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#ifdef CONFIG_AS_AVX
51#include <linux/linkage.h>
52
53.text
54
55
56
57digest = %rdi
58
59msg = %rsi
60
61msglen = %rdx
62T1 = %rcx
63T2 = %r8
64a_64 = %r9
65b_64 = %r10
66c_64 = %r11
67d_64 = %r12
68e_64 = %r13
69f_64 = %r14
70g_64 = %r15
71h_64 = %rbx
72tmp0 = %rax
73
74
75
76
77W_SIZE = 80*8
78
79WK_SIZE = 2*8
80RSPSAVE_SIZE = 1*8
81GPRSAVE_SIZE = 5*8
82
83frame_W = 0
84frame_WK = frame_W + W_SIZE
85frame_RSPSAVE = frame_WK + WK_SIZE
86frame_GPRSAVE = frame_RSPSAVE + RSPSAVE_SIZE
87frame_size = frame_GPRSAVE + GPRSAVE_SIZE
88
89
90
91
92
93
94#define MSG(i) 8*i(msg)
95
96
97#define DIGEST(i) 8*i(digest)
98
99
100#define K_t(i) 8*i+K512(%rip)
101
102
103#define W_t(i) 8*i+frame_W(%rsp)
104
105
106#define WK_2(i) 8*((i%2))+frame_WK(%rsp)
107
108.macro RotateState
109
110 TMP = h_64
111 h_64 = g_64
112 g_64 = f_64
113 f_64 = e_64
114 e_64 = d_64
115 d_64 = c_64
116 c_64 = b_64
117 b_64 = a_64
118 a_64 = TMP
119.endm
120
121.macro RORQ p1 p2
122
123 shld $(64-\p2), \p1, \p1
124.endm
125
126.macro SHA512_Round rnd
127
128 mov f_64, T1
129 mov e_64, tmp0
130 xor g_64, T1
131 RORQ tmp0, 23
132 and e_64, T1
133 xor e_64, tmp0
134 xor g_64, T1
135 idx = \rnd
136 add WK_2(idx), T1
137 RORQ tmp0, 4
138 xor e_64, tmp0
139 mov a_64, T2
140 add h_64, T1
141 RORQ tmp0, 14
142 add tmp0, T1
143 mov a_64, tmp0
144 xor c_64, T2
145 and c_64, tmp0
146 and b_64, T2
147 xor tmp0, T2
148 mov a_64, tmp0
149 RORQ tmp0, 5
150 xor a_64, tmp0
151 add T1, d_64
152 RORQ tmp0, 6
153 xor a_64, tmp0
154 lea (T1, T2), h_64
155 RORQ tmp0, 28
156 add tmp0, h_64
157 RotateState
158.endm
159
160.macro SHA512_2Sched_2Round_avx rnd
161
162
163
164
165
166
167
168
169
170
171
172
173
174
175
176 idx = \rnd - 2
177 vmovdqa W_t(idx), %xmm4
178 idx = \rnd - 15
179 vmovdqu W_t(idx), %xmm5
180 mov f_64, T1
181 vpsrlq $61, %xmm4, %xmm0
182 mov e_64, tmp0
183 vpsrlq $1, %xmm5, %xmm6
184 xor g_64, T1
185 RORQ tmp0, 23
186 vpsrlq $19, %xmm4, %xmm1
187 and e_64, T1
188 xor e_64, tmp0
189 vpxor %xmm1, %xmm0, %xmm0
190 xor g_64, T1
191 idx = \rnd
192 add WK_2(idx), T1
193 vpsrlq $8, %xmm5, %xmm7
194 RORQ tmp0, 4
195 vpsrlq $6, %xmm4, %xmm2
196 xor e_64, tmp0
197 mov a_64, T2
198 add h_64, T1
199 vpxor %xmm7, %xmm6, %xmm6
200 RORQ tmp0, 14
201 add tmp0, T1
202 vpsrlq $7, %xmm5, %xmm8
203 mov a_64, tmp0
204 xor c_64, T2
205 vpsllq $(64-61), %xmm4, %xmm3
206 and c_64, tmp0
207 and b_64, T2
208 vpxor %xmm3, %xmm2, %xmm2
209 xor tmp0, T2
210 mov a_64, tmp0
211 vpsllq $(64-1), %xmm5, %xmm9
212 RORQ tmp0, 5
213 vpxor %xmm9, %xmm8, %xmm8
214 xor a_64, tmp0
215 add T1, d_64
216 RORQ tmp0, 6
217 xor a_64, tmp0
218 vpxor %xmm8, %xmm6, %xmm6
219
220 lea (T1, T2), h_64
221 RORQ tmp0, 28
222 vpsllq $(64-19), %xmm4, %xmm4
223 add tmp0, h_64
224 RotateState
225 vpxor %xmm4, %xmm0, %xmm0
226
227 mov f_64, T1
228 vpxor %xmm2, %xmm0, %xmm0
229 mov e_64, tmp0
230 xor g_64, T1
231 idx = \rnd - 16
232 vpaddq W_t(idx), %xmm0, %xmm0
233 idx = \rnd - 7
234 vmovdqu W_t(idx), %xmm1
235 RORQ tmp0, 23
236 and e_64, T1
237 xor e_64, tmp0
238 xor g_64, T1
239 vpsllq $(64-8), %xmm5, %xmm5
240 idx = \rnd + 1
241 add WK_2(idx), T1
242 vpxor %xmm5, %xmm6, %xmm6
243 RORQ tmp0, 4
244 vpaddq %xmm6, %xmm0, %xmm0
245 xor e_64, tmp0
246 vpaddq %xmm1, %xmm0, %xmm0
247
248 mov a_64, T2
249 add h_64, T1
250 RORQ tmp0, 14
251 add tmp0, T1
252 idx = \rnd
253 vmovdqa %xmm0, W_t(idx)
254 vpaddq K_t(idx), %xmm0, %xmm0
255 vmovdqa %xmm0, WK_2(idx)
256 mov a_64, tmp0
257 xor c_64, T2
258 and c_64, tmp0
259 and b_64, T2
260 xor tmp0, T2
261 mov a_64, tmp0
262 RORQ tmp0, 5
263 xor a_64, tmp0
264 add T1, d_64
265 RORQ tmp0, 6
266 xor a_64, tmp0
267 lea (T1, T2), h_64
268 RORQ tmp0, 28
269 add tmp0, h_64
270 RotateState
271.endm
272
273
274
275
276
277
278
279
280ENTRY(sha512_transform_avx)
281 cmp $0, msglen
282 je nowork
283
284
285 mov %rsp, %rax
286 sub $frame_size, %rsp
287 and $~(0x20 - 1), %rsp
288 mov %rax, frame_RSPSAVE(%rsp)
289
290
291 mov %rbx, frame_GPRSAVE(%rsp)
292 mov %r12, frame_GPRSAVE +8*1(%rsp)
293 mov %r13, frame_GPRSAVE +8*2(%rsp)
294 mov %r14, frame_GPRSAVE +8*3(%rsp)
295 mov %r15, frame_GPRSAVE +8*4(%rsp)
296
297updateblock:
298
299
300 mov DIGEST(0), a_64
301 mov DIGEST(1), b_64
302 mov DIGEST(2), c_64
303 mov DIGEST(3), d_64
304 mov DIGEST(4), e_64
305 mov DIGEST(5), f_64
306 mov DIGEST(6), g_64
307 mov DIGEST(7), h_64
308
309 t = 0
310 .rept 80/2 + 1
311
312
313 .if t < 2
314
315 vmovdqa XMM_QWORD_BSWAP(%rip), %xmm1
316 vmovdqu MSG(t), %xmm0
317 vpshufb %xmm1, %xmm0, %xmm0
318 vmovdqa %xmm0, W_t(t)
319 vpaddq K_t(t), %xmm0, %xmm0
320 vmovdqa %xmm0, WK_2(t)
321 .elseif t < 16
322
323 vmovdqu MSG(t), %xmm0
324 vpshufb %xmm1, %xmm0, %xmm0
325 SHA512_Round t-2
326 vmovdqa %xmm0, W_t(t)
327 vpaddq K_t(t), %xmm0, %xmm0
328 SHA512_Round t-1
329 vmovdqa %xmm0, WK_2(t)
330 .elseif t < 79
331
332 SHA512_2Sched_2Round_avx t
333 .else
334
335 SHA512_Round t-2
336 SHA512_Round t-1
337 .endif
338 t = t+2
339 .endr
340
341
342 add a_64, DIGEST(0)
343 add b_64, DIGEST(1)
344 add c_64, DIGEST(2)
345 add d_64, DIGEST(3)
346 add e_64, DIGEST(4)
347 add f_64, DIGEST(5)
348 add g_64, DIGEST(6)
349 add h_64, DIGEST(7)
350
351
352 add $16*8, msg
353 dec msglen
354 jnz updateblock
355
356
357 mov frame_GPRSAVE(%rsp), %rbx
358 mov frame_GPRSAVE +8*1(%rsp), %r12
359 mov frame_GPRSAVE +8*2(%rsp), %r13
360 mov frame_GPRSAVE +8*3(%rsp), %r14
361 mov frame_GPRSAVE +8*4(%rsp), %r15
362
363
364 mov frame_RSPSAVE(%rsp), %rsp
365
366nowork:
367 ret
368ENDPROC(sha512_transform_avx)
369
370
371
372
373.section .rodata.cst16.XMM_QWORD_BSWAP, "aM", @progbits, 16
374.align 16
375
376XMM_QWORD_BSWAP:
377 .octa 0x08090a0b0c0d0e0f0001020304050607
378
379
380
381
382.section .rodata.cst640.K512, "aM", @progbits, 640
383.align 64
384
385K512:
386 .quad 0x428a2f98d728ae22,0x7137449123ef65cd
387 .quad 0xb5c0fbcfec4d3b2f,0xe9b5dba58189dbbc
388 .quad 0x3956c25bf348b538,0x59f111f1b605d019
389 .quad 0x923f82a4af194f9b,0xab1c5ed5da6d8118
390 .quad 0xd807aa98a3030242,0x12835b0145706fbe
391 .quad 0x243185be4ee4b28c,0x550c7dc3d5ffb4e2
392 .quad 0x72be5d74f27b896f,0x80deb1fe3b1696b1
393 .quad 0x9bdc06a725c71235,0xc19bf174cf692694
394 .quad 0xe49b69c19ef14ad2,0xefbe4786384f25e3
395 .quad 0x0fc19dc68b8cd5b5,0x240ca1cc77ac9c65
396 .quad 0x2de92c6f592b0275,0x4a7484aa6ea6e483
397 .quad 0x5cb0a9dcbd41fbd4,0x76f988da831153b5
398 .quad 0x983e5152ee66dfab,0xa831c66d2db43210
399 .quad 0xb00327c898fb213f,0xbf597fc7beef0ee4
400 .quad 0xc6e00bf33da88fc2,0xd5a79147930aa725
401 .quad 0x06ca6351e003826f,0x142929670a0e6e70
402 .quad 0x27b70a8546d22ffc,0x2e1b21385c26c926
403 .quad 0x4d2c6dfc5ac42aed,0x53380d139d95b3df
404 .quad 0x650a73548baf63de,0x766a0abb3c77b2a8
405 .quad 0x81c2c92e47edaee6,0x92722c851482353b
406 .quad 0xa2bfe8a14cf10364,0xa81a664bbc423001
407 .quad 0xc24b8b70d0f89791,0xc76c51a30654be30
408 .quad 0xd192e819d6ef5218,0xd69906245565a910
409 .quad 0xf40e35855771202a,0x106aa07032bbd1b8
410 .quad 0x19a4c116b8d2d0c8,0x1e376c085141ab53
411 .quad 0x2748774cdf8eeb99,0x34b0bcb5e19b48a8
412 .quad 0x391c0cb3c5c95a63,0x4ed8aa4ae3418acb
413 .quad 0x5b9cca4f7763e373,0x682e6ff3d6b2b8a3
414 .quad 0x748f82ee5defb2fc,0x78a5636f43172f60
415 .quad 0x84c87814a1f0ab72,0x8cc702081a6439ec
416 .quad 0x90befffa23631e28,0xa4506cebde82bde9
417 .quad 0xbef9a3f7b2c67915,0xc67178f2e372532b
418 .quad 0xca273eceea26619c,0xd186b8c721c0c207
419 .quad 0xeada7dd6cde0eb1e,0xf57d4f7fee6ed178
420 .quad 0x06f067aa72176fba,0x0a637dc5a2c898a6
421 .quad 0x113f9804bef90dae,0x1b710b35131c471b
422 .quad 0x28db77f523047d84,0x32caab7b40c72493
423 .quad 0x3c9ebe0a15c9bebc,0x431d67c49c100d4c
424 .quad 0x4cc5d4becb3e42b6,0x597f299cfc657e2a
425 .quad 0x5fcb6fab3ad6faec,0x6c44198c4a475817
426#endif
427