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#include <linux/linkage.h>
51
52.text
53
54
55
56digest = %rdi
57
58msg = %rsi
59
60msglen = %rdx
61T1 = %rcx
62T2 = %r8
63a_64 = %r9
64b_64 = %r10
65c_64 = %r11
66d_64 = %r12
67e_64 = %r13
68f_64 = %r14
69g_64 = %r15
70h_64 = %rbx
71tmp0 = %rax
72
73
74
75
76W_SIZE = 80*8
77
78WK_SIZE = 2*8
79RSPSAVE_SIZE = 1*8
80GPRSAVE_SIZE = 5*8
81
82frame_W = 0
83frame_WK = frame_W + W_SIZE
84frame_RSPSAVE = frame_WK + WK_SIZE
85frame_GPRSAVE = frame_RSPSAVE + RSPSAVE_SIZE
86frame_size = frame_GPRSAVE + GPRSAVE_SIZE
87
88
89
90
91
92
93#define MSG(i) 8*i(msg)
94
95
96#define DIGEST(i) 8*i(digest)
97
98
99#define K_t(i) 8*i+K512(%rip)
100
101
102#define W_t(i) 8*i+frame_W(%rsp)
103
104
105#define WK_2(i) 8*((i%2))+frame_WK(%rsp)
106
107.macro RotateState
108
109 TMP = h_64
110 h_64 = g_64
111 g_64 = f_64
112 f_64 = e_64
113 e_64 = d_64
114 d_64 = c_64
115 c_64 = b_64
116 b_64 = a_64
117 a_64 = TMP
118.endm
119
120.macro RORQ p1 p2
121
122 shld $(64-\p2), \p1, \p1
123.endm
124
125.macro SHA512_Round rnd
126
127 mov f_64, T1
128 mov e_64, tmp0
129 xor g_64, T1
130 RORQ tmp0, 23
131 and e_64, T1
132 xor e_64, tmp0
133 xor g_64, T1
134 idx = \rnd
135 add WK_2(idx), T1
136 RORQ tmp0, 4
137 xor e_64, tmp0
138 mov a_64, T2
139 add h_64, T1
140 RORQ tmp0, 14
141 add tmp0, T1
142 mov a_64, tmp0
143 xor c_64, T2
144 and c_64, tmp0
145 and b_64, T2
146 xor tmp0, T2
147 mov a_64, tmp0
148 RORQ tmp0, 5
149 xor a_64, tmp0
150 add T1, d_64
151 RORQ tmp0, 6
152 xor a_64, tmp0
153 lea (T1, T2), h_64
154 RORQ tmp0, 28
155 add tmp0, h_64
156 RotateState
157.endm
158
159.macro SHA512_2Sched_2Round_avx rnd
160
161
162
163
164
165
166
167
168
169
170
171
172
173
174
175 idx = \rnd - 2
176 vmovdqa W_t(idx), %xmm4
177 idx = \rnd - 15
178 vmovdqu W_t(idx), %xmm5
179 mov f_64, T1
180 vpsrlq $61, %xmm4, %xmm0
181 mov e_64, tmp0
182 vpsrlq $1, %xmm5, %xmm6
183 xor g_64, T1
184 RORQ tmp0, 23
185 vpsrlq $19, %xmm4, %xmm1
186 and e_64, T1
187 xor e_64, tmp0
188 vpxor %xmm1, %xmm0, %xmm0
189 xor g_64, T1
190 idx = \rnd
191 add WK_2(idx), T1
192 vpsrlq $8, %xmm5, %xmm7
193 RORQ tmp0, 4
194 vpsrlq $6, %xmm4, %xmm2
195 xor e_64, tmp0
196 mov a_64, T2
197 add h_64, T1
198 vpxor %xmm7, %xmm6, %xmm6
199 RORQ tmp0, 14
200 add tmp0, T1
201 vpsrlq $7, %xmm5, %xmm8
202 mov a_64, tmp0
203 xor c_64, T2
204 vpsllq $(64-61), %xmm4, %xmm3
205 and c_64, tmp0
206 and b_64, T2
207 vpxor %xmm3, %xmm2, %xmm2
208 xor tmp0, T2
209 mov a_64, tmp0
210 vpsllq $(64-1), %xmm5, %xmm9
211 RORQ tmp0, 5
212 vpxor %xmm9, %xmm8, %xmm8
213 xor a_64, tmp0
214 add T1, d_64
215 RORQ tmp0, 6
216 xor a_64, tmp0
217 vpxor %xmm8, %xmm6, %xmm6
218
219 lea (T1, T2), h_64
220 RORQ tmp0, 28
221 vpsllq $(64-19), %xmm4, %xmm4
222 add tmp0, h_64
223 RotateState
224 vpxor %xmm4, %xmm0, %xmm0
225
226 mov f_64, T1
227 vpxor %xmm2, %xmm0, %xmm0
228 mov e_64, tmp0
229 xor g_64, T1
230 idx = \rnd - 16
231 vpaddq W_t(idx), %xmm0, %xmm0
232 idx = \rnd - 7
233 vmovdqu W_t(idx), %xmm1
234 RORQ tmp0, 23
235 and e_64, T1
236 xor e_64, tmp0
237 xor g_64, T1
238 vpsllq $(64-8), %xmm5, %xmm5
239 idx = \rnd + 1
240 add WK_2(idx), T1
241 vpxor %xmm5, %xmm6, %xmm6
242 RORQ tmp0, 4
243 vpaddq %xmm6, %xmm0, %xmm0
244 xor e_64, tmp0
245 vpaddq %xmm1, %xmm0, %xmm0
246
247 mov a_64, T2
248 add h_64, T1
249 RORQ tmp0, 14
250 add tmp0, T1
251 idx = \rnd
252 vmovdqa %xmm0, W_t(idx)
253 vpaddq K_t(idx), %xmm0, %xmm0
254 vmovdqa %xmm0, WK_2(idx)
255 mov a_64, tmp0
256 xor c_64, T2
257 and c_64, tmp0
258 and b_64, T2
259 xor tmp0, T2
260 mov a_64, tmp0
261 RORQ tmp0, 5
262 xor a_64, tmp0
263 add T1, d_64
264 RORQ tmp0, 6
265 xor a_64, tmp0
266 lea (T1, T2), h_64
267 RORQ tmp0, 28
268 add tmp0, h_64
269 RotateState
270.endm
271
272
273
274
275
276
277
278
279
280SYM_FUNC_START(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
368SYM_FUNC_END(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