forked from jancarlsson/snarklib
-
Notifications
You must be signed in to change notification settings - Fork 0
/
Copy pathAsmMacros.hpp
383 lines (354 loc) · 18.8 KB
/
AsmMacros.hpp
1
2
3
4
5
6
7
8
9
10
11
12
13
14
15
16
17
18
19
20
21
22
23
24
25
26
27
28
29
30
31
32
33
34
35
36
37
38
39
40
41
42
43
44
45
46
47
48
49
50
51
52
53
54
55
56
57
58
59
60
61
62
63
64
65
66
67
68
69
70
71
72
73
74
75
76
77
78
79
80
81
82
83
84
85
86
87
88
89
90
91
92
93
94
95
96
97
98
99
100
101
102
103
104
105
106
107
108
109
110
111
112
113
114
115
116
117
118
119
120
121
122
123
124
125
126
127
128
129
130
131
132
133
134
135
136
137
138
139
140
141
142
143
144
145
146
147
148
149
150
151
152
153
154
155
156
157
158
159
160
161
162
163
164
165
166
167
168
169
170
171
172
173
174
175
176
177
178
179
180
181
182
183
184
185
186
187
188
189
190
191
192
193
194
195
196
197
198
199
200
201
202
203
204
205
206
207
208
209
210
211
212
213
214
215
216
217
218
219
220
221
222
223
224
225
226
227
228
229
230
231
232
233
234
235
236
237
238
239
240
241
242
243
244
245
246
247
248
249
250
251
252
253
254
255
256
257
258
259
260
261
262
263
264
265
266
267
268
269
270
271
272
273
274
275
276
277
278
279
280
281
282
283
284
285
286
287
288
289
290
291
292
293
294
295
296
297
298
299
300
301
302
303
304
305
306
307
308
309
310
311
312
313
314
315
316
317
318
319
320
321
322
323
324
325
326
327
328
329
330
331
332
333
334
335
336
337
338
339
340
341
342
343
344
345
346
347
348
349
350
351
352
353
354
355
356
357
358
359
360
361
362
363
364
365
366
367
368
369
370
371
372
373
374
375
376
377
378
379
380
381
382
383
#ifndef _SNARKLIB_ASM_MACROS_HPP_
#define _SNARKLIB_ASM_MACROS_HPP_
namespace snarklib {
////////////////////////////////////////////////////////////////////////////////
// From libsnark file: src/algebra/fields/fp_aux.tcc
// There are no changes or modifications.
//
#define STR_HELPER(x) #x
#define STR(x) STR_HELPER(x)
/* addq is faster than adcq, even if preceded by clc */
#define ADD_FIRSTADD \
"movq (%[B]), %%rax \n\t" \
"addq %%rax, (%[A]) \n\t"
#define ADD_NEXTADD(ofs) \
"movq " STR(ofs) "(%[B]), %%rax \n\t" \
"adcq %%rax, " STR(ofs) "(%[A]) \n\t"
#define ADD_CMP(ofs) \
"movq " STR(ofs) "(%[mod]), %%rax \n\t" \
"cmpq %%rax, " STR(ofs) "(%[A]) \n\t" \
"jb done%= \n\t" \
"ja subtract%= \n\t"
#define ADD_FIRSTSUB \
"movq (%[mod]), %%rax \n\t" \
"subq %%rax, (%[A]) \n\t"
#define ADD_FIRSTSUB \
"movq (%[mod]), %%rax \n\t" \
"subq %%rax, (%[A]) \n\t"
#define ADD_NEXTSUB(ofs) \
"movq " STR(ofs) "(%[mod]), %%rax \n\t" \
"sbbq %%rax, " STR(ofs) "(%[A]) \n\t"
#define SUB_FIRSTSUB \
"movq (%[B]), %%rax\n\t" \
"subq %%rax, (%[A])\n\t"
#define SUB_NEXTSUB(ofs) \
"movq " STR(ofs) "(%[B]), %%rax\n\t" \
"sbbq %%rax, " STR(ofs) "(%[A])\n\t"
#define SUB_FIRSTADD \
"movq (%[mod]), %%rax\n\t" \
"addq %%rax, (%[A])\n\t"
#define SUB_NEXTADD(ofs) \
"movq " STR(ofs) "(%[mod]), %%rax\n\t" \
"adcq %%rax, " STR(ofs) "(%[A])\n\t"
#define MONT_CMP(ofs) \
"movq " STR(ofs) "(%[M]), %%rax \n\t" \
"cmpq %%rax, " STR(ofs) "(%[tmp]) \n\t" \
"jb done%= \n\t" \
"ja subtract%= \n\t"
#define MONT_FIRSTSUB \
"movq (%[M]), %%rax \n\t" \
"subq %%rax, (%[tmp]) \n\t"
#define MONT_NEXTSUB(ofs) \
"movq " STR(ofs) "(%[M]), %%rax \n\t" \
"sbbq %%rax, " STR(ofs) "(%[tmp]) \n\t"
/*
The x86-64 Montgomery multiplication here is similar
to Algorithm 2 (CIOS method) in http://eprint.iacr.org/2012/140.pdf
and the PowerPC pseudocode of gmp-ecm library (c) Paul Zimmermann and Alexander Kruppa
(see comments on top of powerpc64/mulredc.m4).
*/
#define MONT_PRECOMPUTE \
"xorq %[cy], %[cy] \n\t" \
"movq 0(%[A]), %%rax \n\t" \
"mulq 0(%[B]) \n\t" \
"movq %%rax, %[T0] \n\t" \
"movq %%rdx, %[T1] # T1:T0 <- A[0] * B[0] \n\t" \
"mulq %[inv] \n\t" \
"movq %%rax, %[u] # u <- T0 * inv \n\t" \
"mulq 0(%[M]) \n\t" \
"addq %[T0], %%rax \n\t" \
"adcq %%rdx, %[T1] \n\t" \
"adcq $0, %[cy] # cy:T1 <- (M[0]*u + T1 * b + T0) / b\n\t"
#define MONT_FIRSTITER(j) \
"xorq %[T0], %[T0] \n\t" \
"movq 0(%[A]), %%rax \n\t" \
"mulq " STR((j*8)) "(%[B]) \n\t" \
"addq %[T1], %%rax \n\t" \
"movq %%rax, " STR(((j-1)*8)) "(%[tmp]) \n\t" \
"adcq $0, %%rdx \n\t" \
"movq %%rdx, %[T1] # now T1:tmp[j-1] <-- X[0] * Y[j] + T1\n\t" \
"movq " STR((j*8)) "(%[M]), %%rax \n\t" \
"mulq %[u] \n\t" \
"addq %%rax, " STR(((j-1)*8)) "(%[tmp]) \n\t" \
"adcq %[cy], %%rdx \n\t" \
"adcq $0, %[T0] \n\t" \
"xorq %[cy], %[cy] \n\t" \
"addq %%rdx, %[T1] \n\t" \
"adcq %[T0], %[cy] # cy:T1:tmp[j-1] <---- (X[0] * Y[j] + T1) + (M[j] * u + cy * b) \n\t"
#define MONT_ITERFIRST(i) \
"xorq %[cy], %[cy] \n\t" \
"movq " STR((i*8)) "(%[A]), %%rax \n\t" \
"mulq 0(%[B]) \n\t" \
"addq 0(%[tmp]), %%rax \n\t" \
"adcq 8(%[tmp]), %%rdx \n\t" \
"adcq $0, %[cy] \n\t" \
"movq %%rax, %[T0] \n\t" \
"movq %%rdx, %[T1] # cy:T1:T0 <- A[i] * B[0] + tmp[1] * b + tmp[0]\n\t" \
"mulq %[inv] \n\t" \
"movq %%rax, %[u] # u <- T0 * inv\n\t" \
"mulq 0(%[M]) \n\t" \
"addq %[T0], %%rax \n\t" \
"adcq %%rdx, %[T1] \n\t" \
"adcq $0, %[cy] # cy:T1 <- (M[0]*u + cy * b * b + T1 * b + T0) / b\n\t"
#define MONT_ITERITER(i, j) \
"xorq %[T0], %[T0] \n\t" \
"movq " STR((i*8)) "(%[A]), %%rax \n\t" \
"mulq " STR((j*8)) "(%[B]) \n\t" \
"addq %[T1], %%rax \n\t" \
"movq %%rax, " STR(((j-1)*8)) "(%[tmp]) \n\t" \
"adcq $0, %%rdx \n\t" \
"movq %%rdx, %[T1] # now T1:tmp[j-1] <-- X[i] * Y[j] + T1 \n\t" \
"movq " STR((j*8)) "(%[M]), %%rax \n\t" \
"mulq %[u] \n\t" \
"addq %%rax, " STR(((j-1)*8)) "(%[tmp]) \n\t" \
"adcq %[cy], %%rdx \n\t" \
"adcq $0, %[T0] \n\t" \
"xorq %[cy], %[cy] \n\t" \
"addq %%rdx, %[T1] \n\t" \
"adcq %[T0], %[cy] # cy:T1:tmp[j-1] <-- (X[i] * Y[j] + T1) + M[j] * u + cy * b \n\t" \
"addq " STR(((j+1)*8)) "(%[tmp]), %[T1] \n\t" \
"adcq $0, %[cy] # cy:T1:tmp[j-1] <-- (X[i] * Y[j] + T1) + M[j] * u + (tmp[j+1] + cy) * b \n\t"
#define MONT_FINALIZE(j) \
"movq %[T1], " STR((j*8)) "(%[tmp]) \n\t" \
"movq %[cy], " STR(((j+1)*8)) "(%[tmp]) \n\t"
/*
Comba multiplication and squaring routines are based on the
public-domain tomsfastmath library by Tom St Denis
<http://www.libtom.org/>
<https://github.com/libtom/tomsfastmath/blob/master/src/sqr/fp_sqr_comba.c
<https://github.com/libtom/tomsfastmath/blob/master/src/mul/fp_mul_comba.c>
Compared to the above, we save 5-20% of cycles by using careful register
renaming to implement Comba forward operation.
*/
#define COMBA_3_BY_3_MUL(c0_, c1_, c2_, res_, A_, B_) \
asm volatile ( \
"movq 0(%[A]), %%rax \n\t" \
"mulq 0(%[B]) \n\t" \
"movq %%rax, 0(%[res]) \n\t" \
"movq %%rdx, %[c0] \n\t" \
\
"xorq %[c1], %[c1] \n\t" \
"movq 0(%[A]), %%rax \n\t" \
"mulq 8(%[B]) \n\t" \
"addq %%rax, %[c0] \n\t" \
"adcq %%rdx, %[c1] \n\t" \
\
"xorq %[c2], %[c2] \n\t" \
"movq 8(%[A]), %%rax \n\t" \
"mulq 0(%[B]) \n\t" \
"addq %%rax, %[c0] \n\t" \
"movq %[c0], 8(%[res]) \n\t" \
"adcq %%rdx, %[c1] \n\t" \
"adcq $0, %[c2] \n\t" \
\
"// register renaming (c1, c2, c0)\n\t" \
"xorq %[c0], %[c0] \n\t" \
"movq 0(%[A]), %%rax \n\t" \
"mulq 16(%[B]) \n\t" \
"addq %%rax, %[c1] \n\t" \
"adcq %%rdx, %[c2] \n\t" \
"adcq $0, %[c0] \n\t" \
\
"movq 8(%[A]), %%rax \n\t" \
"mulq 8(%[B]) \n\t" \
"addq %%rax, %[c1] \n\t" \
"adcq %%rdx, %[c2] \n\t" \
"adcq $0, %[c0] \n\t" \
\
"movq 16(%[A]), %%rax \n\t" \
"mulq 0(%[B]) \n\t" \
"addq %%rax, %[c1] \n\t" \
"movq %[c1], 16(%[res]) \n\t" \
"adcq %%rdx, %[c2] \n\t" \
"adcq $0, %[c0] \n\t" \
\
"// register renaming (c2, c0, c1)\n\t" \
"xorq %[c1], %[c1] \n\t" \
"movq 8(%[A]), %%rax \n\t" \
"mulq 16(%[B]) \n\t" \
"addq %%rax, %[c2] \n\t" \
"adcq %%rdx, %[c0] \n\t" \
"adcq $0, %[c1] \n\t" \
\
"movq 16(%[A]), %%rax \n\t" \
"mulq 8(%[B]) \n\t" \
"addq %%rax, %[c2] \n\t" \
"movq %[c2], 24(%[res]) \n\t" \
"adcq %%rdx, %[c0] \n\t" \
"adcq $0, %[c1] \n\t" \
\
"// register renaming (c0, c1, c2)\n\t" \
"xorq %[c2], %[c2] \n\t" \
"movq 16(%[A]), %%rax \n\t" \
"mulq 16(%[B]) \n\t" \
"addq %%rax, %[c0] \n\t" \
"movq %[c0], 32(%[res]) \n\t" \
"adcq %%rdx, %[c1] \n\t" \
"movq %[c1], 40(%[res]) \n\t" \
: [c0] "=&r" (c0_), [c1] "=&r" (c1_), [c2] "=&r" (c2_) \
: [res] "r" (res_), [A] "r" (A_), [B] "r" (B_) \
: "%rax", "%rdx", "cc", "memory")
#define COMBA_3_BY_3_SQR(c0_, c1_, c2_, res_, A_) \
asm volatile ( \
"xorq %[c1], %[c1] \n\t" \
"xorq %[c2], %[c2] \n\t" \
"movq 0(%[A]), %%rax \n\t" \
"mulq %%rax \n\t" \
"movq %%rax, 0(%[res]) \n\t" \
"movq %%rdx, %[c0] \n\t" \
\
"movq 0(%[A]), %%rax \n\t" \
"mulq 8(%[A]) \n\t" \
"addq %%rax, %[c0] \n\t" \
"adcq %%rdx, %[c1] \n\t" \
"addq %%rax, %[c0] \n\t" \
"movq %[c0], 8(%[res]) \n\t" \
"adcq %%rdx, %[c1] \n\t" \
"adcq $0, %[c2] \n\t" \
\
"// register renaming (c1, c2, c0)\n\t" \
"movq 0(%[A]), %%rax \n\t" \
"xorq %[c0], %[c0] \n\t" \
"mulq 16(%[A]) \n\t" \
"addq %%rax, %[c1] \n\t" \
"adcq %%rdx, %[c2] \n\t" \
"adcq $0, %[c0] \n\t" \
"addq %%rax, %[c1] \n\t" \
"adcq %%rdx, %[c2] \n\t" \
"adcq $0, %[c0] \n\t" \
\
"movq 8(%[A]), %%rax \n\t" \
"mulq %%rax \n\t" \
"addq %%rax, %[c1] \n\t" \
"movq %[c1], 16(%[res]) \n\t" \
"adcq %%rdx, %[c2] \n\t" \
"adcq $0, %[c0] \n\t" \
\
"// register renaming (c2, c0, c1)\n\t" \
"movq 8(%[A]), %%rax \n\t" \
"xorq %[c1], %[c1] \n\t" \
"mulq 16(%[A]) \n\t" \
"addq %%rax, %[c2] \n\t" \
"adcq %%rdx, %[c0] \n\t" \
"adcq $0, %[c1] \n\t" \
"addq %%rax, %[c2] \n\t" \
"movq %[c2], 24(%[res]) \n\t" \
"adcq %%rdx, %[c0] \n\t" \
"adcq $0, %[c1] \n\t" \
\
"// register renaming (c0, c1, c2)\n\t" \
"movq 16(%[A]), %%rax \n\t" \
"mulq %%rax \n\t" \
"addq %%rax, %[c0] \n\t" \
"movq %[c0], 32(%[res]) \n\t" \
"adcq %%rdx, %[c1] \n\t" \
"movq %[c1], 40(%[res]) \n\t" \
\
: [c0] "=&r" (c0_), [c1] "=&r" (c1_), [c2] "=&r" (c2_) \
: [res] "r" (res_), [A] "r" (A_) \
: "%rax", "%rdx", "cc", "memory")
/*
The Montgomery reduction here is based on Algorithm 14.32 in
Handbook of Applied Cryptography
<http://cacr.uwaterloo.ca/hac/about/chap14.pdf>.
*/
#define REDUCE_6_LIMB_PRODUCT(k_, tmp1_, tmp2_, tmp3_, inv_, res_, mod_) \
__asm__ volatile \
("///////////////////////////////////\n\t" \
"movq 0(%[res]), %%rax \n\t" \
"mulq %[modprime] \n\t" \
"movq %%rax, %[k] \n\t" \
\
"movq (%[mod]), %%rax \n\t" \
"mulq %[k] \n\t" \
"movq %%rax, %[tmp1] \n\t" \
"movq %%rdx, %[tmp2] \n\t" \
\
"xorq %[tmp3], %[tmp3] \n\t" \
"movq 8(%[mod]), %%rax \n\t" \
"mulq %[k] \n\t" \
"addq %[tmp1], 0(%[res]) \n\t" \
"adcq %%rax, %[tmp2] \n\t" \
"adcq %%rdx, %[tmp3] \n\t" \
\
"xorq %[tmp1], %[tmp1] \n\t" \
"movq 16(%[mod]), %%rax \n\t" \
"mulq %[k] \n\t" \
"addq %[tmp2], 8(%[res]) \n\t" \
"adcq %%rax, %[tmp3] \n\t" \
"adcq %%rdx, %[tmp1] \n\t" \
\
"addq %[tmp3], 16(%[res]) \n\t" \
"adcq %[tmp1], 24(%[res]) \n\t" \
"adcq $0, 32(%[res]) \n\t" \
"adcq $0, 40(%[res]) \n\t" \
\
"///////////////////////////////////\n\t" \
"movq 8(%[res]), %%rax \n\t" \
"mulq %[modprime] \n\t" \
"movq %%rax, %[k] \n\t" \
\
"movq (%[mod]), %%rax \n\t" \
"mulq %[k] \n\t" \
"movq %%rax, %[tmp1] \n\t" \
"movq %%rdx, %[tmp2] \n\t" \
\
"xorq %[tmp3], %[tmp3] \n\t" \
"movq 8(%[mod]), %%rax \n\t" \
"mulq %[k] \n\t" \
"addq %[tmp1], 8(%[res]) \n\t" \
"adcq %%rax, %[tmp2] \n\t" \
"adcq %%rdx, %[tmp3] \n\t" \
\
"xorq %[tmp1], %[tmp1] \n\t" \
"movq 16(%[mod]), %%rax \n\t" \
"mulq %[k] \n\t" \
"addq %[tmp2], 16(%[res]) \n\t" \
"adcq %%rax, %[tmp3] \n\t" \
"adcq %%rdx, %[tmp1] \n\t" \
\
"addq %[tmp3], 24(%[res]) \n\t" \
"adcq %[tmp1], 32(%[res]) \n\t" \
"adcq $0, 40(%[res]) \n\t" \
\
"///////////////////////////////////\n\t" \
"movq 16(%[res]), %%rax \n\t" \
"mulq %[modprime] \n\t" \
"movq %%rax, %[k] \n\t" \
\
"movq (%[mod]), %%rax \n\t" \
"mulq %[k] \n\t" \
"movq %%rax, %[tmp1] \n\t" \
"movq %%rdx, %[tmp2] \n\t" \
\
"xorq %[tmp3], %[tmp3] \n\t" \
"movq 8(%[mod]), %%rax \n\t" \
"mulq %[k] \n\t" \
"addq %[tmp1], 16(%[res]) \n\t" \
"adcq %%rax, %[tmp2] \n\t" \
"adcq %%rdx, %[tmp3] \n\t" \
\
"xorq %[tmp1], %[tmp1] \n\t" \
"movq 16(%[mod]), %%rax \n\t" \
"mulq %[k] \n\t" \
"addq %[tmp2], 24(%[res]) \n\t" \
"adcq %%rax, %[tmp3] \n\t" \
"adcq %%rdx, %[tmp1] \n\t" \
\
"addq %[tmp3], 32(%[res]) \n\t" \
"adcq %[tmp1], 40(%[res]) \n\t" \
: [k] "=&r" (k_), [tmp1] "=&r" (tmp1_), [tmp2] "=&r" (tmp2_), [tmp3] "=&r" (tmp3_) \
: [modprime] "r" (inv_), [res] "r" (res_), [mod] "r" (mod_) \
: "%rax", "%rdx", "cc", "memory")
} // namespace snarklib
#endif