crypto/sha/asm/keccak1600-ppc64.pl

   1 #!/usr/bin/env perl
   2 # Copyright 2017-2018 The OpenSSL Project Authors. All Rights Reserved.
   3 #
   4 # Licensed under the OpenSSL license (the "License").  You may not use
   5 # this file except in compliance with the License.  You can obtain a copy
   6 # in the file LICENSE in the source distribution or at
   7 # https://www.openssl.org/source/license.html
   8 #
   9 # ====================================================================
  10 # Written by Andy Polyakov <appro@openssl.org> for the OpenSSL
  11 # project. The module is, however, dual licensed under OpenSSL and
  12 # CRYPTOGAMS licenses depending on where you obtain it. For further
  13 # details see http://www.openssl.org/~appro/cryptogams/.
  14 # ====================================================================
  15 #
  16 # Keccak-1600 for PPC64.
  17 #
  18 # June 2017.
  19 #
  20 # This is straightforward KECCAK_1X_ALT implementation that works on
  21 # *any* PPC64. Then PowerISA 2.07 adds 2x64-bit vector rotate, and
  22 # it's possible to achieve performance better than below, but that is
  23 # naturally option only for POWER8 and successors...
  24 #
  25 ######################################################################
  26 # Numbers are cycles per processed byte.
  27 #
  28 #               r=1088(*)
  29 #
  30 # PPC970/G5     14.6/+120%
  31 # POWER7        10.3/+100%
  32 # POWER8        11.5/+85%
  33 # POWER9        9.4/+45%
  34 #
  35 # (*)   Corresponds to SHA3-256. Percentage after slash is improvement
  36 #       over gcc-4.x-generated KECCAK_1X_ALT code. Newer compilers do
  37 #       much better (but watch out for them generating code specific
  38 #       to processor they execute on).
  39
  40 $flavour = shift;
  41
  42 if ($flavour =~ /64/) {
  43         $SIZE_T =8;
  44         $LRSAVE =2*$SIZE_T;
  45         $UCMP   ="cmpld";
  46         $STU    ="stdu";
  47         $POP    ="ld";
  48         $PUSH   ="std";
  49 } else { die "nonsense $flavour"; }
  50
  51 $0 =~ m/(.*[\/\\])[^\/\\]+$/; $dir=$1;
  52 ( $xlate="${dir}ppc-xlate.pl" and -f $xlate ) or
  53 ( $xlate="${dir}../../perlasm/ppc-xlate.pl" and -f $xlate) or
  54 die "can't locate ppc-xlate.pl";
  55
  56 open STDOUT,"| $^X $xlate $flavour ".shift || die "can't call $xlate: $!";
  57
  58 $FRAME=24*$SIZE_T+6*$SIZE_T+32;
  59 $LOCALS=6*$SIZE_T;
  60 $TEMP=$LOCALS+6*$SIZE_T;
  61
  62 my $sp ="r1";
  63
  64 my @A = map([ "r$_", "r".($_+1), "r".($_+2), "r".($_+3), "r".($_+4) ],
  65             (7, 12, 17, 22, 27));
  66    $A[1][1] = "r6"; # r13 is reserved
  67
  68 my @C = map("r$_", (0,3,4,5));
  69
  70 my @rhotates = ([  0,  1, 62, 28, 27 ],
  71                 [ 36, 44,  6, 55, 20 ],
  72                 [  3, 10, 43, 25, 39 ],
  73                 [ 41, 45, 15, 21,  8 ],
  74                 [ 18,  2, 61, 56, 14 ]);
  75
  76 $code.=<<___;
  77 .text
  78
  79 .type   KeccakF1600_int,\@function
  80 .align  5
  81 KeccakF1600_int:
  82         li      r0,24
  83         mtctr   r0
  84         b       .Loop
  85 .align  4
  86 .Loop:
  87         xor     $C[0],$A[0][0],$A[1][0]         ; Theta
  88         std     $A[0][4],`$TEMP+0`($sp)
  89         xor     $C[1],$A[0][1],$A[1][1]
  90         std     $A[1][4],`$TEMP+8`($sp)
  91         xor     $C[2],$A[0][2],$A[1][2]
  92         std     $A[2][4],`$TEMP+16`($sp)
  93         xor     $C[3],$A[0][3],$A[1][3]
  94         std     $A[3][4],`$TEMP+24`($sp)
  95 ___
  96         $C[4]=$A[0][4];
  97         $C[5]=$A[1][4];
  98         $C[6]=$A[2][4];
  99         $C[7]=$A[3][4];
 100 $code.=<<___;
 101         xor     $C[4],$A[0][4],$A[1][4]
 102         xor     $C[0],$C[0],$A[2][0]
 103         xor     $C[1],$C[1],$A[2][1]
 104         xor     $C[2],$C[2],$A[2][2]
 105         xor     $C[3],$C[3],$A[2][3]
 106         xor     $C[4],$C[4],$A[2][4]
 107         xor     $C[0],$C[0],$A[3][0]
 108         xor     $C[1],$C[1],$A[3][1]
 109         xor     $C[2],$C[2],$A[3][2]
 110         xor     $C[3],$C[3],$A[3][3]
 111         xor     $C[4],$C[4],$A[3][4]
 112         xor     $C[0],$C[0],$A[4][0]
 113         xor     $C[2],$C[2],$A[4][2]
 114         xor     $C[1],$C[1],$A[4][1]
 115         xor     $C[3],$C[3],$A[4][3]
 116         rotldi  $C[5],$C[2],1
 117         xor     $C[4],$C[4],$A[4][4]
 118         rotldi  $C[6],$C[3],1
 119         xor     $C[5],$C[5],$C[0]
 120         rotldi  $C[7],$C[4],1
 121
 122         xor     $A[0][1],$A[0][1],$C[5]
 123         xor     $A[1][1],$A[1][1],$C[5]
 124         xor     $A[2][1],$A[2][1],$C[5]
 125         xor     $A[3][1],$A[3][1],$C[5]
 126         xor     $A[4][1],$A[4][1],$C[5]
 127
 128         rotldi  $C[5],$C[0],1
 129         xor     $C[6],$C[6],$C[1]
 130         xor     $C[2],$C[2],$C[7]
 131         rotldi  $C[7],$C[1],1
 132         xor     $C[3],$C[3],$C[5]
 133         xor     $C[4],$C[4],$C[7]
 134
 135         xor     $C[1],   $A[0][2],$C[6]                 ;mr     $C[1],$A[0][2]
 136         xor     $A[1][2],$A[1][2],$C[6]
 137         xor     $A[2][2],$A[2][2],$C[6]
 138         xor     $A[3][2],$A[3][2],$C[6]
 139         xor     $A[4][2],$A[4][2],$C[6]
 140
 141         xor     $A[0][0],$A[0][0],$C[4]
 142         xor     $A[1][0],$A[1][0],$C[4]
 143         xor     $A[2][0],$A[2][0],$C[4]
 144         xor     $A[3][0],$A[3][0],$C[4]
 145         xor     $A[4][0],$A[4][0],$C[4]
 146 ___
 147         $C[4]=undef;
 148         $C[5]=undef;
 149         $C[6]=undef;
 150         $C[7]=undef;
 151 $code.=<<___;
 152         ld      $A[0][4],`$TEMP+0`($sp)
 153         xor     $C[0],   $A[0][3],$C[2]                 ;mr     $C[0],$A[0][3]
 154         ld      $A[1][4],`$TEMP+8`($sp)
 155         xor     $A[1][3],$A[1][3],$C[2]
 156         ld      $A[2][4],`$TEMP+16`($sp)
 157         xor     $A[2][3],$A[2][3],$C[2]
 158         ld      $A[3][4],`$TEMP+24`($sp)
 159         xor     $A[3][3],$A[3][3],$C[2]
 160         xor     $A[4][3],$A[4][3],$C[2]
 161
 162         xor     $C[2],   $A[0][4],$C[3]                 ;mr     $C[2],$A[0][4]
 163         xor     $A[1][4],$A[1][4],$C[3]
 164         xor     $A[2][4],$A[2][4],$C[3]
 165         xor     $A[3][4],$A[3][4],$C[3]
 166         xor     $A[4][4],$A[4][4],$C[3]
 167
 168         mr      $C[3],$A[0][1]                          ; Rho+Pi
 169         rotldi  $A[0][1],$A[1][1],$rhotates[1][1]
 170         ;mr     $C[1],$A[0][2]
 171         rotldi  $A[0][2],$A[2][2],$rhotates[2][2]
 172         ;mr     $C[0],$A[0][3]
 173         rotldi  $A[0][3],$A[3][3],$rhotates[3][3]
 174         ;mr     $C[2],$A[0][4]
 175         rotldi  $A[0][4],$A[4][4],$rhotates[4][4]
 176
 177         rotldi  $A[1][1],$A[1][4],$rhotates[1][4]
 178         rotldi  $A[2][2],$A[2][3],$rhotates[2][3]
 179         rotldi  $A[3][3],$A[3][2],$rhotates[3][2]
 180         rotldi  $A[4][4],$A[4][1],$rhotates[4][1]
 181
 182         rotldi  $A[1][4],$A[4][2],$rhotates[4][2]
 183         rotldi  $A[2][3],$A[3][4],$rhotates[3][4]
 184         rotldi  $A[3][2],$A[2][1],$rhotates[2][1]
 185         rotldi  $A[4][1],$A[1][3],$rhotates[1][3]
 186
 187         rotldi  $A[4][2],$A[2][4],$rhotates[2][4]
 188         rotldi  $A[3][4],$A[4][3],$rhotates[4][3]
 189         rotldi  $A[2][1],$A[1][2],$rhotates[1][2]
 190         rotldi  $A[1][3],$A[3][1],$rhotates[3][1]
 191
 192         rotldi  $A[2][4],$A[4][0],$rhotates[4][0]
 193         rotldi  $A[4][3],$A[3][0],$rhotates[3][0]
 194         rotldi  $A[1][2],$A[2][0],$rhotates[2][0]
 195         rotldi  $A[3][1],$A[1][0],$rhotates[1][0]
 196
 197         rotldi  $A[1][0],$C[0],$rhotates[0][3]
 198         rotldi  $A[2][0],$C[3],$rhotates[0][1]
 199         rotldi  $A[3][0],$C[2],$rhotates[0][4]
 200         rotldi  $A[4][0],$C[1],$rhotates[0][2]
 201
 202         andc    $C[0],$A[0][2],$A[0][1]                 ; Chi+Iota
 203         andc    $C[1],$A[0][3],$A[0][2]
 204         andc    $C[2],$A[0][0],$A[0][4]
 205         andc    $C[3],$A[0][1],$A[0][0]
 206         xor     $A[0][0],$A[0][0],$C[0]
 207         andc    $C[0],$A[0][4],$A[0][3]
 208         xor     $A[0][1],$A[0][1],$C[1]
 209          ld     $C[1],`$LOCALS+4*$SIZE_T`($sp)
 210         xor     $A[0][3],$A[0][3],$C[2]
 211         xor     $A[0][4],$A[0][4],$C[3]
 212         xor     $A[0][2],$A[0][2],$C[0]
 213          ldu    $C[3],8($C[1])                          ; Iota[i++]
 214
 215         andc    $C[0],$A[1][2],$A[1][1]
 216          std    $C[1],`$LOCALS+4*$SIZE_T`($sp)
 217         andc    $C[1],$A[1][3],$A[1][2]
 218         andc    $C[2],$A[1][0],$A[1][4]
 219          xor    $A[0][0],$A[0][0],$C[3]                 ; A[0][0] ^= Iota
 220         andc    $C[3],$A[1][1],$A[1][0]
 221         xor     $A[1][0],$A[1][0],$C[0]
 222         andc    $C[0],$A[1][4],$A[1][3]
 223         xor     $A[1][1],$A[1][1],$C[1]
 224         xor     $A[1][3],$A[1][3],$C[2]
 225         xor     $A[1][4],$A[1][4],$C[3]
 226         xor     $A[1][2],$A[1][2],$C[0]
 227
 228         andc    $C[0],$A[2][2],$A[2][1]
 229         andc    $C[1],$A[2][3],$A[2][2]
 230         andc    $C[2],$A[2][0],$A[2][4]
 231         andc    $C[3],$A[2][1],$A[2][0]
 232         xor     $A[2][0],$A[2][0],$C[0]
 233         andc    $C[0],$A[2][4],$A[2][3]
 234         xor     $A[2][1],$A[2][1],$C[1]
 235         xor     $A[2][3],$A[2][3],$C[2]
 236         xor     $A[2][4],$A[2][4],$C[3]
 237         xor     $A[2][2],$A[2][2],$C[0]
 238
 239         andc    $C[0],$A[3][2],$A[3][1]
 240         andc    $C[1],$A[3][3],$A[3][2]
 241         andc    $C[2],$A[3][0],$A[3][4]
 242         andc    $C[3],$A[3][1],$A[3][0]
 243         xor     $A[3][0],$A[3][0],$C[0]
 244         andc    $C[0],$A[3][4],$A[3][3]
 245         xor     $A[3][1],$A[3][1],$C[1]
 246         xor     $A[3][3],$A[3][3],$C[2]
 247         xor     $A[3][4],$A[3][4],$C[3]
 248         xor     $A[3][2],$A[3][2],$C[0]
 249
 250         andc    $C[0],$A[4][2],$A[4][1]
 251         andc    $C[1],$A[4][3],$A[4][2]
 252         andc    $C[2],$A[4][0],$A[4][4]
 253         andc    $C[3],$A[4][1],$A[4][0]
 254         xor     $A[4][0],$A[4][0],$C[0]
 255         andc    $C[0],$A[4][4],$A[4][3]
 256         xor     $A[4][1],$A[4][1],$C[1]
 257         xor     $A[4][3],$A[4][3],$C[2]
 258         xor     $A[4][4],$A[4][4],$C[3]
 259         xor     $A[4][2],$A[4][2],$C[0]
 260
 261         bdnz    .Loop
 262
 263         blr
 264         .long   0
 265         .byte   0,12,0x14,0,0,0,0,0
 266 .size   KeccakF1600_int,.-KeccakF1600_int
 267
 268 .type   KeccakF1600,\@function
 269 .align  5
 270 KeccakF1600:
 271         $STU    $sp,-$FRAME($sp)
 272         mflr    r0
 273         $PUSH   r14,`$FRAME-$SIZE_T*18`($sp)
 274         $PUSH   r15,`$FRAME-$SIZE_T*17`($sp)
 275         $PUSH   r16,`$FRAME-$SIZE_T*16`($sp)
 276         $PUSH   r17,`$FRAME-$SIZE_T*15`($sp)
 277         $PUSH   r18,`$FRAME-$SIZE_T*14`($sp)
 278         $PUSH   r19,`$FRAME-$SIZE_T*13`($sp)
 279         $PUSH   r20,`$FRAME-$SIZE_T*12`($sp)
 280         $PUSH   r21,`$FRAME-$SIZE_T*11`($sp)
 281         $PUSH   r22,`$FRAME-$SIZE_T*10`($sp)
 282         $PUSH   r23,`$FRAME-$SIZE_T*9`($sp)
 283         $PUSH   r24,`$FRAME-$SIZE_T*8`($sp)
 284         $PUSH   r25,`$FRAME-$SIZE_T*7`($sp)
 285         $PUSH   r26,`$FRAME-$SIZE_T*6`($sp)
 286         $PUSH   r27,`$FRAME-$SIZE_T*5`($sp)
 287         $PUSH   r28,`$FRAME-$SIZE_T*4`($sp)
 288         $PUSH   r29,`$FRAME-$SIZE_T*3`($sp)
 289         $PUSH   r30,`$FRAME-$SIZE_T*2`($sp)
 290         $PUSH   r31,`$FRAME-$SIZE_T*1`($sp)
 291         $PUSH   r0,`$FRAME+$LRSAVE`($sp)
 292
 293         bl      PICmeup
 294         subi    r12,r12,8                       ; prepare for ldu
 295
 296         $PUSH   r3,`$LOCALS+0*$SIZE_T`($sp)
 297         ;$PUSH  r4,`$LOCALS+1*$SIZE_T`($sp)
 298         ;$PUSH  r5,`$LOCALS+2*$SIZE_T`($sp)
 299         ;$PUSH  r6,`$LOCALS+3*$SIZE_T`($sp)
 300         $PUSH   r12,`$LOCALS+4*$SIZE_T`($sp)
 301
 302         ld      $A[0][0],`8*0`(r3)              ; load A[5][5]
 303         ld      $A[0][1],`8*1`(r3)
 304         ld      $A[0][2],`8*2`(r3)
 305         ld      $A[0][3],`8*3`(r3)
 306         ld      $A[0][4],`8*4`(r3)
 307         ld      $A[1][0],`8*5`(r3)
 308         ld      $A[1][1],`8*6`(r3)
 309         ld      $A[1][2],`8*7`(r3)
 310         ld      $A[1][3],`8*8`(r3)
 311         ld      $A[1][4],`8*9`(r3)
 312         ld      $A[2][0],`8*10`(r3)
 313         ld      $A[2][1],`8*11`(r3)
 314         ld      $A[2][2],`8*12`(r3)
 315         ld      $A[2][3],`8*13`(r3)
 316         ld      $A[2][4],`8*14`(r3)
 317         ld      $A[3][0],`8*15`(r3)
 318         ld      $A[3][1],`8*16`(r3)
 319         ld      $A[3][2],`8*17`(r3)
 320         ld      $A[3][3],`8*18`(r3)
 321         ld      $A[3][4],`8*19`(r3)
 322         ld      $A[4][0],`8*20`(r3)
 323         ld      $A[4][1],`8*21`(r3)
 324         ld      $A[4][2],`8*22`(r3)
 325         ld      $A[4][3],`8*23`(r3)
 326         ld      $A[4][4],`8*24`(r3)
 327
 328         bl      KeccakF1600_int
 329
 330         $POP    r3,`$LOCALS+0*$SIZE_T`($sp)
 331         std     $A[0][0],`8*0`(r3)              ; return A[5][5]
 332         std     $A[0][1],`8*1`(r3)
 333         std     $A[0][2],`8*2`(r3)
 334         std     $A[0][3],`8*3`(r3)
 335         std     $A[0][4],`8*4`(r3)
 336         std     $A[1][0],`8*5`(r3)
 337         std     $A[1][1],`8*6`(r3)
 338         std     $A[1][2],`8*7`(r3)
 339         std     $A[1][3],`8*8`(r3)
 340         std     $A[1][4],`8*9`(r3)
 341         std     $A[2][0],`8*10`(r3)
 342         std     $A[2][1],`8*11`(r3)
 343         std     $A[2][2],`8*12`(r3)
 344         std     $A[2][3],`8*13`(r3)
 345         std     $A[2][4],`8*14`(r3)
 346         std     $A[3][0],`8*15`(r3)
 347         std     $A[3][1],`8*16`(r3)
 348         std     $A[3][2],`8*17`(r3)
 349         std     $A[3][3],`8*18`(r3)
 350         std     $A[3][4],`8*19`(r3)
 351         std     $A[4][0],`8*20`(r3)
 352         std     $A[4][1],`8*21`(r3)
 353         std     $A[4][2],`8*22`(r3)
 354         std     $A[4][3],`8*23`(r3)
 355         std     $A[4][4],`8*24`(r3)
 356
 357         $POP    r0,`$FRAME+$LRSAVE`($sp)
 358         $POP    r14,`$FRAME-$SIZE_T*18`($sp)
 359         $POP    r15,`$FRAME-$SIZE_T*17`($sp)
 360         $POP    r16,`$FRAME-$SIZE_T*16`($sp)
 361         $POP    r17,`$FRAME-$SIZE_T*15`($sp)
 362         $POP    r18,`$FRAME-$SIZE_T*14`($sp)
 363         $POP    r19,`$FRAME-$SIZE_T*13`($sp)
 364         $POP    r20,`$FRAME-$SIZE_T*12`($sp)
 365         $POP    r21,`$FRAME-$SIZE_T*11`($sp)
 366         $POP    r22,`$FRAME-$SIZE_T*10`($sp)
 367         $POP    r23,`$FRAME-$SIZE_T*9`($sp)
 368         $POP    r24,`$FRAME-$SIZE_T*8`($sp)
 369         $POP    r25,`$FRAME-$SIZE_T*7`($sp)
 370         $POP    r26,`$FRAME-$SIZE_T*6`($sp)
 371         $POP    r27,`$FRAME-$SIZE_T*5`($sp)
 372         $POP    r28,`$FRAME-$SIZE_T*4`($sp)
 373         $POP    r29,`$FRAME-$SIZE_T*3`($sp)
 374         $POP    r30,`$FRAME-$SIZE_T*2`($sp)
 375         $POP    r31,`$FRAME-$SIZE_T*1`($sp)
 376         mtlr    r0
 377         addi    $sp,$sp,$FRAME
 378         blr
 379         .long   0
 380         .byte   0,12,4,1,0x80,18,1,0
 381         .long   0
 382 .size   KeccakF1600,.-KeccakF1600
 383
 384 .type   dword_le_load,\@function
 385 .align  5
 386 dword_le_load:
 387         lbzu    r0,1(r3)
 388         lbzu    r4,1(r3)
 389         lbzu    r5,1(r3)
 390         insrdi  r0,r4,8,48
 391         lbzu    r4,1(r3)
 392         insrdi  r0,r5,8,40
 393         lbzu    r5,1(r3)
 394         insrdi  r0,r4,8,32
 395         lbzu    r4,1(r3)
 396         insrdi  r0,r5,8,24
 397         lbzu    r5,1(r3)
 398         insrdi  r0,r4,8,16
 399         lbzu    r4,1(r3)
 400         insrdi  r0,r5,8,8
 401         insrdi  r0,r4,8,0
 402         blr
 403         .long   0
 404         .byte   0,12,0x14,0,0,0,1,0
 405         .long   0
 406 .size   dword_le_load,.-dword_le_load
 407
 408 .globl  SHA3_absorb
 409 .type   SHA3_absorb,\@function
 410 .align  5
 411 SHA3_absorb:
 412         $STU    $sp,-$FRAME($sp)
 413         mflr    r0
 414         $PUSH   r14,`$FRAME-$SIZE_T*18`($sp)
 415         $PUSH   r15,`$FRAME-$SIZE_T*17`($sp)
 416         $PUSH   r16,`$FRAME-$SIZE_T*16`($sp)
 417         $PUSH   r17,`$FRAME-$SIZE_T*15`($sp)
 418         $PUSH   r18,`$FRAME-$SIZE_T*14`($sp)
 419         $PUSH   r19,`$FRAME-$SIZE_T*13`($sp)
 420         $PUSH   r20,`$FRAME-$SIZE_T*12`($sp)
 421         $PUSH   r21,`$FRAME-$SIZE_T*11`($sp)
 422         $PUSH   r22,`$FRAME-$SIZE_T*10`($sp)
 423         $PUSH   r23,`$FRAME-$SIZE_T*9`($sp)
 424         $PUSH   r24,`$FRAME-$SIZE_T*8`($sp)
 425         $PUSH   r25,`$FRAME-$SIZE_T*7`($sp)
 426         $PUSH   r26,`$FRAME-$SIZE_T*6`($sp)
 427         $PUSH   r27,`$FRAME-$SIZE_T*5`($sp)
 428         $PUSH   r28,`$FRAME-$SIZE_T*4`($sp)
 429         $PUSH   r29,`$FRAME-$SIZE_T*3`($sp)
 430         $PUSH   r30,`$FRAME-$SIZE_T*2`($sp)
 431         $PUSH   r31,`$FRAME-$SIZE_T*1`($sp)
 432         $PUSH   r0,`$FRAME+$LRSAVE`($sp)
 433
 434         bl      PICmeup
 435         subi    r4,r4,1                         ; prepare for lbzu
 436         subi    r12,r12,8                       ; prepare for ldu
 437
 438         $PUSH   r3,`$LOCALS+0*$SIZE_T`($sp)     ; save A[][]
 439         $PUSH   r4,`$LOCALS+1*$SIZE_T`($sp)     ; save inp
 440         $PUSH   r5,`$LOCALS+2*$SIZE_T`($sp)     ; save len
 441         $PUSH   r6,`$LOCALS+3*$SIZE_T`($sp)     ; save bsz
 442         mr      r0,r6
 443         $PUSH   r12,`$LOCALS+4*$SIZE_T`($sp)
 444
 445         ld      $A[0][0],`8*0`(r3)              ; load A[5][5]
 446         ld      $A[0][1],`8*1`(r3)
 447         ld      $A[0][2],`8*2`(r3)
 448         ld      $A[0][3],`8*3`(r3)
 449         ld      $A[0][4],`8*4`(r3)
 450         ld      $A[1][0],`8*5`(r3)
 451         ld      $A[1][1],`8*6`(r3)
 452         ld      $A[1][2],`8*7`(r3)
 453         ld      $A[1][3],`8*8`(r3)
 454         ld      $A[1][4],`8*9`(r3)
 455         ld      $A[2][0],`8*10`(r3)
 456         ld      $A[2][1],`8*11`(r3)
 457         ld      $A[2][2],`8*12`(r3)
 458         ld      $A[2][3],`8*13`(r3)
 459         ld      $A[2][4],`8*14`(r3)
 460         ld      $A[3][0],`8*15`(r3)
 461         ld      $A[3][1],`8*16`(r3)
 462         ld      $A[3][2],`8*17`(r3)
 463         ld      $A[3][3],`8*18`(r3)
 464         ld      $A[3][4],`8*19`(r3)
 465         ld      $A[4][0],`8*20`(r3)
 466         ld      $A[4][1],`8*21`(r3)
 467         ld      $A[4][2],`8*22`(r3)
 468         ld      $A[4][3],`8*23`(r3)
 469         ld      $A[4][4],`8*24`(r3)
 470
 471         mr      r3,r4
 472         mr      r4,r5
 473         mr      r5,r0
 474
 475         b       .Loop_absorb
 476
 477 .align  4
 478 .Loop_absorb:
 479         $UCMP   r4,r5                           ; len < bsz?
 480         blt     .Labsorbed
 481
 482         sub     r4,r4,r5                        ; len -= bsz
 483         srwi    r5,r5,3
 484         $PUSH   r4,`$LOCALS+2*$SIZE_T`($sp)     ; save len
 485         mtctr   r5
 486         bl      dword_le_load                   ; *inp++
 487         xor     $A[0][0],$A[0][0],r0
 488         bdz     .Lprocess_block
 489         bl      dword_le_load                   ; *inp++
 490         xor     $A[0][1],$A[0][1],r0
 491         bdz     .Lprocess_block
 492         bl      dword_le_load                   ; *inp++
 493         xor     $A[0][2],$A[0][2],r0
 494         bdz     .Lprocess_block
 495         bl      dword_le_load                   ; *inp++
 496         xor     $A[0][3],$A[0][3],r0
 497         bdz     .Lprocess_block
 498         bl      dword_le_load                   ; *inp++
 499         xor     $A[0][4],$A[0][4],r0
 500         bdz     .Lprocess_block
 501         bl      dword_le_load                   ; *inp++
 502         xor     $A[1][0],$A[1][0],r0
 503         bdz     .Lprocess_block
 504         bl      dword_le_load                   ; *inp++
 505         xor     $A[1][1],$A[1][1],r0
 506         bdz     .Lprocess_block
 507         bl      dword_le_load                   ; *inp++
 508         xor     $A[1][2],$A[1][2],r0
 509         bdz     .Lprocess_block
 510         bl      dword_le_load                   ; *inp++
 511         xor     $A[1][3],$A[1][3],r0
 512         bdz     .Lprocess_block
 513         bl      dword_le_load                   ; *inp++
 514         xor     $A[1][4],$A[1][4],r0
 515         bdz     .Lprocess_block
 516         bl      dword_le_load                   ; *inp++
 517         xor     $A[2][0],$A[2][0],r0
 518         bdz     .Lprocess_block
 519         bl      dword_le_load                   ; *inp++
 520         xor     $A[2][1],$A[2][1],r0
 521         bdz     .Lprocess_block
 522         bl      dword_le_load                   ; *inp++
 523         xor     $A[2][2],$A[2][2],r0
 524         bdz     .Lprocess_block
 525         bl      dword_le_load                   ; *inp++
 526         xor     $A[2][3],$A[2][3],r0
 527         bdz     .Lprocess_block
 528         bl      dword_le_load                   ; *inp++
 529         xor     $A[2][4],$A[2][4],r0
 530         bdz     .Lprocess_block
 531         bl      dword_le_load                   ; *inp++
 532         xor     $A[3][0],$A[3][0],r0
 533         bdz     .Lprocess_block
 534         bl      dword_le_load                   ; *inp++
 535         xor     $A[3][1],$A[3][1],r0
 536         bdz     .Lprocess_block
 537         bl      dword_le_load                   ; *inp++
 538         xor     $A[3][2],$A[3][2],r0
 539         bdz     .Lprocess_block
 540         bl      dword_le_load                   ; *inp++
 541         xor     $A[3][3],$A[3][3],r0
 542         bdz     .Lprocess_block
 543         bl      dword_le_load                   ; *inp++
 544         xor     $A[3][4],$A[3][4],r0
 545         bdz     .Lprocess_block
 546         bl      dword_le_load                   ; *inp++
 547         xor     $A[4][0],$A[4][0],r0
 548         bdz     .Lprocess_block
 549         bl      dword_le_load                   ; *inp++
 550         xor     $A[4][1],$A[4][1],r0
 551         bdz     .Lprocess_block
 552         bl      dword_le_load                   ; *inp++
 553         xor     $A[4][2],$A[4][2],r0
 554         bdz     .Lprocess_block
 555         bl      dword_le_load                   ; *inp++
 556         xor     $A[4][3],$A[4][3],r0
 557         bdz     .Lprocess_block
 558         bl      dword_le_load                   ; *inp++
 559         xor     $A[4][4],$A[4][4],r0
 560
 561 .Lprocess_block:
 562         $PUSH   r3,`$LOCALS+1*$SIZE_T`($sp)     ; save inp
 563
 564         bl      KeccakF1600_int
 565
 566         $POP    r0,`$LOCALS+4*$SIZE_T`($sp)     ; pull iotas[24]
 567         $POP    r5,`$LOCALS+3*$SIZE_T`($sp)     ; restore bsz
 568         $POP    r4,`$LOCALS+2*$SIZE_T`($sp)     ; restore len
 569         $POP    r3,`$LOCALS+1*$SIZE_T`($sp)     ; restore inp
 570         addic   r0,r0,`-8*24`                   ; rewind iotas
 571         $PUSH   r0,`$LOCALS+4*$SIZE_T`($sp)
 572
 573         b       .Loop_absorb
 574
 575 .align  4
 576 .Labsorbed:
 577         $POP    r3,`$LOCALS+0*$SIZE_T`($sp)
 578         std     $A[0][0],`8*0`(r3)              ; return A[5][5]
 579         std     $A[0][1],`8*1`(r3)
 580         std     $A[0][2],`8*2`(r3)
 581         std     $A[0][3],`8*3`(r3)
 582         std     $A[0][4],`8*4`(r3)
 583         std     $A[1][0],`8*5`(r3)
 584         std     $A[1][1],`8*6`(r3)
 585         std     $A[1][2],`8*7`(r3)
 586         std     $A[1][3],`8*8`(r3)
 587         std     $A[1][4],`8*9`(r3)
 588         std     $A[2][0],`8*10`(r3)
 589         std     $A[2][1],`8*11`(r3)
 590         std     $A[2][2],`8*12`(r3)
 591         std     $A[2][3],`8*13`(r3)
 592         std     $A[2][4],`8*14`(r3)
 593         std     $A[3][0],`8*15`(r3)
 594         std     $A[3][1],`8*16`(r3)
 595         std     $A[3][2],`8*17`(r3)
 596         std     $A[3][3],`8*18`(r3)
 597         std     $A[3][4],`8*19`(r3)
 598         std     $A[4][0],`8*20`(r3)
 599         std     $A[4][1],`8*21`(r3)
 600         std     $A[4][2],`8*22`(r3)
 601         std     $A[4][3],`8*23`(r3)
 602         std     $A[4][4],`8*24`(r3)
 603
 604         mr      r3,r4                           ; return value
 605         $POP    r0,`$FRAME+$LRSAVE`($sp)
 606         $POP    r14,`$FRAME-$SIZE_T*18`($sp)
 607         $POP    r15,`$FRAME-$SIZE_T*17`($sp)
 608         $POP    r16,`$FRAME-$SIZE_T*16`($sp)
 609         $POP    r17,`$FRAME-$SIZE_T*15`($sp)
 610         $POP    r18,`$FRAME-$SIZE_T*14`($sp)
 611         $POP    r19,`$FRAME-$SIZE_T*13`($sp)
 612         $POP    r20,`$FRAME-$SIZE_T*12`($sp)
 613         $POP    r21,`$FRAME-$SIZE_T*11`($sp)
 614         $POP    r22,`$FRAME-$SIZE_T*10`($sp)
 615         $POP    r23,`$FRAME-$SIZE_T*9`($sp)
 616         $POP    r24,`$FRAME-$SIZE_T*8`($sp)
 617         $POP    r25,`$FRAME-$SIZE_T*7`($sp)
 618         $POP    r26,`$FRAME-$SIZE_T*6`($sp)
 619         $POP    r27,`$FRAME-$SIZE_T*5`($sp)
 620         $POP    r28,`$FRAME-$SIZE_T*4`($sp)
 621         $POP    r29,`$FRAME-$SIZE_T*3`($sp)
 622         $POP    r30,`$FRAME-$SIZE_T*2`($sp)
 623         $POP    r31,`$FRAME-$SIZE_T*1`($sp)
 624         mtlr    r0
 625         addi    $sp,$sp,$FRAME
 626         blr
 627         .long   0
 628         .byte   0,12,4,1,0x80,18,4,0
 629         .long   0
 630 .size   SHA3_absorb,.-SHA3_absorb
 631 ___
 632 {
 633 my ($A_flat,$out,$len,$bsz) = map("r$_",(28..31));
 634 $code.=<<___;
 635 .globl  SHA3_squeeze
 636 .type   SHA3_squeeze,\@function
 637 .align  5
 638 SHA3_squeeze:
 639         $STU    $sp,`-10*$SIZE_T`($sp)
 640         mflr    r0
 641         $PUSH   r28,`6*$SIZE_T`($sp)
 642         $PUSH   r29,`7*$SIZE_T`($sp)
 643         $PUSH   r30,`8*$SIZE_T`($sp)
 644         $PUSH   r31,`9*$SIZE_T`($sp)
 645         $PUSH   r0,`10*$SIZE_T+$LRSAVE`($sp)
 646
 647         mr      $A_flat,r3
 648         subi    r3,r3,8                 ; prepare for ldu
 649         subi    $out,r4,1               ; prepare for stbu
 650         mr      $len,r5
 651         mr      $bsz,r6
 652         b       .Loop_squeeze
 653
 654 .align  4
 655 .Loop_squeeze:
 656         ldu     r0,8(r3)
 657         ${UCMP}i $len,8
 658         blt     .Lsqueeze_tail
 659
 660         stbu    r0,1($out)
 661         srdi    r0,r0,8
 662         stbu    r0,1($out)
 663         srdi    r0,r0,8
 664         stbu    r0,1($out)
 665         srdi    r0,r0,8
 666         stbu    r0,1($out)
 667         srdi    r0,r0,8
 668         stbu    r0,1($out)
 669         srdi    r0,r0,8
 670         stbu    r0,1($out)
 671         srdi    r0,r0,8
 672         stbu    r0,1($out)
 673         srdi    r0,r0,8
 674         stbu    r0,1($out)
 675
 676         subic.  $len,$len,8
 677         beq     .Lsqueeze_done
 678
 679         subic.  r6,r6,8
 680         bgt     .Loop_squeeze
 681
 682         mr      r3,$A_flat
 683         bl      KeccakF1600
 684         subi    r3,$A_flat,8            ; prepare for ldu
 685         mr      r6,$bsz
 686         b       .Loop_squeeze
 687
 688 .align  4
 689 .Lsqueeze_tail:
 690         mtctr   $len
 691 .Loop_tail:
 692         stbu    r0,1($out)
 693         srdi    r0,r0,8
 694         bdnz    .Loop_tail
 695
 696 .Lsqueeze_done:
 697         $POP    r0,`10*$SIZE_T+$LRSAVE`($sp)
 698         $POP    r28,`6*$SIZE_T`($sp)
 699         $POP    r29,`7*$SIZE_T`($sp)
 700         $POP    r30,`8*$SIZE_T`($sp)
 701         $POP    r31,`9*$SIZE_T`($sp)
 702         mtlr    r0
 703         addi    $sp,$sp,`10*$SIZE_T`
 704         blr
 705         .long   0
 706         .byte   0,12,4,1,0x80,4,4,0
 707         .long   0
 708 .size   SHA3_squeeze,.-SHA3_squeeze
 709 ___
 710 }
 711
 712 # Ugly hack here, because PPC assembler syntax seem to vary too
 713 # much from platforms to platform...
 714 $code.=<<___;
 715 .align  6
 716 PICmeup:
 717         mflr    r0
 718         bcl     20,31,\$+4
 719         mflr    r12   ; vvvvvv "distance" between . and 1st data entry
 720         addi    r12,r12,`64-8`
 721         mtlr    r0
 722         blr
 723         .long   0
 724         .byte   0,12,0x14,0,0,0,0,0
 725         .space  `64-9*4`
 726 .type   iotas,\@object
 727 iotas:
 728         .quad   0x0000000000000001
 729         .quad   0x0000000000008082
 730         .quad   0x800000000000808a
 731         .quad   0x8000000080008000
 732         .quad   0x000000000000808b
 733         .quad   0x0000000080000001
 734         .quad   0x8000000080008081
 735         .quad   0x8000000000008009
 736         .quad   0x000000000000008a
 737         .quad   0x0000000000000088
 738         .quad   0x0000000080008009
 739         .quad   0x000000008000000a
 740         .quad   0x000000008000808b
 741         .quad   0x800000000000008b
 742         .quad   0x8000000000008089
 743         .quad   0x8000000000008003
 744         .quad   0x8000000000008002
 745         .quad   0x8000000000000080
 746         .quad   0x000000000000800a
 747         .quad   0x800000008000000a
 748         .quad   0x8000000080008081
 749         .quad   0x8000000000008080
 750         .quad   0x0000000080000001
 751         .quad   0x8000000080008008
 752 .size   iotas,.-iotas
 753 .asciz  "Keccak-1600 absorb and squeeze for PPC64, CRYPTOGAMS by <appro\@openssl.org>"
 754 ___
 755
 756 $code =~ s/\`([^\`]*)\`/eval $1/gem;
 757 print $code;
 758 close STDOUT;