OLD | NEW |
(Empty) | |
| 1 #!/usr/bin/env perl |
| 2 |
| 3 # ==================================================================== |
| 4 # Written by Andy Polyakov <appro@fy.chalmers.se> for the OpenSSL |
| 5 # project. The module is, however, dual licensed under OpenSSL and |
| 6 # CRYPTOGAMS licenses depending on where you obtain it. For further |
| 7 # details see http://www.openssl.org/~appro/cryptogams/. |
| 8 # ==================================================================== |
| 9 |
| 10 # SHA256/512 block procedures for s390x. |
| 11 |
| 12 # April 2007. |
| 13 # |
| 14 # sha256_block_data_order is reportedly >3 times faster than gcc 3.3 |
| 15 # generated code (must be a bug in compiler, as improvement is |
| 16 # "pathologically" high, in particular in comparison to other SHA |
| 17 # modules). But the real twist is that it detects if hardware support |
| 18 # for SHA256 is available and in such case utilizes it. Then the |
| 19 # performance can reach >6.5x of assembler one for larger chunks. |
| 20 # |
| 21 # sha512_block_data_order is ~70% faster than gcc 3.3 generated code. |
| 22 |
| 23 # January 2009. |
| 24 # |
| 25 # Add support for hardware SHA512 and reschedule instructions to |
| 26 # favour dual-issue z10 pipeline. Hardware SHA256/512 is ~4.7x faster |
| 27 # than software. |
| 28 |
| 29 $t0="%r0"; |
| 30 $t1="%r1"; |
| 31 $ctx="%r2"; $t2="%r2"; |
| 32 $inp="%r3"; |
| 33 $len="%r4"; # used as index in inner loop |
| 34 |
| 35 $A="%r5"; |
| 36 $B="%r6"; |
| 37 $C="%r7"; |
| 38 $D="%r8"; |
| 39 $E="%r9"; |
| 40 $F="%r10"; |
| 41 $G="%r11"; |
| 42 $H="%r12"; @V=($A,$B,$C,$D,$E,$F,$G,$H); |
| 43 $tbl="%r13"; |
| 44 $T1="%r14"; |
| 45 $sp="%r15"; |
| 46 |
| 47 $output=shift; |
| 48 open STDOUT,">$output"; |
| 49 |
| 50 if ($output =~ /512/) { |
| 51 $label="512"; |
| 52 $SZ=8; |
| 53 $LD="lg"; # load from memory |
| 54 $ST="stg"; # store to memory |
| 55 $ADD="alg"; # add with memory operand |
| 56 $ROT="rllg"; # rotate left |
| 57 $SHR="srlg"; # logical right shift [see even at the end] |
| 58 @Sigma0=(25,30,36); |
| 59 @Sigma1=(23,46,50); |
| 60 @sigma0=(56,63, 7); |
| 61 @sigma1=( 3,45, 6); |
| 62 $rounds=80; |
| 63 $kimdfunc=3; # 0 means unknown/unsupported/unimplemented/disabled |
| 64 } else { |
| 65 $label="256"; |
| 66 $SZ=4; |
| 67 $LD="llgf"; # load from memory |
| 68 $ST="st"; # store to memory |
| 69 $ADD="al"; # add with memory operand |
| 70 $ROT="rll"; # rotate left |
| 71 $SHR="srl"; # logical right shift |
| 72 @Sigma0=(10,19,30); |
| 73 @Sigma1=( 7,21,26); |
| 74 @sigma0=(14,25, 3); |
| 75 @sigma1=(13,15,10); |
| 76 $rounds=64; |
| 77 $kimdfunc=2; # magic function code for kimd instruction |
| 78 } |
| 79 $Func="sha${label}_block_data_order"; |
| 80 $Table="K${label}"; |
| 81 $frame=160+16*$SZ; |
| 82 |
| 83 sub BODY_00_15 { |
| 84 my ($i,$a,$b,$c,$d,$e,$f,$g,$h) = @_; |
| 85 |
| 86 $code.=<<___ if ($i<16); |
| 87 $LD $T1,`$i*$SZ`($inp) ### $i |
| 88 ___ |
| 89 $code.=<<___; |
| 90 $ROT $t0,$e,$Sigma1[0] |
| 91 $ROT $t1,$e,$Sigma1[1] |
| 92 lgr $t2,$f |
| 93 xgr $t0,$t1 |
| 94 $ROT $t1,$t1,`$Sigma1[2]-$Sigma1[1]` |
| 95 xgr $t2,$g |
| 96 $ST $T1,`160+$SZ*($i%16)`($sp) |
| 97 xgr $t0,$t1 # Sigma1(e) |
| 98 la $T1,0($T1,$h) # T1+=h |
| 99 ngr $t2,$e |
| 100 lgr $t1,$a |
| 101 algr $T1,$t0 # T1+=Sigma1(e) |
| 102 $ROT $h,$a,$Sigma0[0] |
| 103 xgr $t2,$g # Ch(e,f,g) |
| 104 $ADD $T1,`$i*$SZ`($len,$tbl) # T1+=K[i] |
| 105 $ROT $t0,$a,$Sigma0[1] |
| 106 algr $T1,$t2 # T1+=Ch(e,f,g) |
| 107 ogr $t1,$b |
| 108 xgr $h,$t0 |
| 109 lgr $t2,$a |
| 110 ngr $t1,$c |
| 111 $ROT $t0,$t0,`$Sigma0[2]-$Sigma0[1]` |
| 112 xgr $h,$t0 # h=Sigma0(a) |
| 113 ngr $t2,$b |
| 114 algr $h,$T1 # h+=T1 |
| 115 ogr $t2,$t1 # Maj(a,b,c) |
| 116 la $d,0($d,$T1) # d+=T1 |
| 117 algr $h,$t2 # h+=Maj(a,b,c) |
| 118 ___ |
| 119 } |
| 120 |
| 121 sub BODY_16_XX { |
| 122 my ($i,$a,$b,$c,$d,$e,$f,$g,$h) = @_; |
| 123 |
| 124 $code.=<<___; |
| 125 $LD $T1,`160+$SZ*(($i+1)%16)`($sp) ### $i |
| 126 $LD $t1,`160+$SZ*(($i+14)%16)`($sp) |
| 127 $ROT $t0,$T1,$sigma0[0] |
| 128 $SHR $T1,$sigma0[2] |
| 129 $ROT $t2,$t0,`$sigma0[1]-$sigma0[0]` |
| 130 xgr $T1,$t0 |
| 131 $ROT $t0,$t1,$sigma1[0] |
| 132 xgr $T1,$t2 # sigma0(X[i+1]) |
| 133 $SHR $t1,$sigma1[2] |
| 134 $ADD $T1,`160+$SZ*($i%16)`($sp) # +=X[i] |
| 135 xgr $t1,$t0 |
| 136 $ROT $t0,$t0,`$sigma1[1]-$sigma1[0]` |
| 137 $ADD $T1,`160+$SZ*(($i+9)%16)`($sp) # +=X[i+9] |
| 138 xgr $t1,$t0 # sigma1(X[i+14]) |
| 139 algr $T1,$t1 # +=sigma1(X[i+14]) |
| 140 ___ |
| 141 &BODY_00_15(@_); |
| 142 } |
| 143 |
| 144 $code.=<<___; |
| 145 .text |
| 146 .align 64 |
| 147 .type $Table,\@object |
| 148 $Table: |
| 149 ___ |
| 150 $code.=<<___ if ($SZ==4); |
| 151 .long 0x428a2f98,0x71374491,0xb5c0fbcf,0xe9b5dba5 |
| 152 .long 0x3956c25b,0x59f111f1,0x923f82a4,0xab1c5ed5 |
| 153 .long 0xd807aa98,0x12835b01,0x243185be,0x550c7dc3 |
| 154 .long 0x72be5d74,0x80deb1fe,0x9bdc06a7,0xc19bf174 |
| 155 .long 0xe49b69c1,0xefbe4786,0x0fc19dc6,0x240ca1cc |
| 156 .long 0x2de92c6f,0x4a7484aa,0x5cb0a9dc,0x76f988da |
| 157 .long 0x983e5152,0xa831c66d,0xb00327c8,0xbf597fc7 |
| 158 .long 0xc6e00bf3,0xd5a79147,0x06ca6351,0x14292967 |
| 159 .long 0x27b70a85,0x2e1b2138,0x4d2c6dfc,0x53380d13 |
| 160 .long 0x650a7354,0x766a0abb,0x81c2c92e,0x92722c85 |
| 161 .long 0xa2bfe8a1,0xa81a664b,0xc24b8b70,0xc76c51a3 |
| 162 .long 0xd192e819,0xd6990624,0xf40e3585,0x106aa070 |
| 163 .long 0x19a4c116,0x1e376c08,0x2748774c,0x34b0bcb5 |
| 164 .long 0x391c0cb3,0x4ed8aa4a,0x5b9cca4f,0x682e6ff3 |
| 165 .long 0x748f82ee,0x78a5636f,0x84c87814,0x8cc70208 |
| 166 .long 0x90befffa,0xa4506ceb,0xbef9a3f7,0xc67178f2 |
| 167 ___ |
| 168 $code.=<<___ if ($SZ==8); |
| 169 .quad 0x428a2f98d728ae22,0x7137449123ef65cd |
| 170 .quad 0xb5c0fbcfec4d3b2f,0xe9b5dba58189dbbc |
| 171 .quad 0x3956c25bf348b538,0x59f111f1b605d019 |
| 172 .quad 0x923f82a4af194f9b,0xab1c5ed5da6d8118 |
| 173 .quad 0xd807aa98a3030242,0x12835b0145706fbe |
| 174 .quad 0x243185be4ee4b28c,0x550c7dc3d5ffb4e2 |
| 175 .quad 0x72be5d74f27b896f,0x80deb1fe3b1696b1 |
| 176 .quad 0x9bdc06a725c71235,0xc19bf174cf692694 |
| 177 .quad 0xe49b69c19ef14ad2,0xefbe4786384f25e3 |
| 178 .quad 0x0fc19dc68b8cd5b5,0x240ca1cc77ac9c65 |
| 179 .quad 0x2de92c6f592b0275,0x4a7484aa6ea6e483 |
| 180 .quad 0x5cb0a9dcbd41fbd4,0x76f988da831153b5 |
| 181 .quad 0x983e5152ee66dfab,0xa831c66d2db43210 |
| 182 .quad 0xb00327c898fb213f,0xbf597fc7beef0ee4 |
| 183 .quad 0xc6e00bf33da88fc2,0xd5a79147930aa725 |
| 184 .quad 0x06ca6351e003826f,0x142929670a0e6e70 |
| 185 .quad 0x27b70a8546d22ffc,0x2e1b21385c26c926 |
| 186 .quad 0x4d2c6dfc5ac42aed,0x53380d139d95b3df |
| 187 .quad 0x650a73548baf63de,0x766a0abb3c77b2a8 |
| 188 .quad 0x81c2c92e47edaee6,0x92722c851482353b |
| 189 .quad 0xa2bfe8a14cf10364,0xa81a664bbc423001 |
| 190 .quad 0xc24b8b70d0f89791,0xc76c51a30654be30 |
| 191 .quad 0xd192e819d6ef5218,0xd69906245565a910 |
| 192 .quad 0xf40e35855771202a,0x106aa07032bbd1b8 |
| 193 .quad 0x19a4c116b8d2d0c8,0x1e376c085141ab53 |
| 194 .quad 0x2748774cdf8eeb99,0x34b0bcb5e19b48a8 |
| 195 .quad 0x391c0cb3c5c95a63,0x4ed8aa4ae3418acb |
| 196 .quad 0x5b9cca4f7763e373,0x682e6ff3d6b2b8a3 |
| 197 .quad 0x748f82ee5defb2fc,0x78a5636f43172f60 |
| 198 .quad 0x84c87814a1f0ab72,0x8cc702081a6439ec |
| 199 .quad 0x90befffa23631e28,0xa4506cebde82bde9 |
| 200 .quad 0xbef9a3f7b2c67915,0xc67178f2e372532b |
| 201 .quad 0xca273eceea26619c,0xd186b8c721c0c207 |
| 202 .quad 0xeada7dd6cde0eb1e,0xf57d4f7fee6ed178 |
| 203 .quad 0x06f067aa72176fba,0x0a637dc5a2c898a6 |
| 204 .quad 0x113f9804bef90dae,0x1b710b35131c471b |
| 205 .quad 0x28db77f523047d84,0x32caab7b40c72493 |
| 206 .quad 0x3c9ebe0a15c9bebc,0x431d67c49c100d4c |
| 207 .quad 0x4cc5d4becb3e42b6,0x597f299cfc657e2a |
| 208 .quad 0x5fcb6fab3ad6faec,0x6c44198c4a475817 |
| 209 ___ |
| 210 $code.=<<___; |
| 211 .size $Table,.-$Table |
| 212 .globl $Func |
| 213 .type $Func,\@function |
| 214 $Func: |
| 215 ___ |
| 216 $code.=<<___ if ($kimdfunc); |
| 217 larl %r1,OPENSSL_s390xcap_P |
| 218 lg %r0,0(%r1) |
| 219 tmhl %r0,0x4000 # check for message-security assist |
| 220 jz .Lsoftware |
| 221 lghi %r0,0 |
| 222 la %r1,16($sp) |
| 223 .long 0xb93e0002 # kimd %r0,%r2 |
| 224 lg %r0,16($sp) |
| 225 tmhh %r0,`0x8000>>$kimdfunc` |
| 226 jz .Lsoftware |
| 227 lghi %r0,$kimdfunc |
| 228 lgr %r1,$ctx |
| 229 lgr %r2,$inp |
| 230 sllg %r3,$len,`log(16*$SZ)/log(2)` |
| 231 .long 0xb93e0002 # kimd %r0,%r2 |
| 232 brc 1,.-4 # pay attention to "partial completion" |
| 233 br %r14 |
| 234 .align 16 |
| 235 .Lsoftware: |
| 236 ___ |
| 237 $code.=<<___; |
| 238 sllg $len,$len,`log(16*$SZ)/log(2)` |
| 239 lghi %r1,-$frame |
| 240 agr $len,$inp |
| 241 stmg $ctx,%r15,16($sp) |
| 242 lgr %r0,$sp |
| 243 la $sp,0(%r1,$sp) |
| 244 stg %r0,0($sp) |
| 245 |
| 246 larl $tbl,$Table |
| 247 $LD $A,`0*$SZ`($ctx) |
| 248 $LD $B,`1*$SZ`($ctx) |
| 249 $LD $C,`2*$SZ`($ctx) |
| 250 $LD $D,`3*$SZ`($ctx) |
| 251 $LD $E,`4*$SZ`($ctx) |
| 252 $LD $F,`5*$SZ`($ctx) |
| 253 $LD $G,`6*$SZ`($ctx) |
| 254 $LD $H,`7*$SZ`($ctx) |
| 255 |
| 256 .Lloop: |
| 257 lghi $len,0 |
| 258 ___ |
| 259 for ($i=0;$i<16;$i++) { &BODY_00_15($i,@V); unshift(@V,pop(@V)); } |
| 260 $code.=".Lrounds_16_xx:\n"; |
| 261 for (;$i<32;$i++) { &BODY_16_XX($i,@V); unshift(@V,pop(@V)); } |
| 262 $code.=<<___; |
| 263 aghi $len,`16*$SZ` |
| 264 lghi $t0,`($rounds-16)*$SZ` |
| 265 clgr $len,$t0 |
| 266 jne .Lrounds_16_xx |
| 267 |
| 268 lg $ctx,`$frame+16`($sp) |
| 269 la $inp,`16*$SZ`($inp) |
| 270 $ADD $A,`0*$SZ`($ctx) |
| 271 $ADD $B,`1*$SZ`($ctx) |
| 272 $ADD $C,`2*$SZ`($ctx) |
| 273 $ADD $D,`3*$SZ`($ctx) |
| 274 $ADD $E,`4*$SZ`($ctx) |
| 275 $ADD $F,`5*$SZ`($ctx) |
| 276 $ADD $G,`6*$SZ`($ctx) |
| 277 $ADD $H,`7*$SZ`($ctx) |
| 278 $ST $A,`0*$SZ`($ctx) |
| 279 $ST $B,`1*$SZ`($ctx) |
| 280 $ST $C,`2*$SZ`($ctx) |
| 281 $ST $D,`3*$SZ`($ctx) |
| 282 $ST $E,`4*$SZ`($ctx) |
| 283 $ST $F,`5*$SZ`($ctx) |
| 284 $ST $G,`6*$SZ`($ctx) |
| 285 $ST $H,`7*$SZ`($ctx) |
| 286 clg $inp,`$frame+32`($sp) |
| 287 jne .Lloop |
| 288 |
| 289 lmg %r6,%r15,`$frame+48`($sp) |
| 290 br %r14 |
| 291 .size $Func,.-$Func |
| 292 .string "SHA${label} block transform for s390x, CRYPTOGAMS by <appro\@openssl.or
g>" |
| 293 .comm OPENSSL_s390xcap_P,8,8 |
| 294 ___ |
| 295 |
| 296 $code =~ s/\`([^\`]*)\`/eval $1/gem; |
| 297 # unlike 32-bit shift 64-bit one takes three arguments |
| 298 $code =~ s/(srlg\s+)(%r[0-9]+),/$1$2,$2,/gm; |
| 299 |
| 300 print $code; |
| 301 close STDOUT; |
OLD | NEW |