arch/x86/crypto/sha256-ssse3-asm.S

   1 ########################################################################
   2 # Implement fast SHA-256 with SSSE3 instructions. (x86_64)
   3 #
   4 # Copyright (C) 2013 Intel Corporation.
   5 #
   6 # Authors:
   7 #     James Guilford <james.guilford@intel.com>
   8 #     Kirk Yap <kirk.s.yap@intel.com>
   9 #     Tim Chen <tim.c.chen@linux.intel.com>
  10 #
  11 # This software is available to you under a choice of one of two
  12 # licenses.  You may choose to be licensed under the terms of the GNU
  13 # General Public License (GPL) Version 2, available from the file
  14 # COPYING in the main directory of this source tree, or the
  15 # OpenIB.org BSD license below:
  16 #
  17 #     Redistribution and use in source and binary forms, with or
  18 #     without modification, are permitted provided that the following
  19 #     conditions are met:
  20 #
  21 #      - Redistributions of source code must retain the above
  22 #        copyright notice, this list of conditions and the following
  23 #        disclaimer.
  24 #
  25 #      - Redistributions in binary form must reproduce the above
  26 #        copyright notice, this list of conditions and the following
  27 #        disclaimer in the documentation and/or other materials
  28 #        provided with the distribution.
  29 #
  30 # THE SOFTWARE IS PROVIDED "AS IS", WITHOUT WARRANTY OF ANY KIND,
  31 # EXPRESS OR IMPLIED, INCLUDING BUT NOT LIMITED TO THE WARRANTIES OF
  32 # MERCHANTABILITY, FITNESS FOR A PARTICULAR PURPOSE AND
  33 # NONINFRINGEMENT. IN NO EVENT SHALL THE AUTHORS OR COPYRIGHT HOLDERS
  34 # BE LIABLE FOR ANY CLAIM, DAMAGES OR OTHER LIABILITY, WHETHER IN AN
  35 # ACTION OF CONTRACT, TORT OR OTHERWISE, ARISING FROM, OUT OF OR IN
  36 # CONNECTION WITH THE SOFTWARE OR THE USE OR OTHER DEALINGS IN THE
  37 # SOFTWARE.
  38 #
  39 ########################################################################
  40 #
  41 # This code is described in an Intel White-Paper:
  42 # "Fast SHA-256 Implementations on Intel Architecture Processors"
  43 #
  44 # To find it, surf to http://www.intel.com/p/en_US/embedded
  45 # and search for that title.
  46 #
  47 ########################################################################
  48
  49 #include <linux/linkage.h>
  50
  51 ## assume buffers not aligned
  52 #define    MOVDQ movdqu
  53
  54 ################################ Define Macros
  55
  56 # addm [mem], reg
  57 # Add reg to mem using reg-mem add and store
  58 .macro addm p1 p2
  59         add     \p1, \p2
  60         mov     \p2, \p1
  61 .endm
  62
  63 ################################
  64
  65 # COPY_XMM_AND_BSWAP xmm, [mem], byte_flip_mask
  66 # Load xmm with mem and byte swap each dword
  67 .macro COPY_XMM_AND_BSWAP p1 p2 p3
  68         MOVDQ \p2, \p1
  69         pshufb \p3, \p1
  70 .endm
  71
  72 ################################
  73
  74 X0 = %xmm4
  75 X1 = %xmm5
  76 X2 = %xmm6
  77 X3 = %xmm7
  78
  79 XTMP0 = %xmm0
  80 XTMP1 = %xmm1
  81 XTMP2 = %xmm2
  82 XTMP3 = %xmm3
  83 XTMP4 = %xmm8
  84 XFER = %xmm9
  85
  86 SHUF_00BA = %xmm10      # shuffle xBxA -> 00BA
  87 SHUF_DC00 = %xmm11      # shuffle xDxC -> DC00
  88 BYTE_FLIP_MASK = %xmm12
  89
  90 NUM_BLKS = %rdx   # 3rd arg
  91 CTX = %rsi        # 2nd arg
  92 INP = %rdi        # 1st arg
  93
  94 SRND = %rdi       # clobbers INP
  95 c = %ecx
  96 d = %r8d
  97 e = %edx
  98 TBL = %rbp
  99 a = %eax
 100 b = %ebx
 101
 102 f = %r9d
 103 g = %r10d
 104 h = %r11d
 105
 106 y0 = %r13d
 107 y1 = %r14d
 108 y2 = %r15d
 109
 110
 111
 112 _INP_END_SIZE = 8
 113 _INP_SIZE = 8
 114 _XFER_SIZE = 8
 115 _XMM_SAVE_SIZE = 0
 116
 117 _INP_END = 0
 118 _INP            = _INP_END  + _INP_END_SIZE
 119 _XFER           = _INP      + _INP_SIZE
 120 _XMM_SAVE       = _XFER     + _XFER_SIZE
 121 STACK_SIZE      = _XMM_SAVE + _XMM_SAVE_SIZE
 122
 123 # rotate_Xs
 124 # Rotate values of symbols X0...X3
 125 .macro rotate_Xs
 126 X_ = X0
 127 X0 = X1
 128 X1 = X2
 129 X2 = X3
 130 X3 = X_
 131 .endm
 132
 133 # ROTATE_ARGS
 134 # Rotate values of symbols a...h
 135 .macro ROTATE_ARGS
 136 TMP_ = h
 137 h = g
 138 g = f
 139 f = e
 140 e = d
 141 d = c
 142 c = b
 143 b = a
 144 a = TMP_
 145 .endm
 146
 147 .macro FOUR_ROUNDS_AND_SCHED
 148         ## compute s0 four at a time and s1 two at a time
 149         ## compute W[-16] + W[-7] 4 at a time
 150         movdqa  X3, XTMP0
 151         mov     e, y0                   # y0 = e
 152         ror     $(25-11), y0            # y0 = e >> (25-11)
 153         mov     a, y1                   # y1 = a
 154         palignr $4, X2, XTMP0           # XTMP0 = W[-7]
 155         ror     $(22-13), y1            # y1 = a >> (22-13)
 156         xor     e, y0                   # y0 = e ^ (e >> (25-11))
 157         mov     f, y2                   # y2 = f
 158         ror     $(11-6), y0             # y0 = (e >> (11-6)) ^ (e >> (25-6))
 159         movdqa  X1, XTMP1
 160         xor     a, y1                   # y1 = a ^ (a >> (22-13)
 161         xor     g, y2                   # y2 = f^g
 162         paddd   X0, XTMP0               # XTMP0 = W[-7] + W[-16]
 163         xor     e, y0                   # y0 = e ^ (e >> (11-6)) ^ (e >> (25-6))
 164         and     e, y2                   # y2 = (f^g)&e
 165         ror     $(13-2), y1             # y1 = (a >> (13-2)) ^ (a >> (22-2))
 166         ## compute s0
 167         palignr $4, X0, XTMP1           # XTMP1 = W[-15]
 168         xor     a, y1                   # y1 = a ^ (a >> (13-2)) ^ (a >> (22-2))
 169         ror     $6, y0                  # y0 = S1 = (e>>6) & (e>>11) ^ (e>>25)
 170         xor     g, y2                   # y2 = CH = ((f^g)&e)^g
 171         movdqa  XTMP1, XTMP2            # XTMP2 = W[-15]
 172         ror     $2, y1                  # y1 = S0 = (a>>2) ^ (a>>13) ^ (a>>22)
 173         add     y0, y2                  # y2 = S1 + CH
 174         add     _XFER(%rsp) , y2        # y2 = k + w + S1 + CH
 175         movdqa  XTMP1, XTMP3            # XTMP3 = W[-15]
 176         mov     a, y0                   # y0 = a
 177         add     y2, h                   # h = h + S1 + CH + k + w
 178         mov     a, y2                   # y2 = a
 179         pslld   $(32-7), XTMP1          #
 180         or      c, y0                   # y0 = a|c
 181         add     h, d                    # d = d + h + S1 + CH + k + w
 182         and     c, y2                   # y2 = a&c
 183         psrld   $7, XTMP2               #
 184         and     b, y0                   # y0 = (a|c)&b
 185         add     y1, h                   # h = h + S1 + CH + k + w + S0
 186         por     XTMP2, XTMP1            # XTMP1 = W[-15] ror 7
 187         or      y2, y0                  # y0 = MAJ = (a|c)&b)|(a&c)
 188         add     y0, h                   # h = h + S1 + CH + k + w + S0 + MAJ
 189                                         #
 190         ROTATE_ARGS                     #
 191         movdqa  XTMP3, XTMP2            # XTMP2 = W[-15]
 192         mov     e, y0                   # y0 = e
 193         mov     a, y1                   # y1 = a
 194         movdqa  XTMP3, XTMP4            # XTMP4 = W[-15]
 195         ror     $(25-11), y0            # y0 = e >> (25-11)
 196         xor     e, y0                   # y0 = e ^ (e >> (25-11))
 197         mov     f, y2                   # y2 = f
 198         ror     $(22-13), y1            # y1 = a >> (22-13)
 199         pslld   $(32-18), XTMP3         #
 200         xor     a, y1                   # y1 = a ^ (a >> (22-13)
 201         ror     $(11-6), y0             # y0 = (e >> (11-6)) ^ (e >> (25-6))
 202         xor     g, y2                   # y2 = f^g
 203         psrld   $18, XTMP2              #
 204         ror     $(13-2), y1             # y1 = (a >> (13-2)) ^ (a >> (22-2))
 205         xor     e, y0                   # y0 = e ^ (e >> (11-6)) ^ (e >> (25-6))
 206         and     e, y2                   # y2 = (f^g)&e
 207         ror     $6, y0                  # y0 = S1 = (e>>6) & (e>>11) ^ (e>>25)
 208         pxor    XTMP3, XTMP1
 209         xor     a, y1                   # y1 = a ^ (a >> (13-2)) ^ (a >> (22-2))
 210         xor     g, y2                   # y2 = CH = ((f^g)&e)^g
 211         psrld   $3, XTMP4               # XTMP4 = W[-15] >> 3
 212         add     y0, y2                  # y2 = S1 + CH
 213         add     (1*4 + _XFER)(%rsp), y2 # y2 = k + w + S1 + CH
 214         ror     $2, y1                  # y1 = S0 = (a>>2) ^ (a>>13) ^ (a>>22)
 215         pxor    XTMP2, XTMP1            # XTMP1 = W[-15] ror 7 ^ W[-15] ror 18
 216         mov     a, y0                   # y0 = a
 217         add     y2, h                   # h = h + S1 + CH + k + w
 218         mov     a, y2                   # y2 = a
 219         pxor    XTMP4, XTMP1            # XTMP1 = s0
 220         or      c, y0                   # y0 = a|c
 221         add     h, d                    # d = d + h + S1 + CH + k + w
 222         and     c, y2                   # y2 = a&c
 223         ## compute low s1
 224         pshufd  $0b11111010, X3, XTMP2   # XTMP2 = W[-2] {BBAA}
 225         and     b, y0                   # y0 = (a|c)&b
 226         add     y1, h                   # h = h + S1 + CH + k + w + S0
 227         paddd   XTMP1, XTMP0            # XTMP0 = W[-16] + W[-7] + s0
 228         or      y2, y0                  # y0 = MAJ = (a|c)&b)|(a&c)
 229         add     y0, h                   # h = h + S1 + CH + k + w + S0 + MAJ
 230
 231         ROTATE_ARGS
 232         movdqa  XTMP2, XTMP3            # XTMP3 = W[-2] {BBAA}
 233         mov     e, y0                   # y0 = e
 234         mov     a, y1                   # y1 = a
 235         ror     $(25-11), y0            # y0 = e >> (25-11)
 236         movdqa  XTMP2, XTMP4            # XTMP4 = W[-2] {BBAA}
 237         xor     e, y0                   # y0 = e ^ (e >> (25-11))
 238         ror     $(22-13), y1            # y1 = a >> (22-13)
 239         mov     f, y2                   # y2 = f
 240         xor     a, y1                   # y1 = a ^ (a >> (22-13)
 241         ror     $(11-6), y0             # y0 = (e >> (11-6)) ^ (e >> (25-6))
 242         psrlq   $17, XTMP2              # XTMP2 = W[-2] ror 17 {xBxA}
 243         xor     g, y2                   # y2 = f^g
 244         psrlq   $19, XTMP3              # XTMP3 = W[-2] ror 19 {xBxA}
 245         xor     e, y0                   # y0 = e ^ (e >> (11-6)) ^ (e >> (25-6))
 246         and     e, y2                   # y2 = (f^g)&e
 247         psrld   $10, XTMP4              # XTMP4 = W[-2] >> 10 {BBAA}
 248         ror     $(13-2), y1             # y1 = (a >> (13-2)) ^ (a >> (22-2))
 249         xor     a, y1                   # y1 = a ^ (a >> (13-2)) ^ (a >> (22-2))
 250         xor     g, y2                   # y2 = CH = ((f^g)&e)^g
 251         ror     $6, y0                  # y0 = S1 = (e>>6) & (e>>11) ^ (e>>25)
 252         pxor    XTMP3, XTMP2
 253         add     y0, y2                  # y2 = S1 + CH
 254         ror     $2, y1                  # y1 = S0 = (a>>2) ^ (a>>13) ^ (a>>22)
 255         add     (2*4 + _XFER)(%rsp), y2 # y2 = k + w + S1 + CH
 256         pxor    XTMP2, XTMP4            # XTMP4 = s1 {xBxA}
 257         mov     a, y0                   # y0 = a
 258         add     y2, h                   # h = h + S1 + CH + k + w
 259         mov     a, y2                   # y2 = a
 260         pshufb  SHUF_00BA, XTMP4        # XTMP4 = s1 {00BA}
 261         or      c, y0                   # y0 = a|c
 262         add     h, d                    # d = d + h + S1 + CH + k + w
 263         and     c, y2                   # y2 = a&c
 264         paddd   XTMP4, XTMP0            # XTMP0 = {..., ..., W[1], W[0]}
 265         and     b, y0                   # y0 = (a|c)&b
 266         add     y1, h                   # h = h + S1 + CH + k + w + S0
 267         ## compute high s1
 268         pshufd  $0b01010000, XTMP0, XTMP2 # XTMP2 = W[-2] {BBAA}
 269         or      y2, y0                  # y0 = MAJ = (a|c)&b)|(a&c)
 270         add     y0, h                   # h = h + S1 + CH + k + w + S0 + MAJ
 271                                         #
 272         ROTATE_ARGS                     #
 273         movdqa  XTMP2, XTMP3            # XTMP3 = W[-2] {DDCC}
 274         mov     e, y0                   # y0 = e
 275         ror     $(25-11), y0            # y0 = e >> (25-11)
 276         mov     a, y1                   # y1 = a
 277         movdqa  XTMP2, X0               # X0    = W[-2] {DDCC}
 278         ror     $(22-13), y1            # y1 = a >> (22-13)
 279         xor     e, y0                   # y0 = e ^ (e >> (25-11))
 280         mov     f, y2                   # y2 = f
 281         ror     $(11-6), y0             # y0 = (e >> (11-6)) ^ (e >> (25-6))
 282         psrlq   $17, XTMP2              # XTMP2 = W[-2] ror 17 {xDxC}
 283         xor     a, y1                   # y1 = a ^ (a >> (22-13)
 284         xor     g, y2                   # y2 = f^g
 285         psrlq   $19, XTMP3              # XTMP3 = W[-2] ror 19 {xDxC}
 286         xor     e, y0                   # y0 = e ^ (e >> (11-6)) ^ (e >> (25
 287         and     e, y2                   # y2 = (f^g)&e
 288         ror     $(13-2), y1             # y1 = (a >> (13-2)) ^ (a >> (22-2))
 289         psrld   $10, X0                 # X0 = W[-2] >> 10 {DDCC}
 290         xor     a, y1                   # y1 = a ^ (a >> (13-2)) ^ (a >> (22
 291         ror     $6, y0                  # y0 = S1 = (e>>6) & (e>>11) ^ (e>>2
 292         xor     g, y2                   # y2 = CH = ((f^g)&e)^g
 293         pxor    XTMP3, XTMP2            #
 294         ror     $2, y1                  # y1 = S0 = (a>>2) ^ (a>>13) ^ (a>>2
 295         add     y0, y2                  # y2 = S1 + CH
 296         add     (3*4 + _XFER)(%rsp), y2 # y2 = k + w + S1 + CH
 297         pxor    XTMP2, X0               # X0 = s1 {xDxC}
 298         mov     a, y0                   # y0 = a
 299         add     y2, h                   # h = h + S1 + CH + k + w
 300         mov     a, y2                   # y2 = a
 301         pshufb  SHUF_DC00, X0           # X0 = s1 {DC00}
 302         or      c, y0                   # y0 = a|c
 303         add     h, d                    # d = d + h + S1 + CH + k + w
 304         and     c, y2                   # y2 = a&c
 305         paddd   XTMP0, X0               # X0 = {W[3], W[2], W[1], W[0]}
 306         and     b, y0                   # y0 = (a|c)&b
 307         add     y1, h                   # h = h + S1 + CH + k + w + S0
 308         or      y2, y0                  # y0 = MAJ = (a|c)&b)|(a&c)
 309         add     y0, h                   # h = h + S1 + CH + k + w + S0 + MAJ
 310
 311         ROTATE_ARGS
 312         rotate_Xs
 313 .endm
 314
 315 ## input is [rsp + _XFER + %1 * 4]
 316 .macro DO_ROUND round
 317         mov     e, y0                 # y0 = e
 318         ror     $(25-11), y0          # y0 = e >> (25-11)
 319         mov     a, y1                 # y1 = a
 320         xor     e, y0                 # y0 = e ^ (e >> (25-11))
 321         ror     $(22-13), y1          # y1 = a >> (22-13)
 322         mov     f, y2                 # y2 = f
 323         xor     a, y1                 # y1 = a ^ (a >> (22-13)
 324         ror     $(11-6), y0           # y0 = (e >> (11-6)) ^ (e >> (25-6))
 325         xor     g, y2                 # y2 = f^g
 326         xor     e, y0                 # y0 = e ^ (e >> (11-6)) ^ (e >> (25-6))
 327         ror     $(13-2), y1           # y1 = (a >> (13-2)) ^ (a >> (22-2))
 328         and     e, y2                 # y2 = (f^g)&e
 329         xor     a, y1                 # y1 = a ^ (a >> (13-2)) ^ (a >> (22-2))
 330         ror     $6, y0                # y0 = S1 = (e>>6) & (e>>11) ^ (e>>25)
 331         xor     g, y2                 # y2 = CH = ((f^g)&e)^g
 332         add     y0, y2                # y2 = S1 + CH
 333         ror     $2, y1                # y1 = S0 = (a>>2) ^ (a>>13) ^ (a>>22)
 334         offset = \round * 4 + _XFER
 335         add     offset(%rsp), y2      # y2 = k + w + S1 + CH
 336         mov     a, y0                 # y0 = a
 337         add     y2, h                 # h = h + S1 + CH + k + w
 338         mov     a, y2                 # y2 = a
 339         or      c, y0                 # y0 = a|c
 340         add     h, d                  # d = d + h + S1 + CH + k + w
 341         and     c, y2                 # y2 = a&c
 342         and     b, y0                 # y0 = (a|c)&b
 343         add     y1, h                 # h = h + S1 + CH + k + w + S0
 344         or      y2, y0                # y0 = MAJ = (a|c)&b)|(a&c)
 345         add     y0, h                 # h = h + S1 + CH + k + w + S0 + MAJ
 346         ROTATE_ARGS
 347 .endm
 348
 349 ########################################################################
 350 ## void sha256_transform_ssse3(void *input_data, UINT32 digest[8], UINT64 num_blks)
 351 ## arg 1 : pointer to input data
 352 ## arg 2 : pointer to digest
 353 ## arg 3 : Num blocks
 354 ########################################################################
 355 .text
 356 ENTRY(sha256_transform_ssse3)
 357 .align 32
 358         pushq   %rbx
 359         pushq   %rbp
 360         pushq   %r13
 361         pushq   %r14
 362         pushq   %r15
 363         pushq   %r12
 364
 365         mov     %rsp, %r12
 366         subq    $STACK_SIZE, %rsp
 367         and     $~15, %rsp
 368
 369         shl     $6, NUM_BLKS             # convert to bytes
 370         jz      done_hash
 371         add     INP, NUM_BLKS
 372         mov     NUM_BLKS, _INP_END(%rsp) # pointer to end of data
 373
 374         ## load initial digest
 375         mov     4*0(CTX), a
 376         mov     4*1(CTX), b
 377         mov     4*2(CTX), c
 378         mov     4*3(CTX), d
 379         mov     4*4(CTX), e
 380         mov     4*5(CTX), f
 381         mov     4*6(CTX), g
 382         mov     4*7(CTX), h
 383
 384         movdqa  PSHUFFLE_BYTE_FLIP_MASK(%rip), BYTE_FLIP_MASK
 385         movdqa  _SHUF_00BA(%rip), SHUF_00BA
 386         movdqa  _SHUF_DC00(%rip), SHUF_DC00
 387
 388 loop0:
 389         lea     K256(%rip), TBL
 390
 391         ## byte swap first 16 dwords
 392         COPY_XMM_AND_BSWAP      X0, 0*16(INP), BYTE_FLIP_MASK
 393         COPY_XMM_AND_BSWAP      X1, 1*16(INP), BYTE_FLIP_MASK
 394         COPY_XMM_AND_BSWAP      X2, 2*16(INP), BYTE_FLIP_MASK
 395         COPY_XMM_AND_BSWAP      X3, 3*16(INP), BYTE_FLIP_MASK
 396
 397         mov     INP, _INP(%rsp)
 398
 399         ## schedule 48 input dwords, by doing 3 rounds of 16 each
 400         mov     $3, SRND
 401 .align 16
 402 loop1:
 403         movdqa  (TBL), XFER
 404         paddd   X0, XFER
 405         movdqa  XFER, _XFER(%rsp)
 406         FOUR_ROUNDS_AND_SCHED
 407
 408         movdqa  1*16(TBL), XFER
 409         paddd   X0, XFER
 410         movdqa  XFER, _XFER(%rsp)
 411         FOUR_ROUNDS_AND_SCHED
 412
 413         movdqa  2*16(TBL), XFER
 414         paddd   X0, XFER
 415         movdqa  XFER, _XFER(%rsp)
 416         FOUR_ROUNDS_AND_SCHED
 417
 418         movdqa  3*16(TBL), XFER
 419         paddd   X0, XFER
 420         movdqa  XFER, _XFER(%rsp)
 421         add     $4*16, TBL
 422         FOUR_ROUNDS_AND_SCHED
 423
 424         sub     $1, SRND
 425         jne     loop1
 426
 427         mov     $2, SRND
 428 loop2:
 429         paddd   (TBL), X0
 430         movdqa  X0, _XFER(%rsp)
 431         DO_ROUND        0
 432         DO_ROUND        1
 433         DO_ROUND        2
 434         DO_ROUND        3
 435         paddd   1*16(TBL), X1
 436         movdqa  X1, _XFER(%rsp)
 437         add     $2*16, TBL
 438         DO_ROUND        0
 439         DO_ROUND        1
 440         DO_ROUND        2
 441         DO_ROUND        3
 442
 443         movdqa  X2, X0
 444         movdqa  X3, X1
 445
 446         sub     $1, SRND
 447         jne     loop2
 448
 449         addm    (4*0)(CTX),a
 450         addm    (4*1)(CTX),b
 451         addm    (4*2)(CTX),c
 452         addm    (4*3)(CTX),d
 453         addm    (4*4)(CTX),e
 454         addm    (4*5)(CTX),f
 455         addm    (4*6)(CTX),g
 456         addm    (4*7)(CTX),h
 457
 458         mov     _INP(%rsp), INP
 459         add     $64, INP
 460         cmp     _INP_END(%rsp), INP
 461         jne     loop0
 462
 463 done_hash:
 464
 465         mov     %r12, %rsp
 466
 467         popq    %r12
 468         popq    %r15
 469         popq    %r14
 470         popq    %r13
 471         popq    %rbp
 472         popq    %rbx
 473
 474         ret
 475 ENDPROC(sha256_transform_ssse3)
 476
 477 .data
 478 .align 64
 479 K256:
 480         .long 0x428a2f98,0x71374491,0xb5c0fbcf,0xe9b5dba5
 481         .long 0x3956c25b,0x59f111f1,0x923f82a4,0xab1c5ed5
 482         .long 0xd807aa98,0x12835b01,0x243185be,0x550c7dc3
 483         .long 0x72be5d74,0x80deb1fe,0x9bdc06a7,0xc19bf174
 484         .long 0xe49b69c1,0xefbe4786,0x0fc19dc6,0x240ca1cc
 485         .long 0x2de92c6f,0x4a7484aa,0x5cb0a9dc,0x76f988da
 486         .long 0x983e5152,0xa831c66d,0xb00327c8,0xbf597fc7
 487         .long 0xc6e00bf3,0xd5a79147,0x06ca6351,0x14292967
 488         .long 0x27b70a85,0x2e1b2138,0x4d2c6dfc,0x53380d13
 489         .long 0x650a7354,0x766a0abb,0x81c2c92e,0x92722c85
 490         .long 0xa2bfe8a1,0xa81a664b,0xc24b8b70,0xc76c51a3
 491         .long 0xd192e819,0xd6990624,0xf40e3585,0x106aa070
 492         .long 0x19a4c116,0x1e376c08,0x2748774c,0x34b0bcb5
 493         .long 0x391c0cb3,0x4ed8aa4a,0x5b9cca4f,0x682e6ff3
 494         .long 0x748f82ee,0x78a5636f,0x84c87814,0x8cc70208
 495         .long 0x90befffa,0xa4506ceb,0xbef9a3f7,0xc67178f2
 496
 497 PSHUFFLE_BYTE_FLIP_MASK:
 498         .octa 0x0c0d0e0f08090a0b0405060700010203
 499
 500 # shuffle xBxA -> 00BA
 501 _SHUF_00BA:
 502         .octa 0xFFFFFFFFFFFFFFFF0b0a090803020100
 503
 504 # shuffle xDxC -> DC00
 505 _SHUF_DC00:
 506         .octa 0x0b0a090803020100FFFFFFFFFFFFFFFF