1 ;;;;;;;;;;;;;;;;;;;;;;;;;;;;;;;;;;;;;;;;;;;;;;;;;;;;;;;;;;;;;;;;;;;;;;;;
2 ; Copyright(c) 2011-2016 Intel Corporation All rights reserved.
4 ; Redistribution and use in source and binary forms, with or without
5 ; modification, are permitted provided that the following conditions
7 ; * Redistributions of source code must retain the above copyright
8 ; notice, this list of conditions and the following disclaimer.
9 ; * Redistributions in binary form must reproduce the above copyright
10 ; notice, this list of conditions and the following disclaimer in
11 ; the documentation and/or other materials provided with the
13 ; * Neither the name of Intel Corporation nor the names of its
14 ; contributors may be used to endorse or promote products derived
15 ; from this software without specific prior written permission.
17 ; THIS SOFTWARE IS PROVIDED BY THE COPYRIGHT HOLDERS AND CONTRIBUTORS
18 ; "AS IS" AND ANY EXPRESS OR IMPLIED WARRANTIES, INCLUDING, BUT NOT
19 ; LIMITED TO, THE IMPLIED WARRANTIES OF MERCHANTABILITY AND FITNESS FOR
20 ; A PARTICULAR PURPOSE ARE DISCLAIMED. IN NO EVENT SHALL THE COPYRIGHT
21 ; OWNER OR CONTRIBUTORS BE LIABLE FOR ANY DIRECT, INDIRECT, INCIDENTAL,
22 ; SPECIAL, EXEMPLARY, OR CONSEQUENTIAL DAMAGES (INCLUDING, BUT NOT
23 ; LIMITED TO, PROCUREMENT OF SUBSTITUTE GOODS OR SERVICES; LOSS OF USE,
24 ; DATA, OR PROFITS; OR BUSINESS INTERRUPTION) HOWEVER CAUSED AND ON ANY
25 ; THEORY OF LIABILITY, WHETHER IN CONTRACT, STRICT LIABILITY, OR TORT
26 ; (INCLUDING NEGLIGENCE OR OTHERWISE) ARISING IN ANY WAY OUT OF THE USE
27 ; OF THIS SOFTWARE, EVEN IF ADVISED OF THE POSSIBILITY OF SUCH DAMAGE.
28 ;;;;;;;;;;;;;;;;;;;;;;;;;;;;;;;;;;;;;;;;;;;;;;;;;;;;;;;;;;;;;;;;;;;;;;;;
30 %include "reg_sizes.asm"
37 %ifidn __OUTPUT_FORMAT__, win64
38 %define msg rcx ; ARG1
39 %define digest rdx ; ARG2
40 %define msglen r8 ; ARG3
44 %define msg rdi ; ARG1
45 %define digest rsi ; ARG2
46 %define msglen rdx ; ARG3
60 ; Local variables (stack frame)
61 ; Note: frame_size must be an odd multiple of 8 bytes to XMM align RSP
63 .W: resq 80 ; Message Schedule
64 .WK: resq 2 ; W[t] + K[t] | W[t+1] + K[t+1]
66 %ifidn __OUTPUT_FORMAT__, win64
73 ; Useful QWORD "arrays" for simpler memory references
74 %define MSG(i) msg + 8*(i) ; Input message (arg1)
75 %define DIGEST(i) digest + 8*(i) ; Output Digest (arg2)
76 %define K_t(i) K512 + 8*(i) ; SHA Constants (static mem)
77 %define W_t(i) rsp + frame.W + 8*(i) ; Message Schedule (stack frame)
78 %define WK_2(i) rsp + frame.WK + 8*((i) % 2) ; W[t]+K[t] (stack frame)
79 ; MSG, DIGEST, K_t, W_t are arrays
80 ; WK_2(t) points to 1 of 2 qwords at frame.WK depdending on t being odd/even
83 ; Rotate symbles a..h right
100 mov tmp0, e_64 ; tmp = e
101 xor T1, g_64 ; T1 = f ^ g
102 ror tmp0, 23 ; 41 ; tmp = e ror 23
103 and T1, e_64 ; T1 = (f ^ g) & e
104 xor tmp0, e_64 ; tmp = (e ror 23) ^ e
105 xor T1, g_64 ; T1 = ((f ^ g) & e) ^ g = CH(e,f,g)
106 add T1, [WK_2(%%t)] ; W[t] + K[t] from message scheduler
107 ror tmp0, 4 ; 18 ; tmp = ((e ror 23) ^ e) ror 4
108 xor tmp0, e_64 ; tmp = (((e ror 23) ^ e) ror 4) ^ e
109 mov T2, a_64 ; T2 = a
110 add T1, h_64 ; T1 = CH(e,f,g) + W[t] + K[t] + h
111 ror tmp0, 14 ; 14 ; tmp = ((((e ror23)^e)ror4)^e)ror14 = S1(e)
112 add T1, tmp0 ; T1 = CH(e,f,g) + W[t] + K[t] + S1(e)
113 mov tmp0, a_64 ; tmp = a
114 xor T2, c_64 ; T2 = a ^ c
115 and tmp0, c_64 ; tmp = a & c
116 and T2, b_64 ; T2 = (a ^ c) & b
117 xor T2, tmp0 ; T2 = ((a ^ c) & b) ^ (a & c) = Maj(a,b,c)
118 mov tmp0, a_64 ; tmp = a
119 ror tmp0, 5 ; 39 ; tmp = a ror 5
120 xor tmp0, a_64 ; tmp = (a ror 5) ^ a
121 add d_64, T1 ; e(next_state) = d + T1
122 ror tmp0, 6 ; 34 ; tmp = ((a ror 5) ^ a) ror 6
123 xor tmp0, a_64 ; tmp = (((a ror 5) ^ a) ror 6) ^ a
124 lea h_64, [T1 + T2] ; a(next_state) = T1 + Maj(a,b,c)
125 ror tmp0, 28 ; 28 ; tmp = ((((a ror5)^a)ror6)^a)ror28 = S0(a)
126 add h_64, tmp0 ; a(next_state) = T1 + Maj(a,b,c) S0(a)
130 %macro SHA512_2Sched_2Round_sse 1
133 ; Compute rounds %%t-2 and %%t-1
134 ; Compute message schedule QWORDS %%t and %%t+1
136 ; Two rounds are computed based on the values for K[t-2]+W[t-2] and
137 ; K[t-1]+W[t-1] which were previously stored at WK_2 by the message
139 ; The two new schedule QWORDS are stored at [W_t(%%t)] and [W_t(%%t+1)].
140 ; They are then added to their respective SHA512 constants at
141 ; [K_t(%%t)] and [K_t(%%t+1)] and stored at dqword [WK_2(%%t)]
142 ; For brievity, the comments following vectored instructions only refer to
143 ; the first of a pair of QWORDS.
144 ; Eg. XMM2=W[t-2] really means XMM2={W[t-2]|W[t-1]}
145 ; The computation of the message schedule and the rounds are tightly
146 ; stitched to take advantage of instruction-level parallelism.
147 ; For clarity, integer instructions (for the rounds calculation) are indented
148 ; by one tab. Vectored instructions (for the message scheduler) are indented
152 movdqa xmm2, [W_t(%%t-2)] ; XMM2 = W[t-2]
155 movdqa xmm0, xmm2 ; XMM0 = W[t-2]
158 movdqu xmm5, [W_t(%%t-15)] ; XMM5 = W[t-15]
161 movdqa xmm3, xmm5 ; XMM3 = W[t-15]
164 psrlq xmm0, 61 - 19 ; XMM0 = W[t-2] >> 42
167 psrlq xmm3, (8 - 7) ; XMM3 = W[t-15] >> 1
170 pxor xmm0, xmm2 ; XMM0 = (W[t-2] >> 42) ^ W[t-2]
173 pxor xmm3, xmm5 ; XMM3 = (W[t-15] >> 1) ^ W[t-15]
176 psrlq xmm0, 19 - 6 ; XMM0 = ((W[t-2]>>42)^W[t-2])>>13
179 psrlq xmm3, (7 - 1) ; XMM3 = ((W[t-15]>>1)^W[t-15])>>6
182 pxor xmm0, xmm2 ; XMM0 = (((W[t-2]>>42)^W[t-2])>>13)^W[t-2]
185 pxor xmm3, xmm5 ; XMM3 = (((W[t-15]>>1)^W[t-15])>>6)^W[t-15]
188 psrlq xmm0, 6 ; XMM0 = ((((W[t-2]>>42)^W[t-2])>>13)^W[t-2])>>6
191 psrlq xmm3, 1 ; XMM3 = (((W[t-15]>>1)^W[t-15])>>6)^W[t-15]>>1
194 movdqa xmm1, xmm2 ; XMM1 = W[t-2]
197 movdqa xmm4, xmm5 ; XMM4 = W[t-15]
200 psllq xmm1, (64 - 19) - (64 - 61) ; XMM1 = W[t-2] << 42
201 add T1, [WK_2(%%t+1)]
203 psllq xmm4, (64 - 1) - (64 - 8) ; XMM4 = W[t-15] << 7
206 pxor xmm1, xmm2 ; XMM1 = (W[t-2] << 42)^W[t-2]
209 pxor xmm4, xmm5 ; XMM4 = (W[t-15]<<7)^W[t-15]
212 psllq xmm1, (64 - 61) ; XMM1 = ((W[t-2] << 42)^W[t-2])<<3
215 psllq xmm4, (64 - 8) ; XMM4 = ((W[t-15]<<7)^W[t-15])<<56
218 pxor xmm0, xmm1 ; XMM0 = s1(W[t-2])
221 movdqu xmm1, [W_t(%%t- 7)] ; XMM1 = W[t-7]
223 pxor xmm3, xmm4 ; XMM3 = s0(W[t-15])
225 paddq xmm0, xmm3 ; XMM0 = s1(W[t-2]) + s0(W[t-15])
227 paddq xmm0, [W_t(%%t-16)] ; XMM0 = s1(W[t-2]) + s0(W[t-15]) + W[t-16]
229 paddq xmm0, xmm1 ; XMM0 = s1(W[t-2]) + W[t-7] + s0(W[t-15]) + W[t-16]
231 movdqa [W_t(%%t)], xmm0 ; Store scheduled qwords
233 paddq xmm0, [K_t(t)] ; Compute W[t]+K[t]
235 movdqa [WK_2(t)], xmm0 ; Store W[t]+K[t] for next rounds
242 ;;;;;;;;;;;;;;;;;;;;;;;;;;;;;;;;;;;;;;;;;;;;;;;;;;;;;;;;;;;;;;;;;;;;;;;;
243 ; void sha512_sse4(const void* M, void* D, uint64_t L);
244 ; Purpose: Updates the SHA512 digest stored at D with the message stored in M.
245 ; The size of the message pointed to by M must be an integer multiple of SHA512
247 ; L is the message length in SHA512 blocks.
248 mk_global sha512_sse4, function
254 ; Allocate Stack Space
258 mov [rsp + frame.GPRSAVE + 8 * 0], rbx
259 mov [rsp + frame.GPRSAVE + 8 * 1], r12
260 mov [rsp + frame.GPRSAVE + 8 * 2], r13
261 mov [rsp + frame.GPRSAVE + 8 * 3], r14
262 mov [rsp + frame.GPRSAVE + 8 * 4], r15
263 %ifidn __OUTPUT_FORMAT__, win64
264 mov [rsp + frame.GPRSAVE + 8 * 5], rsi
265 mov [rsp + frame.GPRSAVE + 8 * 6], rdi
270 ; Load state variables
271 mov a_64, [DIGEST(0)]
272 mov b_64, [DIGEST(1)]
273 mov c_64, [DIGEST(2)]
274 mov d_64, [DIGEST(3)]
275 mov e_64, [DIGEST(4)]
276 mov f_64, [DIGEST(5)]
277 mov g_64, [DIGEST(6)]
278 mov h_64, [DIGEST(7)]
282 ; (80 rounds) / (2 rounds/iteration) + (1 iteration)
283 ; +1 iteration because the scheduler leads hashing by 1 iteration
286 movdqa xmm1, [XMM_QWORD_BSWAP]
287 movdqu xmm0, [MSG(t)]
288 pshufb xmm0, xmm1 ; BSWAP
289 movdqa [W_t(t)], xmm0 ; Store Scheduled Pair
290 paddq xmm0, [K_t(t)] ; Compute W[t]+K[t]
291 movdqa [WK_2(t)], xmm0 ; Store into WK for rounds
293 ; BSWAP 2 QWORDS; Compute 2 Rounds
294 movdqu xmm0, [MSG(t)]
295 pshufb xmm0, xmm1 ; BSWAP
296 SHA512_Round t - 2 ; Round t-2
297 movdqa [W_t(t)], xmm0 ; Store Scheduled Pair
298 paddq xmm0, [K_t(t)] ; Compute W[t]+K[t]
299 SHA512_Round t - 1 ; Round t-1
300 movdqa [WK_2(t)], xmm0 ; Store W[t]+K[t] into WK
302 ; Schedule 2 QWORDS; Compute 2 Rounds
303 SHA512_2Sched_2Round_sse t
313 add [DIGEST(0)], a_64
314 add [DIGEST(1)], b_64
315 add [DIGEST(2)], c_64
316 add [DIGEST(3)], d_64
317 add [DIGEST(4)], e_64
318 add [DIGEST(5)], f_64
319 add [DIGEST(6)], g_64
320 add [DIGEST(7)], h_64
322 ; Advance to next message block
328 mov rbx, [rsp + frame.GPRSAVE + 8 * 0]
329 mov r12, [rsp + frame.GPRSAVE + 8 * 1]
330 mov r13, [rsp + frame.GPRSAVE + 8 * 2]
331 mov r14, [rsp + frame.GPRSAVE + 8 * 3]
332 mov r15, [rsp + frame.GPRSAVE + 8 * 4]
333 %ifidn __OUTPUT_FORMAT__, win64
334 mov rsi, [rsp + frame.GPRSAVE + 8 * 5]
335 mov rdi, [rsp + frame.GPRSAVE + 8 * 6]
337 ; Restore Stack Pointer
343 ;;;;;;;;;;;;;;;;;;;;;;;;;;;;;;;;;;;;;;;;;;;;;;;;;;;;;;;;;;;;;;;;;;;;;;;;
350 ; Mask for byte-swapping a couple of qwords in an XMM register using (v)pshufb.
352 dq 0x0001020304050607, 0x08090a0b0c0d0e0f
354 ; K[t] used in SHA512 hashing
356 dq 0x428a2f98d728ae22,0x7137449123ef65cd
357 dq 0xb5c0fbcfec4d3b2f,0xe9b5dba58189dbbc
358 dq 0x3956c25bf348b538,0x59f111f1b605d019
359 dq 0x923f82a4af194f9b,0xab1c5ed5da6d8118
360 dq 0xd807aa98a3030242,0x12835b0145706fbe
361 dq 0x243185be4ee4b28c,0x550c7dc3d5ffb4e2
362 dq 0x72be5d74f27b896f,0x80deb1fe3b1696b1
363 dq 0x9bdc06a725c71235,0xc19bf174cf692694
364 dq 0xe49b69c19ef14ad2,0xefbe4786384f25e3
365 dq 0x0fc19dc68b8cd5b5,0x240ca1cc77ac9c65
366 dq 0x2de92c6f592b0275,0x4a7484aa6ea6e483
367 dq 0x5cb0a9dcbd41fbd4,0x76f988da831153b5
368 dq 0x983e5152ee66dfab,0xa831c66d2db43210
369 dq 0xb00327c898fb213f,0xbf597fc7beef0ee4
370 dq 0xc6e00bf33da88fc2,0xd5a79147930aa725
371 dq 0x06ca6351e003826f,0x142929670a0e6e70
372 dq 0x27b70a8546d22ffc,0x2e1b21385c26c926
373 dq 0x4d2c6dfc5ac42aed,0x53380d139d95b3df
374 dq 0x650a73548baf63de,0x766a0abb3c77b2a8
375 dq 0x81c2c92e47edaee6,0x92722c851482353b
376 dq 0xa2bfe8a14cf10364,0xa81a664bbc423001
377 dq 0xc24b8b70d0f89791,0xc76c51a30654be30
378 dq 0xd192e819d6ef5218,0xd69906245565a910
379 dq 0xf40e35855771202a,0x106aa07032bbd1b8
380 dq 0x19a4c116b8d2d0c8,0x1e376c085141ab53
381 dq 0x2748774cdf8eeb99,0x34b0bcb5e19b48a8
382 dq 0x391c0cb3c5c95a63,0x4ed8aa4ae3418acb
383 dq 0x5b9cca4f7763e373,0x682e6ff3d6b2b8a3
384 dq 0x748f82ee5defb2fc,0x78a5636f43172f60
385 dq 0x84c87814a1f0ab72,0x8cc702081a6439ec
386 dq 0x90befffa23631e28,0xa4506cebde82bde9
387 dq 0xbef9a3f7b2c67915,0xc67178f2e372532b
388 dq 0xca273eceea26619c,0xd186b8c721c0c207
389 dq 0xeada7dd6cde0eb1e,0xf57d4f7fee6ed178
390 dq 0x06f067aa72176fba,0x0a637dc5a2c898a6
391 dq 0x113f9804bef90dae,0x1b710b35131c471b
392 dq 0x28db77f523047d84,0x32caab7b40c72493
393 dq 0x3c9ebe0a15c9bebc,0x431d67c49c100d4c
394 dq 0x4cc5d4becb3e42b6,0x597f299cfc657e2a
395 dq 0x5fcb6fab3ad6faec,0x6c44198c4a475817