forked from ocminer/node-multi-hashing-xzc
-
Notifications
You must be signed in to change notification settings - Fork 1
/
Copy pathscrypt-jane-mix_salsa-avx.h
381 lines (357 loc) · 9.94 KB
/
scrypt-jane-mix_salsa-avx.h
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
/* x86 */
#if defined(X86ASM_AVX) && (!defined(SCRYPT_CHOOSE_COMPILETIME) || !defined(SCRYPT_SALSA_INCLUDED))
#define SCRYPT_SALSA_AVX
asm_naked_fn_proto(void, scrypt_ChunkMix_avx)(uint32_t *Bout/*[chunkBytes]*/, uint32_t *Bin/*[chunkBytes]*/, uint32_t *Bxor/*[chunkBytes]*/, uint32_t r)
asm_naked_fn(scrypt_ChunkMix_avx)
a1(push ebx)
a1(push edi)
a1(push esi)
a1(push ebp)
a2(mov ebp,esp)
a2(mov edi,[ebp+20])
a2(mov esi,[ebp+24])
a2(mov eax,[ebp+28])
a2(mov ebx,[ebp+32])
a2(sub esp,32)
a2(and esp,~63)
a2(lea edx,[ebx*2])
a2(shl edx,6)
a2(lea ecx,[edx-64])
a2(and eax, eax)
a2(movdqa xmm0,[ecx+esi+0])
a2(movdqa xmm1,[ecx+esi+16])
a2(movdqa xmm2,[ecx+esi+32])
a2(movdqa xmm3,[ecx+esi+48])
a1(jz scrypt_ChunkMix_avx_no_xor1)
a3(vpxor xmm0,xmm0,[ecx+eax+0])
a3(vpxor xmm1,xmm1,[ecx+eax+16])
a3(vpxor xmm2,xmm2,[ecx+eax+32])
a3(vpxor xmm3,xmm3,[ecx+eax+48])
a1(scrypt_ChunkMix_avx_no_xor1:)
a2(xor ecx,ecx)
a2(xor ebx,ebx)
a1(scrypt_ChunkMix_avx_loop:)
a2(and eax, eax)
a3(vpxor xmm0,xmm0,[esi+ecx+0])
a3(vpxor xmm1,xmm1,[esi+ecx+16])
a3(vpxor xmm2,xmm2,[esi+ecx+32])
a3(vpxor xmm3,xmm3,[esi+ecx+48])
a1(jz scrypt_ChunkMix_avx_no_xor2)
a3(vpxor xmm0,xmm0,[eax+ecx+0])
a3(vpxor xmm1,xmm1,[eax+ecx+16])
a3(vpxor xmm2,xmm2,[eax+ecx+32])
a3(vpxor xmm3,xmm3,[eax+ecx+48])
a1(scrypt_ChunkMix_avx_no_xor2:)
a2(vmovdqa [esp+0],xmm0)
a2(vmovdqa [esp+16],xmm1)
a2(vmovdqa xmm6,xmm2)
a2(vmovdqa xmm7,xmm3)
a2(mov eax,8)
a1(scrypt_salsa_avx_loop: )
a3(vpaddd xmm4, xmm1, xmm0)
a3(vpsrld xmm5, xmm4, 25)
a3(vpslld xmm4, xmm4, 7)
a3(vpxor xmm3, xmm3, xmm5)
a3(vpxor xmm3, xmm3, xmm4)
a3(vpaddd xmm4, xmm0, xmm3)
a3(vpsrld xmm5, xmm4, 23)
a3(vpslld xmm4, xmm4, 9)
a3(vpxor xmm2, xmm2, xmm5)
a3(vpxor xmm2, xmm2, xmm4)
a3(vpaddd xmm4, xmm3, xmm2)
a3(vpsrld xmm5, xmm4, 19)
a3(vpslld xmm4, xmm4, 13)
a3(vpxor xmm1, xmm1, xmm5)
a3(pshufd xmm3, xmm3, 0x93)
a3(vpxor xmm1, xmm1, xmm4)
a3(vpaddd xmm4, xmm2, xmm1)
a3(vpsrld xmm5, xmm4, 14)
a3(vpslld xmm4, xmm4, 18)
a3(vpxor xmm0, xmm0, xmm5)
a3(pshufd xmm2, xmm2, 0x4e)
a3(vpxor xmm0, xmm0, xmm4)
a2(sub eax, 2)
a3(vpaddd xmm4, xmm3, xmm0)
a3(pshufd xmm1, xmm1, 0x39)
a3(vpsrld xmm5, xmm4, 25)
a3(vpslld xmm4, xmm4, 7)
a3(vpxor xmm1, xmm1, xmm5)
a3(vpxor xmm1, xmm1, xmm4)
a3(vpaddd xmm4, xmm0, xmm1)
a3(vpsrld xmm5, xmm4, 23)
a3(vpslld xmm4, xmm4, 9)
a3(vpxor xmm2, xmm2, xmm5)
a3(vpxor xmm2, xmm2, xmm4)
a3(vpaddd xmm4, xmm1, xmm2)
a3(vpsrld xmm5, xmm4, 19)
a3(vpslld xmm4, xmm4, 13)
a3(vpxor xmm3, xmm3, xmm5)
a3(pshufd xmm1, xmm1, 0x93)
a3(vpxor xmm3, xmm3, xmm4)
a3(vpaddd xmm4, xmm2, xmm3)
a3(vpsrld xmm5, xmm4, 14)
a3(vpslld xmm4, xmm4, 18)
a3(vpxor xmm0, xmm0, xmm5)
a3(pshufd xmm2, xmm2, 0x4e)
a3(vpxor xmm0, xmm0, xmm4)
a3(pshufd xmm3, xmm3, 0x39)
a1(ja scrypt_salsa_avx_loop)
a3(vpaddd xmm0,xmm0,[esp+0])
a3(vpaddd xmm1,xmm1,[esp+16])
a3(vpaddd xmm2,xmm2,xmm6)
a3(vpaddd xmm3,xmm3,xmm7)
a2(lea eax,[ebx+ecx])
a2(xor ebx,edx)
a2(and eax,~0x7f)
a2(add ecx,64)
a2(shr eax,1)
a2(add eax, edi)
a2(cmp ecx,edx)
a2(vmovdqa [eax+0],xmm0)
a2(vmovdqa [eax+16],xmm1)
a2(vmovdqa [eax+32],xmm2)
a2(vmovdqa [eax+48],xmm3)
a2(mov eax,[ebp+28])
a1(jne scrypt_ChunkMix_avx_loop)
a2(mov esp,ebp)
a1(pop ebp)
a1(pop esi)
a1(pop edi)
a1(pop ebx)
a1(ret 16)
asm_naked_fn_end(scrypt_ChunkMix_avx)
#endif
/* x64 */
#if defined(X86_64ASM_AVX) && (!defined(SCRYPT_CHOOSE_COMPILETIME) || !defined(SCRYPT_SALSA_INCLUDED))
#define SCRYPT_SALSA_AVX
asm_naked_fn_proto(void, scrypt_ChunkMix_avx)(uint32_t *Bout/*[chunkBytes]*/, uint32_t *Bin/*[chunkBytes]*/, uint32_t *Bxor/*[chunkBytes]*/, uint32_t r)
asm_naked_fn(scrypt_ChunkMix_avx)
a2(lea rcx,[rcx*2])
a2(shl rcx,6)
a2(lea r9,[rcx-64])
a2(lea rax,[rsi+r9])
a2(lea r9,[rdx+r9])
a2(and rdx, rdx)
a2(vmovdqa xmm0,[rax+0])
a2(vmovdqa xmm1,[rax+16])
a2(vmovdqa xmm2,[rax+32])
a2(vmovdqa xmm3,[rax+48])
a1(jz scrypt_ChunkMix_avx_no_xor1)
a3(vpxor xmm0,xmm0,[r9+0])
a3(vpxor xmm1,xmm1,[r9+16])
a3(vpxor xmm2,xmm2,[r9+32])
a3(vpxor xmm3,xmm3,[r9+48])
a1(scrypt_ChunkMix_avx_no_xor1:)
a2(xor r9,r9)
a2(xor r8,r8)
a1(scrypt_ChunkMix_avx_loop:)
a2(and rdx, rdx)
a3(vpxor xmm0,xmm0,[rsi+r9+0])
a3(vpxor xmm1,xmm1,[rsi+r9+16])
a3(vpxor xmm2,xmm2,[rsi+r9+32])
a3(vpxor xmm3,xmm3,[rsi+r9+48])
a1(jz scrypt_ChunkMix_avx_no_xor2)
a3(vpxor xmm0,xmm0,[rdx+r9+0])
a3(vpxor xmm1,xmm1,[rdx+r9+16])
a3(vpxor xmm2,xmm2,[rdx+r9+32])
a3(vpxor xmm3,xmm3,[rdx+r9+48])
a1(scrypt_ChunkMix_avx_no_xor2:)
a2(vmovdqa xmm8,xmm0)
a2(vmovdqa xmm9,xmm1)
a2(vmovdqa xmm10,xmm2)
a2(vmovdqa xmm11,xmm3)
a2(mov rax,8)
a1(scrypt_salsa_avx_loop: )
a3(vpaddd xmm4, xmm1, xmm0)
a3(vpsrld xmm5, xmm4, 25)
a3(vpslld xmm4, xmm4, 7)
a3(vpxor xmm3, xmm3, xmm5)
a3(vpxor xmm3, xmm3, xmm4)
a3(vpaddd xmm4, xmm0, xmm3)
a3(vpsrld xmm5, xmm4, 23)
a3(vpslld xmm4, xmm4, 9)
a3(vpxor xmm2, xmm2, xmm5)
a3(vpxor xmm2, xmm2, xmm4)
a3(vpaddd xmm4, xmm3, xmm2)
a3(vpsrld xmm5, xmm4, 19)
a3(vpslld xmm4, xmm4, 13)
a3(vpxor xmm1, xmm1, xmm5)
a3(pshufd xmm3, xmm3, 0x93)
a3(vpxor xmm1, xmm1, xmm4)
a3(vpaddd xmm4, xmm2, xmm1)
a3(vpsrld xmm5, xmm4, 14)
a3(vpslld xmm4, xmm4, 18)
a3(vpxor xmm0, xmm0, xmm5)
a3(pshufd xmm2, xmm2, 0x4e)
a3(vpxor xmm0, xmm0, xmm4)
a2(sub rax, 2)
a3(vpaddd xmm4, xmm3, xmm0)
a3(pshufd xmm1, xmm1, 0x39)
a3(vpsrld xmm5, xmm4, 25)
a3(vpslld xmm4, xmm4, 7)
a3(vpxor xmm1, xmm1, xmm5)
a3(vpxor xmm1, xmm1, xmm4)
a3(vpaddd xmm4, xmm0, xmm1)
a3(vpsrld xmm5, xmm4, 23)
a3(vpslld xmm4, xmm4, 9)
a3(vpxor xmm2, xmm2, xmm5)
a3(vpxor xmm2, xmm2, xmm4)
a3(vpaddd xmm4, xmm1, xmm2)
a3(vpsrld xmm5, xmm4, 19)
a3(vpslld xmm4, xmm4, 13)
a3(vpxor xmm3, xmm3, xmm5)
a3(pshufd xmm1, xmm1, 0x93)
a3(vpxor xmm3, xmm3, xmm4)
a3(vpaddd xmm4, xmm2, xmm3)
a3(vpsrld xmm5, xmm4, 14)
a3(vpslld xmm4, xmm4, 18)
a3(vpxor xmm0, xmm0, xmm5)
a3(pshufd xmm2, xmm2, 0x4e)
a3(vpxor xmm0, xmm0, xmm4)
a3(pshufd xmm3, xmm3, 0x39)
a1(ja scrypt_salsa_avx_loop)
a3(vpaddd xmm0,xmm0,xmm8)
a3(vpaddd xmm1,xmm1,xmm9)
a3(vpaddd xmm2,xmm2,xmm10)
a3(vpaddd xmm3,xmm3,xmm11)
a2(lea rax,[r8+r9])
a2(xor r8,rcx)
a2(and rax,~0x7f)
a2(add r9,64)
a2(shr rax,1)
a2(add rax, rdi)
a2(cmp r9,rcx)
a2(vmovdqa [rax+0],xmm0)
a2(vmovdqa [rax+16],xmm1)
a2(vmovdqa [rax+32],xmm2)
a2(vmovdqa [rax+48],xmm3)
a1(jne scrypt_ChunkMix_avx_loop)
a1(ret)
asm_naked_fn_end(scrypt_ChunkMix_avx)
#endif
/* intrinsic */
#if defined(X86_INTRINSIC_AVX) && (!defined(SCRYPT_CHOOSE_COMPILETIME) || !defined(SCRYPT_SALSA_INCLUDED))
#define SCRYPT_SALSA_AVX
static void NOINLINE
scrypt_ChunkMix_avx(uint32_t *Bout/*[chunkBytes]*/, uint32_t *Bin/*[chunkBytes]*/, uint32_t *Bxor/*[chunkBytes]*/, uint32_t r) {
uint32_t i, blocksPerChunk = r * 2, half = 0;
xmmi *xmmp,x0,x1,x2,x3,x4,x5,t0,t1,t2,t3;
size_t rounds;
/* 1: X = B_{2r - 1} */
xmmp = (xmmi *)scrypt_block(Bin, blocksPerChunk - 1);
x0 = xmmp[0];
x1 = xmmp[1];
x2 = xmmp[2];
x3 = xmmp[3];
if (Bxor) {
xmmp = (xmmi *)scrypt_block(Bxor, blocksPerChunk - 1);
x0 = _mm_xor_si128(x0, xmmp[0]);
x1 = _mm_xor_si128(x1, xmmp[1]);
x2 = _mm_xor_si128(x2, xmmp[2]);
x3 = _mm_xor_si128(x3, xmmp[3]);
}
/* 2: for i = 0 to 2r - 1 do */
for (i = 0; i < blocksPerChunk; i++, half ^= r) {
/* 3: X = H(X ^ B_i) */
xmmp = (xmmi *)scrypt_block(Bin, i);
x0 = _mm_xor_si128(x0, xmmp[0]);
x1 = _mm_xor_si128(x1, xmmp[1]);
x2 = _mm_xor_si128(x2, xmmp[2]);
x3 = _mm_xor_si128(x3, xmmp[3]);
if (Bxor) {
xmmp = (xmmi *)scrypt_block(Bxor, i);
x0 = _mm_xor_si128(x0, xmmp[0]);
x1 = _mm_xor_si128(x1, xmmp[1]);
x2 = _mm_xor_si128(x2, xmmp[2]);
x3 = _mm_xor_si128(x3, xmmp[3]);
}
t0 = x0;
t1 = x1;
t2 = x2;
t3 = x3;
for (rounds = 8; rounds; rounds -= 2) {
x4 = x1;
x4 = _mm_add_epi32(x4, x0);
x5 = x4;
x4 = _mm_slli_epi32(x4, 7);
x5 = _mm_srli_epi32(x5, 25);
x3 = _mm_xor_si128(x3, x4);
x4 = x0;
x3 = _mm_xor_si128(x3, x5);
x4 = _mm_add_epi32(x4, x3);
x5 = x4;
x4 = _mm_slli_epi32(x4, 9);
x5 = _mm_srli_epi32(x5, 23);
x2 = _mm_xor_si128(x2, x4);
x4 = x3;
x2 = _mm_xor_si128(x2, x5);
x3 = _mm_shuffle_epi32(x3, 0x93);
x4 = _mm_add_epi32(x4, x2);
x5 = x4;
x4 = _mm_slli_epi32(x4, 13);
x5 = _mm_srli_epi32(x5, 19);
x1 = _mm_xor_si128(x1, x4);
x4 = x2;
x1 = _mm_xor_si128(x1, x5);
x2 = _mm_shuffle_epi32(x2, 0x4e);
x4 = _mm_add_epi32(x4, x1);
x5 = x4;
x4 = _mm_slli_epi32(x4, 18);
x5 = _mm_srli_epi32(x5, 14);
x0 = _mm_xor_si128(x0, x4);
x4 = x3;
x0 = _mm_xor_si128(x0, x5);
x1 = _mm_shuffle_epi32(x1, 0x39);
x4 = _mm_add_epi32(x4, x0);
x5 = x4;
x4 = _mm_slli_epi32(x4, 7);
x5 = _mm_srli_epi32(x5, 25);
x1 = _mm_xor_si128(x1, x4);
x4 = x0;
x1 = _mm_xor_si128(x1, x5);
x4 = _mm_add_epi32(x4, x1);
x5 = x4;
x4 = _mm_slli_epi32(x4, 9);
x5 = _mm_srli_epi32(x5, 23);
x2 = _mm_xor_si128(x2, x4);
x4 = x1;
x2 = _mm_xor_si128(x2, x5);
x1 = _mm_shuffle_epi32(x1, 0x93);
x4 = _mm_add_epi32(x4, x2);
x5 = x4;
x4 = _mm_slli_epi32(x4, 13);
x5 = _mm_srli_epi32(x5, 19);
x3 = _mm_xor_si128(x3, x4);
x4 = x2;
x3 = _mm_xor_si128(x3, x5);
x2 = _mm_shuffle_epi32(x2, 0x4e);
x4 = _mm_add_epi32(x4, x3);
x5 = x4;
x4 = _mm_slli_epi32(x4, 18);
x5 = _mm_srli_epi32(x5, 14);
x0 = _mm_xor_si128(x0, x4);
x3 = _mm_shuffle_epi32(x3, 0x39);
x0 = _mm_xor_si128(x0, x5);
}
x0 = _mm_add_epi32(x0, t0);
x1 = _mm_add_epi32(x1, t1);
x2 = _mm_add_epi32(x2, t2);
x3 = _mm_add_epi32(x3, t3);
/* 4: Y_i = X */
/* 6: B'[0..r-1] = Y_even */
/* 6: B'[r..2r-1] = Y_odd */
xmmp = (xmmi *)scrypt_block(Bout, (i / 2) + half);
xmmp[0] = x0;
xmmp[1] = x1;
xmmp[2] = x2;
xmmp[3] = x3;
}
}
#endif
#if defined(SCRYPT_SALSA_AVX)
/* uses salsa_core_tangle_sse2 */
#undef SCRYPT_MIX
#define SCRYPT_MIX "Salsa/8-AVX"
#undef SCRYPT_SALSA_INCLUDED
#define SCRYPT_SALSA_INCLUDED
#endif