl >= 0) break; | ||
117 | |||
118 | t = b[3]; | ||
119 | r[3] = (0-t-c)&BN_MASK2; | ||
120 | if (t != 0) c=1; | ||
121 | if (++dl >= 0) break; | ||
122 | |||
123 | b += 4; | ||
124 | r += 4; | ||
125 | } | ||
126 | } | ||
127 | else | ||
128 | { | ||
129 | int save_dl = dl; | ||
130 | #ifdef BN_COUNT | ||
131 | fprintf(stderr, " bn_sub_part_words %d + %d (dl > 0, c = %d)\n", cl, dl, c); | ||
132 | #endif | ||
133 | while(c) | ||
134 | { | ||
135 | t = a[0]; | ||
136 | r[0] = (t-c)&BN_MASK2; | ||
137 | if (t != 0) c=0; | ||
138 | if (--dl <= 0) break; | ||
139 | |||
140 | t = a[1]; | ||
141 | r[1] = (t-c)&BN_MASK2; | ||
142 | if (t != 0) c=0; | ||
143 | if (--dl <= 0) break; | ||
144 | |||
145 | t = a[2]; | ||
146 | r[2] = (t-c)&BN_MASK2; | ||
147 | if (t != 0) c=0; | ||
148 | if (--dl <= 0) break; | ||
149 | |||
150 | t = a[3]; | ||
151 | r[3] = (t-c)&BN_MASK2; | ||
152 | if (t != 0) c=0; | ||
153 | if (--dl <= 0) break; | ||
154 | |||
155 | save_dl = dl; | ||
156 | a += 4; | ||
157 | r += 4; | ||
158 | } | ||
159 | if (dl > 0) | ||
160 | { | ||
161 | #ifdef BN_COUNT | ||
162 | fprintf(stderr, " bn_sub_part_words %d + %d (dl > 0, c == 0)\n", cl, dl); | ||
163 | #endif | ||
164 | if (save_dl > dl) | ||
165 | { | ||
166 | switch (save_dl - dl) | ||
167 | { | ||
168 | case 1: | ||
169 | r[1] = a[1]; | ||
170 | if (--dl <= 0) break; | ||
171 | case 2: | ||
172 | r[2] = a[2]; | ||
173 | if (--dl <= 0) break; | ||
174 | case 3: | ||
175 | r[3] = a[3]; | ||
176 | if (--dl <= 0) break; | ||
177 | } | ||
178 | a += 4; | ||
179 | r += 4; | ||
180 | } | ||
181 | } | ||
182 | if (dl > 0) | ||
183 | { | ||
184 | #ifdef BN_COUNT | ||
185 | fprintf(stderr, " bn_sub_part_words %d + %d (dl > 0, copy)\n", cl, dl); | ||
186 | #endif | ||
187 | for(;;) | ||
188 | { | ||
189 | r[0] = a[0]; | ||
190 | if (--dl <= 0) break; | ||
191 | r[1] = a[1]; | ||
192 | if (--dl <= 0) break; | ||
193 | r[2] = a[2]; | ||
194 | if (--dl <= 0) break; | ||
195 | r[3] = a[3]; | ||
196 | if (--dl <= 0) break; | ||
197 | |||
198 | a += 4; | ||
199 | r += 4; | ||
200 | } | ||
201 | } | ||
202 | } | ||
203 | return c; | ||
204 | } | ||
205 | #endif | ||
206 | |||
207 | BN_ULONG bn_add_part_words(BN_ULONG *r, | ||
208 | const BN_ULONG *a, const BN_ULONG *b, | ||
209 | int cl, int dl) | ||
210 | { | ||
211 | BN_ULONG c, l, t; | ||
212 | |||
213 | assert(cl >= 0); | ||
214 | c = bn_add_words(r, a, b, cl); | ||
215 | |||
216 | if (dl == 0) | ||
217 | return c; | ||
218 | |||
219 | r += cl; | ||
220 | a += cl; | ||
221 | b += cl; | ||
222 | |||
223 | if (dl < 0) | ||
224 | { | ||
225 | int save_dl = dl; | ||
226 | #ifdef BN_COUNT | ||
227 | fprintf(stderr, " bn_add_part_words %d + %d (dl < 0, c = %d)\n", cl, dl, c); | ||
228 | #endif | ||
229 | while (c) | ||
230 | { | ||
231 | l=(c+b[0])&BN_MASK2; | ||
232 | c=(l < c); | ||
233 | r[0]=l; | ||
234 | if (++dl >= 0) break; | ||
235 | |||
236 | l=(c+b[1])&BN_MASK2; | ||
237 | c=(l < c); | ||
238 | r[1]=l; | ||
239 | if (++dl >= 0) break; | ||
240 | |||
241 | l=(c+b[2])&BN_MASK2; | ||
242 | c=(l < c); | ||
243 | r[2]=l; | ||
244 | if (++dl >= 0) break; | ||
245 | |||
246 | l=(c+b[3])&BN_MASK2; | ||
247 | c=(l < c); | ||
248 | r[3]=l; | ||
249 | if (++dl >= 0) break; | ||
250 | |||
251 | save_dl = dl; | ||
252 | b+=4; | ||
253 | r+=4; | ||
254 | } | ||
255 | if (dl < 0) | ||
256 | { | ||
257 | #ifdef BN_COUNT | ||
258 | fprintf(stderr, " bn_add_part_words %d + %d (dl < 0, c == 0)\n", cl, dl); | ||
259 | #endif | ||
260 | if (save_dl < dl) | ||
261 | { | ||
262 | switch (dl - save_dl) | ||
263 | { | ||
264 | case 1: | ||
265 | r[1] = b[1]; | ||
266 | if (++dl >= 0) break; | ||
267 | case 2: | ||
268 | r[2] = b[2]; | ||
269 | if (++dl >= 0) break; | ||
270 | case 3: | ||
271 | r[3] = b[3]; | ||
272 | if (++dl >= 0) break; | ||
273 | } | ||
274 | b += 4; | ||
275 | r += 4; | ||
276 | } | ||
277 | } | ||
278 | if (dl < 0) | ||
279 | { | ||
280 | #ifdef BN_COUNT | ||
281 | fprintf(stderr, " bn_add_part_words %d + %d (dl < 0, copy)\n", cl, dl); | ||
282 | #endif | ||
283 | for(;;) | ||
284 | { | ||
285 | r[0] = b[0]; | ||
286 | if (++dl >= 0) break; | ||
287 | r[1] = b[1]; | ||
288 | if (++dl >= 0) break; | ||
289 | r[2] = b[2]; | ||
290 | if (++dl >= 0) break; | ||
291 | r[3] = b[3]; | ||
292 | if (++dl >= 0) break; | ||
293 | |||
294 | b += 4; | ||
295 | r += 4; | ||
296 | } | ||
297 | } | ||
298 | } | ||
299 | else | ||
300 | { | ||
301 | int save_dl = dl; | ||
302 | #ifdef BN_COUNT | ||
303 | fprintf(stderr, " bn_add_part_words %d + %d (dl > 0)\n", cl, dl); | ||
304 | #endif | ||
305 | while (c) | ||
306 | { | ||
307 | t=(a[0]+c)&BN_MASK2; | ||
308 | c=(t < c); | ||
309 | r[0]=t; | ||
310 | if (--dl <= 0) break; | ||
311 | |||
312 | t=(a[1]+c)&BN_MASK2; | ||
313 | c=(t < c); | ||
314 | r[1]=t; | ||
315 | if (--dl <= 0) break; | ||
316 | |||
317 | t=(a[2]+c)&BN_MASK2; | ||
318 | c=(t < c); | ||
319 | r[2]=t; | ||
320 | if (--dl <= 0) break; | ||
321 | |||
322 | t=(a[3]+c)&BN_MASK2; | ||
323 | c=(t < c); | ||
324 | r[3]=t; | ||
325 | if (--dl <= 0) break; | ||
326 | |||
327 | save_dl = dl; | ||
328 | a+=4; | ||
329 | r+=4; | ||
330 | } | ||
331 | #ifdef BN_COUNT | ||
332 | fprintf(stderr, " bn_add_part_words %d + %d (dl > 0, c == 0)\n", cl, dl); | ||
333 | #endif | ||
334 | if (dl > 0) | ||
335 | { | ||
336 | if (save_dl > dl) | ||
337 | { | ||
338 | switch (save_dl - dl) | ||
339 | { | ||
340 | case 1: | ||
341 | r[1] = a[1]; | ||
342 | if (--dl <= 0) break; | ||
343 | case 2: | ||
344 | r[2] = a[2]; | ||
345 | if (--dl <= 0) break; | ||
346 | case 3: | ||
347 | r[3] = a[3]; | ||
348 | if (--dl <= 0) break; | ||
349 | } | ||
350 | a += 4; | ||
351 | r += 4; | ||
352 | } | ||
353 | } | ||
354 | if (dl > 0) | ||
355 | { | ||
356 | #ifdef BN_COUNT | ||
357 | fprintf(stderr, " bn_add_part_words %d + %d (dl > 0, copy)\n", cl, dl); | ||
358 | #endif | ||
359 | for(;;) | ||
360 | { | ||
361 | r[0] = a[0]; | ||
362 | if (--dl <= 0) break; | ||
363 | r[1] = a[1]; | ||
364 | if (--dl <= 0) break; | ||
365 | r[2] = a[2]; | ||
366 | if (--dl <= 0) break; | ||
367 | r[3] = a[3]; | ||
368 | if (--dl <= 0) break; | ||
369 | |||
370 | a += 4; | ||
371 | r += 4; | ||
372 | } | ||
373 | } | ||
374 | } | ||
375 | return c; | ||
376 | } | ||
377 | |||
378 | #ifdef BN_RECURSION | 63 | #ifdef BN_RECURSION |
379 | /* Karatsuba recursive multiplication algorithm | 64 | /* Karatsuba recursive multiplication algorithm |
380 | * (cf. Knuth, The Art of Computer Programming, Vol. 2) */ | 65 | * (cf. Knuth, The Art of Computer Programming, Vol. 2) */ |
@@ -390,15 +75,14 @@ BN_ULONG bn_add_part_words(BN_ULONG *r, | |||
390 | * a[1]*b[1] | 75 | * a[1]*b[1] |
391 | */ | 76 | */ |
392 | void bn_mul_recursive(BN_ULONG *r, BN_ULONG *a, BN_ULONG *b, int n2, | 77 | void bn_mul_recursive(BN_ULONG *r, BN_ULONG *a, BN_ULONG *b, int n2, |
393 | int dna, int dnb, BN_ULONG *t) | 78 | BN_ULONG *t) |
394 | { | 79 | { |
395 | int n=n2/2,c1,c2; | 80 | int n=n2/2,c1,c2; |
396 | int tna=n+dna, tnb=n+dnb; | ||
397 | unsigned int neg,zero; | 81 | unsigned int neg,zero; |
398 | BN_ULONG ln,lo,*p; | 82 | BN_ULONG ln,lo,*p; |
399 | 83 | ||
400 | # ifdef BN_COUNT | 84 | # ifdef BN_COUNT |
401 | fprintf(stderr," bn_mul_recursive %d * %d\n",n2,n2); | 85 | printf(" bn_mul_recursive %d * %d\n",n2,n2); |
402 | # endif | 86 | # endif |
403 | # ifdef BN_MUL_COMBA | 87 | # ifdef BN_MUL_COMBA |
404 | # if 0 | 88 | # if 0 |
@@ -408,40 +92,34 @@ void bn_mul_recursive(BN_ULONG *r, BN_ULONG *a, BN_ULONG *b, int n2, | |||
408 | return; | 92 | return; |
409 | } | 93 | } |
410 | # endif | 94 | # endif |
411 | /* Only call bn_mul_comba 8 if n2 == 8 and the | 95 | if (n2 == 8) |
412 | * two arrays are complete [steve] | ||
413 | */ | ||
414 | if (n2 == 8 && dna == 0 && dnb == 0) | ||
415 | { | 96 | { |
416 | bn_mul_comba8(r,a,b); | 97 | bn_mul_comba8(r,a,b); |
417 | return; | 98 | return; |
418 | } | 99 | } |
419 | # endif /* BN_MUL_COMBA */ | 100 | # endif /* BN_MUL_COMBA */ |
420 | /* Else do normal multiply */ | ||
421 | if (n2 < BN_MUL_RECURSIVE_SIZE_NORMAL) | 101 | if (n2 < BN_MUL_RECURSIVE_SIZE_NORMAL) |
422 | { | 102 | { |
423 | bn_mul_normal(r,a,n2+dna,b,n2+dnb); | 103 | /* This should not happen */ |
424 | if ((dna + dnb) < 0) | 104 | bn_mul_normal(r,a,n2,b,n2); |
425 | memset(&r[2*n2 + dna + dnb], 0, | ||
426 | sizeof(BN_ULONG) * -(dna + dnb)); | ||
427 | return; | 105 | return; |
428 | } | 106 | } |
429 | /* r=(a[0]-a[1])*(b[1]-b[0]) */ | 107 | /* r=(a[0]-a[1])*(b[1]-b[0]) */ |
430 | c1=bn_cmp_part_words(a,&(a[n]),tna,n-tna); | 108 | c1=bn_cmp_words(a,&(a[n]),n); |
431 | c2=bn_cmp_part_words(&(b[n]),b,tnb,tnb-n); | 109 | c2=bn_cmp_words(&(b[n]),b,n); |
432 | zero=neg=0; | 110 | zero=neg=0; |
433 | switch (c1*3+c2) | 111 | switch (c1*3+c2) |
434 | { | 112 | { |
435 | case -4: | 113 | case -4: |
436 | bn_sub_part_words(t, &(a[n]),a, tna,tna-n); /* - */ | 114 | bn_sub_words(t, &(a[n]),a, n); /* - */ |
437 | bn_sub_part_words(&(t[n]),b, &(b[n]),tnb,n-tnb); /* - */ | 115 | bn_sub_words(&(t[n]),b, &(b[n]),n); /* - */ |
438 | break; | 116 | break; |
439 | case -3: | 117 | case -3: |
440 | zero=1; | 118 | zero=1; |
441 | break; | 119 | break; |
442 | case -2: | 120 | case -2: |
443 | bn_sub_part_words(t, &(a[n]),a, tna,tna-n); /* - */ | 121 | bn_sub_words(t, &(a[n]),a, n); /* - */ |
444 | bn_sub_part_words(&(t[n]),&(b[n]),b, tnb,tnb-n); /* + */ | 122 | bn_sub_words(&(t[n]),&(b[n]),b, n); /* + */ |
445 | neg=1; | 123 | neg=1; |
446 | break; | 124 | break; |
447 | case -1: | 125 | case -1: |
@@ -450,22 +128,21 @@ void bn_mul_recursive(BN_ULONG *r, BN_ULONG *a, BN_ULONG *b, int n2, | |||
450 | zero=1; | 128 | zero=1; |
451 | break; | 129 | break; |
452 | case 2: | 130 | case 2: |
453 | bn_sub_part_words(t, a, &(a[n]),tna,n-tna); /* + */ | 131 | bn_sub_words(t, a, &(a[n]),n); /* + */ |
454 | bn_sub_part_words(&(t[n]),b, &(b[n]),tnb,n-tnb); /* - */ | 132 | bn_sub_words(&(t[n]),b, &(b[n]),n); /* - */ |
455 | neg=1; | 133 | neg=1; |
456 | break; | 134 | break; |
457 | case 3: | 135 | case 3: |
458 | zero=1; | 136 | zero=1; |
459 | break; | 137 | break; |
460 | case 4: | 138 | case 4: |
461 | bn_sub_part_words(t, a, &(a[n]),tna,n-tna); | 139 | bn_sub_words(t, a, &(a[n]),n); |
462 | bn_sub_part_words(&(t[n]),&(b[n]),b, tnb,tnb-n); | 140 | bn_sub_words(&(t[n]),&(b[n]),b, n); |
463 | break; | 141 | break; |
464 | } | 142 | } |
465 | 143 | ||
466 | # ifdef BN_MUL_COMBA | 144 | # ifdef BN_MUL_COMBA |
467 | if (n == 4 && dna == 0 && dnb == 0) /* XXX: bn_mul_comba4 could take | 145 | if (n == 4) |
468 | extra args to do this well */ | ||
469 | { | 146 | { |
470 | if (!zero) | 147 | if (!zero) |
471 | bn_mul_comba4(&(t[n2]),t,&(t[n])); | 148 | bn_mul_comba4(&(t[n2]),t,&(t[n])); |
@@ -475,9 +152,7 @@ void bn_mul_recursive(BN_ULONG *r, BN_ULONG *a, BN_ULONG *b, int n2, | |||
475 | bn_mul_comba4(r,a,b); | 152 | bn_mul_comba4(r,a,b); |
476 | bn_mul_comba4(&(r[n2]),&(a[n]),&(b[n])); | 153 | bn_mul_comba4(&(r[n2]),&(a[n]),&(b[n])); |
477 | } | 154 | } |
478 | else if (n == 8 && dna == 0 && dnb == 0) /* XXX: bn_mul_comba8 could | 155 | else if (n == 8) |
479 | take extra args to do this | ||
480 | well */ | ||
481 | { | 156 | { |
482 | if (!zero) | 157 | if (!zero) |
483 | bn_mul_comba8(&(t[n2]),t,&(t[n])); | 158 | bn_mul_comba8(&(t[n2]),t,&(t[n])); |
@@ -492,11 +167,11 @@ void bn_mul_recursive(BN_ULONG *r, BN_ULONG *a, BN_ULONG *b, int n2, | |||
492 | { | 167 | { |
493 | p= &(t[n2*2]); | 168 | p= &(t[n2*2]); |
494 | if (!zero) | 169 | if (!zero) |
495 | bn_mul_recursive(&(t[n2]),t,&(t[n]),n,0,0,p); | 170 | bn_mul_recursive(&(t[n2]),t,&(t[n]),n,p); |
496 | else | 171 | else |
497 | memset(&(t[n2]),0,n2*sizeof(BN_ULONG)); | 172 | memset(&(t[n2]),0,n2*sizeof(BN_ULONG)); |
498 | bn_mul_recursive(r,a,b,n,0,0,p); | 173 | bn_mul_recursive(r,a,b,n,p); |
499 | bn_mul_recursive(&(r[n2]),&(a[n]),&(b[n]),n,dna,dnb,p); | 174 | bn_mul_recursive(&(r[n2]),&(a[n]),&(b[n]),n,p); |
500 | } | 175 | } |
501 | 176 | ||
502 | /* t[32] holds (a[0]-a[1])*(b[1]-b[0]), c1 is the sign | 177 | /* t[32] holds (a[0]-a[1])*(b[1]-b[0]), c1 is the sign |
@@ -545,39 +220,39 @@ void bn_mul_recursive(BN_ULONG *r, BN_ULONG *a, BN_ULONG *b, int n2, | |||
545 | 220 | ||
546 | /* n+tn is the word length | 221 | /* n+tn is the word length |
547 | * t needs to be n*4 is size, as does r */ | 222 | * t needs to be n*4 is size, as does r */ |
548 | void bn_mul_part_recursive(BN_ULONG *r, BN_ULONG *a, BN_ULONG *b, int n, | 223 | void bn_mul_part_recursive(BN_ULONG *r, BN_ULONG *a, BN_ULONG *b, int tn, |
549 | int tna, int tnb, BN_ULONG *t) | 224 | int n, BN_ULONG *t) |
550 | { | 225 | { |
551 | int i,j,n2=n*2; | 226 | int i,j,n2=n*2; |
552 | unsigned int c1,c2,neg,zero; | 227 | unsigned int c1,c2,neg,zero; |
553 | BN_ULONG ln,lo,*p; | 228 | BN_ULONG ln,lo,*p; |
554 | 229 | ||
555 | # ifdef BN_COUNT | 230 | # ifdef BN_COUNT |
556 | fprintf(stderr," bn_mul_part_recursive (%d+%d) * (%d+%d)\n", | 231 | printf(" bn_mul_part_recursive %d * %d\n",tn+n,tn+n); |
557 | tna, n, tnb, n); | ||
558 | # endif | 232 | # endif |
559 | if (n < 8) | 233 | if (n < 8) |
560 | { | 234 | { |
561 | bn_mul_normal(r,a,n+tna,b,n+tnb); | 235 | i=tn+n; |
236 | bn_mul_normal(r,a,i,b,i); | ||
562 | return; | 237 | return; |
563 | } | 238 | } |
564 | 239 | ||
565 | /* r=(a[0]-a[1])*(b[1]-b[0]) */ | 240 | /* r=(a[0]-a[1])*(b[1]-b[0]) */ |
566 | c1=bn_cmp_part_words(a,&(a[n]),tna,n-tna); | 241 | c1=bn_cmp_words(a,&(a[n]),n); |
567 | c2=bn_cmp_part_words(&(b[n]),b,tnb,tnb-n); | 242 | c2=bn_cmp_words(&(b[n]),b,n); |
568 | zero=neg=0; | 243 | zero=neg=0; |
569 | switch (c1*3+c2) | 244 | switch (c1*3+c2) |
570 | { | 245 | { |
571 | case -4: | 246 | case -4: |
572 | bn_sub_part_words(t, &(a[n]),a, tna,tna-n); /* - */ | 247 | bn_sub_words(t, &(a[n]),a, n); /* - */ |
573 | bn_sub_part_words(&(t[n]),b, &(b[n]),tnb,n-tnb); /* - */ | 248 | bn_sub_words(&(t[n]),b, &(b[n]),n); /* - */ |
574 | break; | 249 | break; |
575 | case -3: | 250 | case -3: |
576 | zero=1; | 251 | zero=1; |
577 | /* break; */ | 252 | /* break; */ |
578 | case -2: | 253 | case -2: |
579 | bn_sub_part_words(t, &(a[n]),a, tna,tna-n); /* - */ | 254 | bn_sub_words(t, &(a[n]),a, n); /* - */ |
580 | bn_sub_part_words(&(t[n]),&(b[n]),b, tnb,tnb-n); /* + */ | 255 | bn_sub_words(&(t[n]),&(b[n]),b, n); /* + */ |
581 | neg=1; | 256 | neg=1; |
582 | break; | 257 | break; |
583 | case -1: | 258 | case -1: |
@@ -586,16 +261,16 @@ void bn_mul_part_recursive(BN_ULONG *r, BN_ULONG *a, BN_ULONG *b, int n, | |||
586 | zero=1; | 261 | zero=1; |
587 | /* break; */ | 262 | /* break; */ |
588 | case 2: | 263 | case 2: |
589 | bn_sub_part_words(t, a, &(a[n]),tna,n-tna); /* + */ | 264 | bn_sub_words(t, a, &(a[n]),n); /* + */ |
590 | bn_sub_part_words(&(t[n]),b, &(b[n]),tnb,n-tnb); /* - */ | 265 | bn_sub_words(&(t[n]),b, &(b[n]),n); /* - */ |
591 | neg=1; | 266 | neg=1; |
592 | break; | 267 | break; |
593 | case 3: | 268 | case 3: |
594 | zero=1; | 269 | zero=1; |
595 | /* break; */ | 270 | /* break; */ |
596 | case 4: | 271 | case 4: |
597 | bn_sub_part_words(t, a, &(a[n]),tna,n-tna); | 272 | bn_sub_words(t, a, &(a[n]),n); |
598 | bn_sub_part_words(&(t[n]),&(b[n]),b, tnb,tnb-n); | 273 | bn_sub_words(&(t[n]),&(b[n]),b, n); |
599 | break; | 274 | break; |
600 | } | 275 | } |
601 | /* The zero case isn't yet implemented here. The speedup | 276 | /* The zero case isn't yet implemented here. The speedup |
@@ -614,59 +289,54 @@ void bn_mul_part_recursive(BN_ULONG *r, BN_ULONG *a, BN_ULONG *b, int n, | |||
614 | { | 289 | { |
615 | bn_mul_comba8(&(t[n2]),t,&(t[n])); | 290 | bn_mul_comba8(&(t[n2]),t,&(t[n])); |
616 | bn_mul_comba8(r,a,b); | 291 | bn_mul_comba8(r,a,b); |
617 | bn_mul_normal(&(r[n2]),&(a[n]),tna,&(b[n]),tnb); | 292 | bn_mul_normal(&(r[n2]),&(a[n]),tn,&(b[n]),tn); |
618 | memset(&(r[n2+tna+tnb]),0,sizeof(BN_ULONG)*(n2-tna-tnb)); | 293 | memset(&(r[n2+tn*2]),0,sizeof(BN_ULONG)*(n2-tn*2)); |
619 | } | 294 | } |
620 | else | 295 | else |
621 | { | 296 | { |
622 | p= &(t[n2*2]); | 297 | p= &(t[n2*2]); |
623 | bn_mul_recursive(&(t[n2]),t,&(t[n]),n,0,0,p); | 298 | bn_mul_recursive(&(t[n2]),t,&(t[n]),n,p); |
624 | bn_mul_recursive(r,a,b,n,0,0,p); | 299 | bn_mul_recursive(r,a,b,n,p); |
625 | i=n/2; | 300 | i=n/2; |
626 | /* If there is only a bottom half to the number, | 301 | /* If there is only a bottom half to the number, |
627 | * just do it */ | 302 | * just do it */ |
628 | if (tna > tnb) | 303 | j=tn-i; |
629 | j = tna - i; | ||
630 | else | ||
631 | j = tnb - i; | ||
632 | if (j == 0) | 304 | if (j == 0) |
633 | { | 305 | { |
634 | bn_mul_recursive(&(r[n2]),&(a[n]),&(b[n]), | 306 | bn_mul_recursive(&(r[n2]),&(a[n]),&(b[n]),i,p); |
635 | i,tna-i,tnb-i,p); | ||
636 | memset(&(r[n2+i*2]),0,sizeof(BN_ULONG)*(n2-i*2)); | 307 | memset(&(r[n2+i*2]),0,sizeof(BN_ULONG)*(n2-i*2)); |
637 | } | 308 | } |
638 | else if (j > 0) /* eg, n == 16, i == 8 and tn == 11 */ | 309 | else if (j > 0) /* eg, n == 16, i == 8 and tn == 11 */ |
639 | { | 310 | { |
640 | bn_mul_part_recursive(&(r[n2]),&(a[n]),&(b[n]), | 311 | bn_mul_part_recursive(&(r[n2]),&(a[n]),&(b[n]), |
641 | i,tna-i,tnb-i,p); | 312 | j,i,p); |
642 | memset(&(r[n2+tna+tnb]),0, | 313 | memset(&(r[n2+tn*2]),0, |
643 | sizeof(BN_ULONG)*(n2-tna-tnb)); | 314 | sizeof(BN_ULONG)*(n2-tn*2)); |
644 | } | 315 | } |
645 | else /* (j < 0) eg, n == 16, i == 8 and tn == 5 */ | 316 | else /* (j < 0) eg, n == 16, i == 8 and tn == 5 */ |
646 | { | 317 | { |
647 | memset(&(r[n2]),0,sizeof(BN_ULONG)*n2); | 318 | memset(&(r[n2]),0,sizeof(BN_ULONG)*n2); |
648 | if (tna < BN_MUL_RECURSIVE_SIZE_NORMAL | 319 | if (tn < BN_MUL_RECURSIVE_SIZE_NORMAL) |
649 | && tnb < BN_MUL_RECURSIVE_SIZE_NORMAL) | ||
650 | { | 320 | { |
651 | bn_mul_normal(&(r[n2]),&(a[n]),tna,&(b[n]),tnb); | 321 | bn_mul_normal(&(r[n2]),&(a[n]),tn,&(b[n]),tn); |
652 | } | 322 | } |
653 | else | 323 | else |
654 | { | 324 | { |
655 | for (;;) | 325 | for (;;) |
656 | { | 326 | { |
657 | i/=2; | 327 | i/=2; |
658 | if (i < tna && i < tnb) | 328 | if (i < tn) |
659 | { | 329 | { |
660 | bn_mul_part_recursive(&(r[n2]), | 330 | bn_mul_part_recursive(&(r[n2]), |
661 | &(a[n]),&(b[n]), | 331 | &(a[n]),&(b[n]), |
662 | i,tna-i,tnb-i,p); | 332 | tn-i,i,p); |
663 | break; | 333 | break; |
664 | } | 334 | } |
665 | else if (i <= tna && i <= tnb) | 335 | else if (i == tn) |
666 | { | 336 | { |
667 | bn_mul_recursive(&(r[n2]), | 337 | bn_mul_recursive(&(r[n2]), |
668 | &(a[n]),&(b[n]), | 338 | &(a[n]),&(b[n]), |
669 | i,tna-i,tnb-i,p); | 339 | i,p); |
670 | break; | 340 | break; |
671 | } | 341 | } |
672 | } | 342 | } |
@@ -727,10 +397,10 @@ void bn_mul_low_recursive(BN_ULONG *r, BN_ULONG *a, BN_ULONG *b, int n2, | |||
727 | int n=n2/2; | 397 | int n=n2/2; |
728 | 398 | ||
729 | # ifdef BN_COUNT | 399 | # ifdef BN_COUNT |
730 | fprintf(stderr," bn_mul_low_recursive %d * %d\n",n2,n2); | 400 | printf(" bn_mul_low_recursive %d * %d\n",n2,n2); |
731 | # endif | 401 | # endif |
732 | 402 | ||
733 | bn_mul_recursive(r,a,b,n,0,0,&(t[0])); | 403 | bn_mul_recursive(r,a,b,n,&(t[0])); |
734 | if (n >= BN_MUL_LOW_RECURSIVE_SIZE_NORMAL) | 404 | if (n >= BN_MUL_LOW_RECURSIVE_SIZE_NORMAL) |
735 | { | 405 | { |
736 | bn_mul_low_recursive(&(t[0]),&(a[0]),&(b[n]),n,&(t[n2])); | 406 | bn_mul_low_recursive(&(t[0]),&(a[0]),&(b[n]),n,&(t[n2])); |
@@ -761,7 +431,7 @@ void bn_mul_high(BN_ULONG *r, BN_ULONG *a, BN_ULONG *b, BN_ULONG *l, int n2, | |||
761 | BN_ULONG ll,lc,*lp,*mp; | 431 | BN_ULONG ll,lc,*lp,*mp; |
762 | 432 | ||
763 | # ifdef BN_COUNT | 433 | # ifdef BN_COUNT |
764 | fprintf(stderr," bn_mul_high %d * %d\n",n2,n2); | 434 | printf(" bn_mul_high %d * %d\n",n2,n2); |
765 | # endif | 435 | # endif |
766 | n=n2/2; | 436 | n=n2/2; |
767 | 437 | ||
@@ -814,8 +484,8 @@ void bn_mul_high(BN_ULONG *r, BN_ULONG *a, BN_ULONG *b, BN_ULONG *l, int n2, | |||
814 | else | 484 | else |
815 | # endif | 485 | # endif |
816 | { | 486 | { |
817 | bn_mul_recursive(&(t[0]),&(r[0]),&(r[n]),n,0,0,&(t[n2])); | 487 | bn_mul_recursive(&(t[0]),&(r[0]),&(r[n]),n,&(t[n2])); |
818 | bn_mul_recursive(r,&(a[n]),&(b[n]),n,0,0,&(t[n2])); | 488 | bn_mul_recursive(r,&(a[n]),&(b[n]),n,&(t[n2])); |
819 | } | 489 | } |
820 | 490 | ||
821 | /* s0 == low(al*bl) | 491 | /* s0 == low(al*bl) |
@@ -940,19 +610,19 @@ void bn_mul_high(BN_ULONG *r, BN_ULONG *a, BN_ULONG *b, BN_ULONG *l, int n2, | |||
940 | 610 | ||
941 | int BN_mul(BIGNUM *r, const BIGNUM *a, const BIGNUM *b, BN_CTX *ctx) | 611 | int BN_mul(BIGNUM *r, const BIGNUM *a, const BIGNUM *b, BN_CTX *ctx) |
942 | { | 612 | { |
943 | int ret=0; | ||
944 | int top,al,bl; | 613 | int top,al,bl; |
945 | BIGNUM *rr; | 614 | BIGNUM *rr; |
615 | int ret = 0; | ||
946 | #if defined(BN_MUL_COMBA) || defined(BN_RECURSION) | 616 | #if defined(BN_MUL_COMBA) || defined(BN_RECURSION) |
947 | int i; | 617 | int i; |
948 | #endif | 618 | #endif |
949 | #ifdef BN_RECURSION | 619 | #ifdef BN_RECURSION |
950 | BIGNUM *t=NULL; | 620 | BIGNUM *t; |
951 | int j=0,k; | 621 | int j,k; |
952 | #endif | 622 | #endif |
953 | 623 | ||
954 | #ifdef BN_COUNT | 624 | #ifdef BN_COUNT |
955 | fprintf(stderr,"BN_mul %d * %d\n",a->top,b->top); | 625 | printf("BN_mul %d * %d\n",a->top,b->top); |
956 | #endif | 626 | #endif |
957 | 627 | ||
958 | bn_check_top(a); | 628 | bn_check_top(a); |
@@ -1005,55 +675,21 @@ int BN_mul(BIGNUM *r, const BIGNUM *a, const BIGNUM *b, BN_CTX *ctx) | |||
1005 | #ifdef BN_RECURSION | 675 | #ifdef BN_RECURSION |
1006 | if ((al >= BN_MULL_SIZE_NORMAL) && (bl >= BN_MULL_SIZE_NORMAL)) | 676 | if ((al >= BN_MULL_SIZE_NORMAL) && (bl >= BN_MULL_SIZE_NORMAL)) |
1007 | { | 677 | { |
1008 | if (i >= -1 && i <= 1) | 678 | if (i == 1 && !BN_get_flags(b,BN_FLG_STATIC_DATA) && bl<b->dmax) |
1009 | { | 679 | { |
1010 | int sav_j =0; | 680 | #if 0 /* tribute to const-ification, bl<b->dmax above covers for this */ |
1011 | /* Find out the power of two lower or equal | 681 | if (bn_wexpand(b,al) == NULL) goto err; |
1012 | to the longest of the two numbers */ | 682 | #endif |
1013 | if (i >= 0) | 683 | b->d[bl]=0; |
1014 | { | ||
1015 | j = BN_num_bits_word((BN_ULONG)al); | ||
1016 | } | ||
1017 | if (i == -1) | ||
1018 | { | ||
1019 | j = BN_num_bits_word((BN_ULONG)bl); | ||
1020 | } | ||
1021 | sav_j = j; | ||
1022 | j = 1<<(j-1); | ||
1023 | assert(j <= al || j <= bl); | ||
1024 | k = j+j; | ||
1025 | t = BN_CTX_get(ctx); | ||
1026 | if (al > j || bl > j) | ||
1027 | { | ||
1028 | bn_wexpand(t,k*4); | ||
1029 | bn_wexpand(rr,k*4); | ||
1030 | bn_mul_part_recursive(rr->d,a->d,b->d, | ||
1031 | j,al-j,bl-j,t->d); | ||
1032 | } | ||
1033 | else /* al <= j || bl <= j */ | ||
1034 | { | ||
1035 | bn_wexpand(t,k*2); | ||
1036 | bn_wexpand(rr,k*2); | ||
1037 | bn_mul_recursive(rr->d,a->d,b->d, | ||
1038 | j,al-j,bl-j,t->d); | ||
1039 | } | ||
1040 | rr->top=top; | ||
1041 | goto end; | ||
1042 | } | ||
1043 | #if 0 | ||
1044 | if (i == 1 && !BN_get_flags(b,BN_FLG_STATIC_DATA)) | ||
1045 | { | ||
1046 | BIGNUM *tmp_bn = (BIGNUM *)b; | ||
1047 | if (bn_wexpand(tmp_bn,al) == NULL) goto err; | ||
1048 | tmp_bn->d[bl]=0; | ||
1049 | bl++; | 684 | bl++; |
1050 | i--; | 685 | i--; |
1051 | } | 686 | } |
1052 | else if (i == -1 && !BN_get_flags(a,BN_FLG_STATIC_DATA)) | 687 | else if (i == -1 && !BN_get_flags(a,BN_FLG_STATIC_DATA) && al<a->dmax) |
1053 | { | 688 | { |
1054 | BIGNUM *tmp_bn = (BIGNUM *)a; | 689 | #if 0 /* tribute to const-ification, al<a->dmax above covers for this */ |
1055 | if (bn_wexpand(tmp_bn,bl) == NULL) goto err; | 690 | if (bn_wexpand(a,bl) == NULL) goto err; |
1056 | tmp_bn->d[al]=0; | 691 | #endif |
692 | a->d[al]=0; | ||
1057 | al++; | 693 | al++; |
1058 | i++; | 694 | i++; |
1059 | } | 695 | } |
@@ -1070,17 +706,26 @@ int BN_mul(BIGNUM *r, const BIGNUM *a, const BIGNUM *b, BN_CTX *ctx) | |||
1070 | if (bn_wexpand(t,k*2) == NULL) goto err; | 706 | if (bn_wexpand(t,k*2) == NULL) goto err; |
1071 | if (bn_wexpand(rr,k*2) == NULL) goto err; | 707 | if (bn_wexpand(rr,k*2) == NULL) goto err; |
1072 | bn_mul_recursive(rr->d,a->d,b->d,al,t->d); | 708 | bn_mul_recursive(rr->d,a->d,b->d,al,t->d); |
709 | rr->top=top; | ||
710 | goto end; | ||
1073 | } | 711 | } |
712 | #if 0 /* tribute to const-ification, rsa/dsa performance is not affected */ | ||
1074 | else | 713 | else |
1075 | { | 714 | { |
1076 | if (bn_wexpand(t,k*4) == NULL) goto err; | 715 | if (bn_wexpand(a,k) == NULL ) goto err; |
1077 | if (bn_wexpand(rr,k*4) == NULL) goto err; | 716 | if (bn_wexpand(b,k) == NULL ) goto err; |
717 | if (bn_wexpand(t,k*4) == NULL ) goto err; | ||
718 | if (bn_wexpand(rr,k*4) == NULL ) goto err; | ||
719 | for (i=a->top; i<k; i++) | ||
720 | a->d[i]=0; | ||
721 | for (i=b->top; i<k; i++) | ||
722 | b->d[i]=0; | ||
1078 | bn_mul_part_recursive(rr->d,a->d,b->d,al-j,j,t->d); | 723 | bn_mul_part_recursive(rr->d,a->d,b->d,al-j,j,t->d); |
1079 | } | 724 | } |
1080 | rr->top=top; | 725 | rr->top=top; |
1081 | goto end; | 726 | goto end; |
1082 | } | ||
1083 | #endif | 727 | #endif |
728 | } | ||
1084 | } | 729 | } |
1085 | #endif /* BN_RECURSION */ | 730 | #endif /* BN_RECURSION */ |
1086 | if (bn_wexpand(rr,top) == NULL) goto err; | 731 | if (bn_wexpand(rr,top) == NULL) goto err; |
@@ -1103,7 +748,7 @@ void bn_mul_normal(BN_ULONG *r, BN_ULONG *a, int na, BN_ULONG *b, int nb) | |||
1103 | BN_ULONG *rr; | 748 | BN_ULONG *rr; |
1104 | 749 | ||
1105 | #ifdef BN_COUNT | 750 | #ifdef BN_COUNT |
1106 | fprintf(stderr," bn_mul_normal %d * %d\n",na,nb); | 751 | printf(" bn_mul_normal %d * %d\n",na,nb); |
1107 | #endif | 752 | #endif |
1108 | 753 | ||
1109 | if (na < nb) | 754 | if (na < nb) |
@@ -1116,13 +761,7 @@ void bn_mul_normal(BN_ULONG *r, BN_ULONG *a, int na, BN_ULONG *b, int nb) | |||
1116 | 761 | ||
1117 | } | 762 | } |
1118 | rr= &(r[na]); | 763 | rr= &(r[na]); |
1119 | if (nb <= 0) | 764 | rr[0]=bn_mul_words(r,a,na,b[0]); |
1120 | { | ||
1121 | (void)bn_mul_words(r,a,na,0); | ||
1122 | return; | ||
1123 | } | ||
1124 | else | ||
1125 | rr[0]=bn_mul_words(r,a,na,b[0]); | ||
1126 | 765 | ||
1127 | for (;;) | 766 | for (;;) |
1128 | { | 767 | { |
@@ -1143,7 +782,7 @@ void bn_mul_normal(BN_ULONG *r, BN_ULONG *a, int na, BN_ULONG *b, int nb) | |||
1143 | void bn_mul_low_normal(BN_ULONG *r, BN_ULONG *a, BN_ULONG *b, int n) | 782 | void bn_mul_low_normal(BN_ULONG *r, BN_ULONG *a, BN_ULONG *b, int n) |
1144 | { | 783 | { |
1145 | #ifdef BN_COUNT | 784 | #ifdef BN_COUNT |
1146 | fprintf(stderr," bn_mul_low_normal %d * %d\n",n,n); | 785 | printf(" bn_mul_low_normal %d * %d\n",n,n); |
1147 | #endif | 786 | #endif |
1148 | bn_mul_words(r,a,n,b[0]); | 787 | bn_mul_words(r,a,n,b[0]); |
1149 | 788 | ||