projects / thirdparty / openssl.git / blame
| Commit | Line | Data |
|---|---|---|
| 8a97a330 AP |
1 | #!/usr/bin/env perl |
| 2 | # | |
| 3 | # ==================================================================== | |
| 4 | # Written by Andy Polyakov <appro@openssl.org> 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 | # January 2013 | |
| 11 | # | |
| 12 | # This is AESNI-CBC+SHA256 stitch implementation. The idea, as spelled | |
| 13 | # in http://download.intel.com/design/intarch/papers/323686.pdf, is | |
| 14 | # that since AESNI-CBC encrypt exhibit *very* low instruction-level | |
| 15 | # parallelism, interleaving it with another algorithm would allow to | |
| 16 | # utilize processor resources better and achieve better performance. | |
| 17 | # SHA256 instruction sequences(*) are taken from sha512-x86_64.pl and | |
| 18 | # AESNI code is weaved into it. As SHA256 dominates execution time, | |
| 19 | # stitch performance does not depend on AES key length. Below are | |
| 20 | # performance numbers in cycles per processed byte, less is better, | |
| 21 | # for standalone AESNI-CBC encrypt, standalone SHA256, and stitched | |
| 22 | # subroutine: | |
| 23 | # | |
| 24 | # AES-128/-192/-256+SHA256 this(**)gain | |
| 25 | # Sandy Bridge 5.05/6.05/7.05+11.6 13.0 +28%/36%/43% | |
| 26 | # Ivy Bridge 5.05/6.05/7.05+10.3 11.6 +32%/41%/50% | |
| 42b9a417 | 27 | # Haswell 4.43/5.29/6.19+7.80 8.79 +39%/49%/59% |
| 8a97a330 AP |
28 | # Bulldozer 5.77/6.89/8.00+13.7 13.7 +42%/50%/58% |
| 29 | # | |
| 30 | # (*) there are XOP, AVX1 and AVX2 code pathes, meaning that | |
| 31 | # Westmere is omitted from loop, this is because gain was not | |
| 32 | # estimated high enough to justify the effort; | |
| 33 | # (**) these are EVP-free results, results obtained with 'speed | |
| 34 | # -evp aes-256-cbc-hmac-sha256' will vary by percent or two; | |
| 35 | ||
| 36 | $flavour = shift; | |
| 37 | $output = shift; | |
| 38 | if ($flavour =~ /\./) { $output = $flavour; undef $flavour; } | |
| 39 | ||
| 40 | $win64=0; $win64=1 if ($flavour =~ /[nm]asm|mingw64/ || $output =~ /\.asm$/); | |
| 41 | ||
| 42 | $0 =~ m/(.*[\/\\])[^\/\\]+$/; $dir=$1; | |
| 43 | ( $xlate="${dir}x86_64-xlate.pl" and -f $xlate ) or | |
| 44 | ( $xlate="${dir}../../perlasm/x86_64-xlate.pl" and -f $xlate) or | |
| 45 | die "can't locate x86_64-xlate.pl"; | |
| 46 | ||
| 47 | if (`$ENV{CC} -Wa,-v -c -o /dev/null -x assembler /dev/null 2>&1` | |
| 48 | =~ /GNU assembler version ([2-9]\.[0-9]+)/) { | |
| 49 | $avx = ($1>=2.19) + ($1>=2.22); | |
| 50 | } | |
| 51 | ||
| 52 | if (!$avx && $win64 && ($flavour =~ /nasm/ || $ENV{ASM} =~ /nasm/) && | |
| 53 | `nasm -v 2>&1` =~ /NASM version ([2-9]\.[0-9]+)/) { | |
| 54 | $avx = ($1>=2.09) + ($1>=2.10); | |
| 55 | } | |
| 56 | ||
| 57 | if (!$avx && $win64 && ($flavour =~ /masm/ || $ENV{ASM} =~ /ml64/) && | |
| 58 | `ml64 2>&1` =~ /Version ([0-9]+)\./) { | |
| 59 | $avx = ($1>=10) + ($1>=11); | |
| 60 | } | |
| 61 | ||
| 977f32e8 AP |
62 | $shaext=1; ### set to zero if compiling for 1.0.1 |
| 63 | $avx=1 if (!$shaext && $avx); | |
| 64 | ||
| 8a97a330 AP |
65 | open OUT,"| \"$^X\" $xlate $flavour $output"; |
| 66 | *STDOUT=*OUT; | |
| 67 | ||
| 68 | $func="aesni_cbc_sha256_enc"; | |
| 69 | $TABLE="K256"; | |
| 70 | $SZ=4; | |
| 71 | @ROT=($A,$B,$C,$D,$E,$F,$G,$H)=("%eax","%ebx","%ecx","%edx", | |
| 72 | "%r8d","%r9d","%r10d","%r11d"); | |
| 73 | ($T1,$a0,$a1,$a2,$a3)=("%r12d","%r13d","%r14d","%r15d","%esi"); | |
| 74 | @Sigma0=( 2,13,22); | |
| 75 | @Sigma1=( 6,11,25); | |
| 76 | @sigma0=( 7,18, 3); | |
| 77 | @sigma1=(17,19,10); | |
| 78 | $rounds=64; | |
| 79 | ||
| 80 | ######################################################################## | |
| 81 | # void aesni_cbc_sha256_enc(const void *inp, | |
| 82 | # void *out, | |
| 83 | # size_t length, | |
| 84 | # const AES_KEY *key, | |
| 85 | # unsigned char *iv, | |
| 86 | # SHA256_CTX *ctx, | |
| 87 | # const void *in0); | |
| 88 | ($inp, $out, $len, $key, $ivp, $ctx, $in0) = | |
| 89 | ("%rdi","%rsi","%rdx","%rcx","%r8","%r9","%r10"); | |
| 90 | ||
| 91 | $Tbl="%rbp"; | |
| 92 | ||
| 93 | $_inp="16*$SZ+0*8(%rsp)"; | |
| 94 | $_out="16*$SZ+1*8(%rsp)"; | |
| 95 | $_end="16*$SZ+2*8(%rsp)"; | |
| 96 | $_key="16*$SZ+3*8(%rsp)"; | |
| 97 | $_ivp="16*$SZ+4*8(%rsp)"; | |
| 98 | $_ctx="16*$SZ+5*8(%rsp)"; | |
| 99 | $_in0="16*$SZ+6*8(%rsp)"; | |
| 100 | $_rsp="16*$SZ+7*8(%rsp)"; | |
| 101 | $framesz=16*$SZ+8*8; | |
| 102 | ||
| 103 | $code=<<___; | |
| 104 | .text | |
| 105 | ||
| 106 | .extern OPENSSL_ia32cap_P | |
| 107 | .globl $func | |
| 108 | .type $func,\@abi-omnipotent | |
| 109 | .align 16 | |
| 110 | $func: | |
| 111 | ___ | |
| 112 | $code.=<<___ if ($avx); | |
| 113 | lea OPENSSL_ia32cap_P(%rip),%r11 | |
| 114 | mov \$1,%eax | |
| 115 | cmp \$0,`$win64?"%rcx":"%rdi"` | |
| 116 | je .Lprobe | |
| 117 | mov 0(%r11),%eax | |
| 619b9466 | 118 | mov 4(%r11),%r10 |
| 977f32e8 AP |
119 | ___ |
| 120 | $code.=<<___ if ($shaext); | |
| 619b9466 AP |
121 | bt \$61,%r10 # check for SHA |
| 122 | jc ${func}_shaext | |
| 977f32e8 AP |
123 | ___ |
| 124 | $code.=<<___; | |
| 619b9466 AP |
125 | mov %r10,%r11 |
| 126 | shr \$32,%r11 | |
| 8a97a330 AP |
127 | |
| 128 | test \$`1<<11`,%r10d # check for XOP | |
| 129 | jnz ${func}_xop | |
| 130 | ___ | |
| 131 | $code.=<<___ if ($avx>1); | |
| 132 | and \$`1<<8|1<<5|1<<3`,%r11d # check for BMI2+AVX2+BMI1 | |
| 133 | cmp \$`1<<8|1<<5|1<<3`,%r11d | |
| 134 | je ${func}_avx2 | |
| 135 | ___ | |
| 136 | $code.=<<___ if ($avx); | |
| 137 | and \$`1<<30`,%eax # mask "Intel CPU" bit | |
| 138 | and \$`1<<28|1<<9`,%r10d # mask AVX+SSSE3 bits | |
| 139 | or %eax,%r10d | |
| 140 | cmp \$`1<<28|1<<9|1<<30`,%r10d | |
| 141 | je ${func}_avx | |
| 142 | ud2 | |
| 143 | ___ | |
| 144 | $code.=<<___; | |
| 145 | xor %eax,%eax | |
| 146 | cmp \$0,`$win64?"%rcx":"%rdi"` | |
| 147 | je .Lprobe | |
| 148 | ud2 | |
| 149 | .Lprobe: | |
| 150 | ret | |
| 151 | .size $func,.-$func | |
| 152 | ||
| 153 | .align 64 | |
| 154 | .type $TABLE,\@object | |
| 155 | $TABLE: | |
| 156 | .long 0x428a2f98,0x71374491,0xb5c0fbcf,0xe9b5dba5 | |
| 157 | .long 0x428a2f98,0x71374491,0xb5c0fbcf,0xe9b5dba5 | |
| 158 | .long 0x3956c25b,0x59f111f1,0x923f82a4,0xab1c5ed5 | |
| 159 | .long 0x3956c25b,0x59f111f1,0x923f82a4,0xab1c5ed5 | |
| 160 | .long 0xd807aa98,0x12835b01,0x243185be,0x550c7dc3 | |
| 161 | .long 0xd807aa98,0x12835b01,0x243185be,0x550c7dc3 | |
| 162 | .long 0x72be5d74,0x80deb1fe,0x9bdc06a7,0xc19bf174 | |
| 163 | .long 0x72be5d74,0x80deb1fe,0x9bdc06a7,0xc19bf174 | |
| 164 | .long 0xe49b69c1,0xefbe4786,0x0fc19dc6,0x240ca1cc | |
| 165 | .long 0xe49b69c1,0xefbe4786,0x0fc19dc6,0x240ca1cc | |
| 166 | .long 0x2de92c6f,0x4a7484aa,0x5cb0a9dc,0x76f988da | |
| 167 | .long 0x2de92c6f,0x4a7484aa,0x5cb0a9dc,0x76f988da | |
| 168 | .long 0x983e5152,0xa831c66d,0xb00327c8,0xbf597fc7 | |
| 169 | .long 0x983e5152,0xa831c66d,0xb00327c8,0xbf597fc7 | |
| 170 | .long 0xc6e00bf3,0xd5a79147,0x06ca6351,0x14292967 | |
| 171 | .long 0xc6e00bf3,0xd5a79147,0x06ca6351,0x14292967 | |
| 172 | .long 0x27b70a85,0x2e1b2138,0x4d2c6dfc,0x53380d13 | |
| 173 | .long 0x27b70a85,0x2e1b2138,0x4d2c6dfc,0x53380d13 | |
| 174 | .long 0x650a7354,0x766a0abb,0x81c2c92e,0x92722c85 | |
| 175 | .long 0x650a7354,0x766a0abb,0x81c2c92e,0x92722c85 | |
| 176 | .long 0xa2bfe8a1,0xa81a664b,0xc24b8b70,0xc76c51a3 | |
| 177 | .long 0xa2bfe8a1,0xa81a664b,0xc24b8b70,0xc76c51a3 | |
| 178 | .long 0xd192e819,0xd6990624,0xf40e3585,0x106aa070 | |
| 179 | .long 0xd192e819,0xd6990624,0xf40e3585,0x106aa070 | |
| 180 | .long 0x19a4c116,0x1e376c08,0x2748774c,0x34b0bcb5 | |
| 181 | .long 0x19a4c116,0x1e376c08,0x2748774c,0x34b0bcb5 | |
| 182 | .long 0x391c0cb3,0x4ed8aa4a,0x5b9cca4f,0x682e6ff3 | |
| 183 | .long 0x391c0cb3,0x4ed8aa4a,0x5b9cca4f,0x682e6ff3 | |
| 184 | .long 0x748f82ee,0x78a5636f,0x84c87814,0x8cc70208 | |
| 185 | .long 0x748f82ee,0x78a5636f,0x84c87814,0x8cc70208 | |
| 186 | .long 0x90befffa,0xa4506ceb,0xbef9a3f7,0xc67178f2 | |
| 187 | .long 0x90befffa,0xa4506ceb,0xbef9a3f7,0xc67178f2 | |
| 188 | ||
| 189 | .long 0x00010203,0x04050607,0x08090a0b,0x0c0d0e0f | |
| 190 | .long 0x00010203,0x04050607,0x08090a0b,0x0c0d0e0f | |
| 191 | .long 0,0,0,0, 0,0,0,0, -1,-1,-1,-1 | |
| 192 | .long 0,0,0,0, 0,0,0,0 | |
| 193 | .asciz "AESNI-CBC+SHA256 stitch for x86_64, CRYPTOGAMS by <appro\@openssl.org>" | |
| 194 | .align 64 | |
| 195 | ___ | |
| 196 | ||
| 197 | ###################################################################### | |
| 198 | # SIMD code paths | |
| 199 | # | |
| 200 | {{{ | |
| 201 | ($iv,$inout,$roundkey,$temp, | |
| 202 | $mask10,$mask12,$mask14,$offload)=map("%xmm$_",(8..15)); | |
| 203 | ||
| 204 | $aesni_cbc_idx=0; | |
| 205 | @aesni_cbc_block = ( | |
| 206 | ## &vmovdqu ($roundkey,"0x00-0x80($inp)");' | |
| 207 | ## &vmovdqu ($inout,($inp)); | |
| 208 | ## &mov ($_inp,$inp); | |
| 209 | ||
| 210 | '&vpxor ($inout,$inout,$roundkey);'. | |
| 211 | ' &vmovdqu ($roundkey,"0x10-0x80($inp)");', | |
| 212 | ||
| 213 | '&vpxor ($inout,$inout,$iv);', | |
| 214 | ||
| 215 | '&vaesenc ($inout,$inout,$roundkey);'. | |
| 216 | ' &vmovdqu ($roundkey,"0x20-0x80($inp)");', | |
| 217 | ||
| 218 | '&vaesenc ($inout,$inout,$roundkey);'. | |
| 219 | ' &vmovdqu ($roundkey,"0x30-0x80($inp)");', | |
| 220 | ||
| 221 | '&vaesenc ($inout,$inout,$roundkey);'. | |
| 222 | ' &vmovdqu ($roundkey,"0x40-0x80($inp)");', | |
| 223 | ||
| 224 | '&vaesenc ($inout,$inout,$roundkey);'. | |
| 225 | ' &vmovdqu ($roundkey,"0x50-0x80($inp)");', | |
| 226 | ||
| 227 | '&vaesenc ($inout,$inout,$roundkey);'. | |
| 228 | ' &vmovdqu ($roundkey,"0x60-0x80($inp)");', | |
| 229 | ||
| 230 | '&vaesenc ($inout,$inout,$roundkey);'. | |
| 231 | ' &vmovdqu ($roundkey,"0x70-0x80($inp)");', | |
| 232 | ||
| 233 | '&vaesenc ($inout,$inout,$roundkey);'. | |
| 234 | ' &vmovdqu ($roundkey,"0x80-0x80($inp)");', | |
| 235 | ||
| 236 | '&vaesenc ($inout,$inout,$roundkey);'. | |
| 237 | ' &vmovdqu ($roundkey,"0x90-0x80($inp)");', | |
| 238 | ||
| 239 | '&vaesenc ($inout,$inout,$roundkey);'. | |
| 240 | ' &vmovdqu ($roundkey,"0xa0-0x80($inp)");', | |
| 241 | ||
| 242 | '&vaesenclast ($temp,$inout,$roundkey);'. | |
| 243 | ' &vaesenc ($inout,$inout,$roundkey);'. | |
| 244 | ' &vmovdqu ($roundkey,"0xb0-0x80($inp)");', | |
| 245 | ||
| 246 | '&vpand ($iv,$temp,$mask10);'. | |
| 247 | ' &vaesenc ($inout,$inout,$roundkey);'. | |
| 248 | ' &vmovdqu ($roundkey,"0xc0-0x80($inp)");', | |
| 249 | ||
| 250 | '&vaesenclast ($temp,$inout,$roundkey);'. | |
| 251 | ' &vaesenc ($inout,$inout,$roundkey);'. | |
| 252 | ' &vmovdqu ($roundkey,"0xd0-0x80($inp)");', | |
| 253 | ||
| 254 | '&vpand ($temp,$temp,$mask12);'. | |
| 255 | ' &vaesenc ($inout,$inout,$roundkey);'. | |
| 256 | '&vmovdqu ($roundkey,"0xe0-0x80($inp)");', | |
| 257 | ||
| 258 | '&vpor ($iv,$iv,$temp);'. | |
| 259 | ' &vaesenclast ($temp,$inout,$roundkey);'. | |
| 260 | ' &vmovdqu ($roundkey,"0x00-0x80($inp)");' | |
| 261 | ||
| 262 | ## &mov ($inp,$_inp); | |
| 263 | ## &mov ($out,$_out); | |
| 264 | ## &vpand ($temp,$temp,$mask14); | |
| 265 | ## &vpor ($iv,$iv,$temp); | |
| 266 | ## &vmovdqu ($iv,($out,$inp); | |
| 267 | ## &lea (inp,16($inp)); | |
| 268 | ); | |
| 269 | ||
| 270 | my $a4=$T1; | |
| 271 | my ($a,$b,$c,$d,$e,$f,$g,$h); | |
| 272 | ||
| 273 | sub AUTOLOAD() # thunk [simplified] 32-bit style perlasm | |
| 274 | { my $opcode = $AUTOLOAD; $opcode =~ s/.*:://; | |
| 275 | my $arg = pop; | |
| 276 | $arg = "\$$arg" if ($arg*1 eq $arg); | |
| 277 | $code .= "\t$opcode\t".join(',',$arg,reverse @_)."\n"; | |
| 278 | } | |
| 279 | ||
| 280 | sub body_00_15 () { | |
| 281 | ( | |
| 282 | '($a,$b,$c,$d,$e,$f,$g,$h)=@ROT;'. | |
| 283 | ||
| 284 | '&ror ($a0,$Sigma1[2]-$Sigma1[1])', | |
| 285 | '&mov ($a,$a1)', | |
| 286 | '&mov ($a4,$f)', | |
| 287 | ||
| 288 | '&xor ($a0,$e)', | |
| 289 | '&ror ($a1,$Sigma0[2]-$Sigma0[1])', | |
| 290 | '&xor ($a4,$g)', # f^g | |
| 291 | ||
| 292 | '&ror ($a0,$Sigma1[1]-$Sigma1[0])', | |
| 293 | '&xor ($a1,$a)', | |
| 294 | '&and ($a4,$e)', # (f^g)&e | |
| 295 | ||
| 296 | @aesni_cbc_block[$aesni_cbc_idx++]. | |
| 297 | '&xor ($a0,$e)', | |
| 298 | '&add ($h,$SZ*($i&15)."(%rsp)")', # h+=X[i]+K[i] | |
| 299 | '&mov ($a2,$a)', | |
| 300 | ||
| 301 | '&ror ($a1,$Sigma0[1]-$Sigma0[0])', | |
| 302 | '&xor ($a4,$g)', # Ch(e,f,g)=((f^g)&e)^g | |
| 303 | '&xor ($a2,$b)', # a^b, b^c in next round | |
| 304 | ||
| 305 | '&ror ($a0,$Sigma1[0])', # Sigma1(e) | |
| 306 | '&add ($h,$a4)', # h+=Ch(e,f,g) | |
| 307 | '&and ($a3,$a2)', # (b^c)&(a^b) | |
| 308 | ||
| 309 | '&xor ($a1,$a)', | |
| 310 | '&add ($h,$a0)', # h+=Sigma1(e) | |
| 311 | '&xor ($a3,$b)', # Maj(a,b,c)=Ch(a^b,c,b) | |
| 312 | ||
| 313 | '&add ($d,$h)', # d+=h | |
| 314 | '&ror ($a1,$Sigma0[0])', # Sigma0(a) | |
| 315 | '&add ($h,$a3)', # h+=Maj(a,b,c) | |
| 316 | ||
| 317 | '&mov ($a0,$d)', | |
| 318 | '&add ($a1,$h);'. # h+=Sigma0(a) | |
| 319 | '($a2,$a3) = ($a3,$a2); unshift(@ROT,pop(@ROT)); $i++;' | |
| 320 | ); | |
| 321 | } | |
| 322 | ||
| 323 | if ($avx) {{ | |
| 324 | ###################################################################### | |
| 325 | # XOP code path | |
| 326 | # | |
| 327 | $code.=<<___; | |
| 328 | .type ${func}_xop,\@function,6 | |
| 329 | .align 64 | |
| 330 | ${func}_xop: | |
| 331 | .Lxop_shortcut: | |
| 332 | mov `($win64?56:8)`(%rsp),$in0 # load 7th parameter | |
| 333 | push %rbx | |
| 334 | push %rbp | |
| 335 | push %r12 | |
| 336 | push %r13 | |
| 337 | push %r14 | |
| 338 | push %r15 | |
| 339 | mov %rsp,%r11 # copy %rsp | |
| 340 | sub \$`$framesz+$win64*16*10`,%rsp | |
| 341 | and \$-64,%rsp # align stack frame | |
| 342 | ||
| 343 | shl \$6,$len | |
| 344 | sub $inp,$out # re-bias | |
| 345 | sub $inp,$in0 | |
| 346 | add $inp,$len # end of input | |
| 347 | ||
| 348 | #mov $inp,$_inp # saved later | |
| 349 | mov $out,$_out | |
| 350 | mov $len,$_end | |
| 351 | #mov $key,$_key # remains resident in $inp register | |
| 352 | mov $ivp,$_ivp | |
| 353 | mov $ctx,$_ctx | |
| 354 | mov $in0,$_in0 | |
| 355 | mov %r11,$_rsp | |
| 356 | ___ | |
| 357 | $code.=<<___ if ($win64); | |
| 358 | movaps %xmm6,`$framesz+16*0`(%rsp) | |
| 359 | movaps %xmm7,`$framesz+16*1`(%rsp) | |
| 360 | movaps %xmm8,`$framesz+16*2`(%rsp) | |
| 361 | movaps %xmm9,`$framesz+16*3`(%rsp) | |
| 362 | movaps %xmm10,`$framesz+16*4`(%rsp) | |
| 363 | movaps %xmm11,`$framesz+16*5`(%rsp) | |
| 364 | movaps %xmm12,`$framesz+16*6`(%rsp) | |
| 365 | movaps %xmm13,`$framesz+16*7`(%rsp) | |
| 366 | movaps %xmm14,`$framesz+16*8`(%rsp) | |
| 367 | movaps %xmm15,`$framesz+16*9`(%rsp) | |
| 368 | ___ | |
| 369 | $code.=<<___; | |
| 370 | .Lprologue_xop: | |
| 371 | vzeroall | |
| 372 | ||
| 373 | mov $inp,%r12 # borrow $a4 | |
| 374 | lea 0x80($key),$inp # size optimization, reassign | |
| 375 | lea $TABLE+`$SZ*2*$rounds+32`(%rip),%r13 # borrow $a0 | |
| 376 | mov 0xf0-0x80($inp),%r14d # rounds, borrow $a1 | |
| 377 | mov $ctx,%r15 # borrow $a2 | |
| 378 | mov $in0,%rsi # borrow $a3 | |
| 379 | vmovdqu ($ivp),$iv # load IV | |
| 380 | sub \$9,%r14 | |
| 381 | ||
| 382 | mov $SZ*0(%r15),$A | |
| 383 | mov $SZ*1(%r15),$B | |
| 384 | mov $SZ*2(%r15),$C | |
| 385 | mov $SZ*3(%r15),$D | |
| 386 | mov $SZ*4(%r15),$E | |
| 387 | mov $SZ*5(%r15),$F | |
| 388 | mov $SZ*6(%r15),$G | |
| 389 | mov $SZ*7(%r15),$H | |
| 390 | ||
| 391 | vmovdqa 0x00(%r13,%r14,8),$mask14 | |
| 392 | vmovdqa 0x10(%r13,%r14,8),$mask12 | |
| 393 | vmovdqa 0x20(%r13,%r14,8),$mask10 | |
| 394 | vmovdqu 0x00-0x80($inp),$roundkey | |
| 395 | jmp .Lloop_xop | |
| 396 | ___ | |
| 397 | if ($SZ==4) { # SHA256 | |
| 398 | my @X = map("%xmm$_",(0..3)); | |
| 399 | my ($t0,$t1,$t2,$t3) = map("%xmm$_",(4..7)); | |
| 400 | ||
| 401 | $code.=<<___; | |
| 402 | .align 16 | |
| 403 | .Lloop_xop: | |
| 404 | vmovdqa $TABLE+`$SZ*2*$rounds`(%rip),$t3 | |
| 405 | vmovdqu 0x00(%rsi,%r12),@X[0] | |
| 406 | vmovdqu 0x10(%rsi,%r12),@X[1] | |
| 407 | vmovdqu 0x20(%rsi,%r12),@X[2] | |
| 408 | vmovdqu 0x30(%rsi,%r12),@X[3] | |
| 409 | vpshufb $t3,@X[0],@X[0] | |
| 410 | lea $TABLE(%rip),$Tbl | |
| 411 | vpshufb $t3,@X[1],@X[1] | |
| 412 | vpshufb $t3,@X[2],@X[2] | |
| 413 | vpaddd 0x00($Tbl),@X[0],$t0 | |
| 414 | vpshufb $t3,@X[3],@X[3] | |
| 415 | vpaddd 0x20($Tbl),@X[1],$t1 | |
| 416 | vpaddd 0x40($Tbl),@X[2],$t2 | |
| 417 | vpaddd 0x60($Tbl),@X[3],$t3 | |
| 418 | vmovdqa $t0,0x00(%rsp) | |
| 419 | mov $A,$a1 | |
| 420 | vmovdqa $t1,0x10(%rsp) | |
| 421 | mov $B,$a3 | |
| 422 | vmovdqa $t2,0x20(%rsp) | |
| 423 | xor $C,$a3 # magic | |
| 424 | vmovdqa $t3,0x30(%rsp) | |
| 425 | mov $E,$a0 | |
| 426 | jmp .Lxop_00_47 | |
| 427 | ||
| 428 | .align 16 | |
| 429 | .Lxop_00_47: | |
| 430 | sub \$-16*2*$SZ,$Tbl # size optimization | |
| 431 | vmovdqu (%r12),$inout # $a4 | |
| 432 | mov %r12,$_inp # $a4 | |
| 433 | ___ | |
| 434 | sub XOP_256_00_47 () { | |
| 435 | my $j = shift; | |
| 436 | my $body = shift; | |
| 437 | my @X = @_; | |
| 438 | my @insns = (&$body,&$body,&$body,&$body); # 104 instructions | |
| 439 | ||
| 440 | &vpalignr ($t0,@X[1],@X[0],$SZ); # X[1..4] | |
| 441 | eval(shift(@insns)); | |
| 442 | eval(shift(@insns)); | |
| 443 | &vpalignr ($t3,@X[3],@X[2],$SZ); # X[9..12] | |
| 444 | eval(shift(@insns)); | |
| 445 | eval(shift(@insns)); | |
| 446 | &vprotd ($t1,$t0,8*$SZ-$sigma0[1]); | |
| 447 | eval(shift(@insns)); | |
| 448 | eval(shift(@insns)); | |
| 449 | &vpsrld ($t0,$t0,$sigma0[2]); | |
| 450 | eval(shift(@insns)); | |
| 451 | eval(shift(@insns)); | |
| 452 | &vpaddd (@X[0],@X[0],$t3); # X[0..3] += X[9..12] | |
| 453 | eval(shift(@insns)); | |
| 454 | eval(shift(@insns)); | |
| 455 | eval(shift(@insns)); | |
| 456 | eval(shift(@insns)); | |
| 457 | &vprotd ($t2,$t1,$sigma0[1]-$sigma0[0]); | |
| 458 | eval(shift(@insns)); | |
| 459 | eval(shift(@insns)); | |
| 460 | &vpxor ($t0,$t0,$t1); | |
| 461 | eval(shift(@insns)); | |
| 462 | eval(shift(@insns)); | |
| 463 | eval(shift(@insns)); | |
| 464 | eval(shift(@insns)); | |
| 465 | &vprotd ($t3,@X[3],8*$SZ-$sigma1[1]); | |
| 466 | eval(shift(@insns)); | |
| 467 | eval(shift(@insns)); | |
| 468 | &vpxor ($t0,$t0,$t2); # sigma0(X[1..4]) | |
| 469 | eval(shift(@insns)); | |
| 470 | eval(shift(@insns)); | |
| 471 | &vpsrld ($t2,@X[3],$sigma1[2]); | |
| 472 | eval(shift(@insns)); | |
| 473 | eval(shift(@insns)); | |
| 474 | &vpaddd (@X[0],@X[0],$t0); # X[0..3] += sigma0(X[1..4]) | |
| 475 | eval(shift(@insns)); | |
| 476 | eval(shift(@insns)); | |
| 477 | &vprotd ($t1,$t3,$sigma1[1]-$sigma1[0]); | |
| 478 | eval(shift(@insns)); | |
| 479 | eval(shift(@insns)); | |
| 480 | &vpxor ($t3,$t3,$t2); | |
| 481 | eval(shift(@insns)); | |
| 482 | eval(shift(@insns)); | |
| 483 | eval(shift(@insns)); | |
| 484 | eval(shift(@insns)); | |
| 485 | &vpxor ($t3,$t3,$t1); # sigma1(X[14..15]) | |
| 486 | eval(shift(@insns)); | |
| 487 | eval(shift(@insns)); | |
| 488 | eval(shift(@insns)); | |
| 489 | eval(shift(@insns)); | |
| 490 | &vpsrldq ($t3,$t3,8); | |
| 491 | eval(shift(@insns)); | |
| 492 | eval(shift(@insns)); | |
| 493 | eval(shift(@insns)); | |
| 494 | eval(shift(@insns)); | |
| 495 | &vpaddd (@X[0],@X[0],$t3); # X[0..1] += sigma1(X[14..15]) | |
| 496 | eval(shift(@insns)); | |
| 497 | eval(shift(@insns)); | |
| 498 | eval(shift(@insns)); | |
| 499 | eval(shift(@insns)); | |
| 500 | &vprotd ($t3,@X[0],8*$SZ-$sigma1[1]); | |
| 501 | eval(shift(@insns)); | |
| 502 | eval(shift(@insns)); | |
| 503 | &vpsrld ($t2,@X[0],$sigma1[2]); | |
| 504 | eval(shift(@insns)); | |
| 505 | eval(shift(@insns)); | |
| 506 | &vprotd ($t1,$t3,$sigma1[1]-$sigma1[0]); | |
| 507 | eval(shift(@insns)); | |
| 508 | eval(shift(@insns)); | |
| 509 | &vpxor ($t3,$t3,$t2); | |
| 510 | eval(shift(@insns)); | |
| 511 | eval(shift(@insns)); | |
| 512 | eval(shift(@insns)); | |
| 513 | eval(shift(@insns)); | |
| 514 | &vpxor ($t3,$t3,$t1); # sigma1(X[16..17]) | |
| 515 | eval(shift(@insns)); | |
| 516 | eval(shift(@insns)); | |
| 517 | eval(shift(@insns)); | |
| 518 | eval(shift(@insns)); | |
| 519 | &vpslldq ($t3,$t3,8); # 22 instructions | |
| 520 | eval(shift(@insns)); | |
| 521 | eval(shift(@insns)); | |
| 522 | eval(shift(@insns)); | |
| 523 | eval(shift(@insns)); | |
| 524 | &vpaddd (@X[0],@X[0],$t3); # X[2..3] += sigma1(X[16..17]) | |
| 525 | eval(shift(@insns)); | |
| 526 | eval(shift(@insns)); | |
| 527 | eval(shift(@insns)); | |
| 528 | eval(shift(@insns)); | |
| 529 | &vpaddd ($t2,@X[0],16*2*$j."($Tbl)"); | |
| 530 | foreach (@insns) { eval; } # remaining instructions | |
| 531 | &vmovdqa (16*$j."(%rsp)",$t2); | |
| 532 | } | |
| 533 | ||
| 534 | $aesni_cbc_idx=0; | |
| 535 | for ($i=0,$j=0; $j<4; $j++) { | |
| 536 | &XOP_256_00_47($j,\&body_00_15,@X); | |
| 537 | push(@X,shift(@X)); # rotate(@X) | |
| 538 | } | |
| 539 | &mov ("%r12",$_inp); # borrow $a4 | |
| 540 | &vpand ($temp,$temp,$mask14); | |
| 541 | &mov ("%r15",$_out); # borrow $a2 | |
| 542 | &vpor ($iv,$iv,$temp); | |
| 543 | &vmovdqu ("(%r15,%r12)",$iv); # write output | |
| 544 | &lea ("%r12","16(%r12)"); # inp++ | |
| 545 | ||
| 546 | &cmpb ($SZ-1+16*2*$SZ."($Tbl)",0); | |
| 547 | &jne (".Lxop_00_47"); | |
| 548 | ||
| 549 | &vmovdqu ($inout,"(%r12)"); | |
| 550 | &mov ($_inp,"%r12"); | |
| 551 | ||
| 552 | $aesni_cbc_idx=0; | |
| 553 | for ($i=0; $i<16; ) { | |
| 554 | foreach(body_00_15()) { eval; } | |
| 555 | } | |
| 556 | } | |
| 557 | $code.=<<___; | |
| 558 | mov $_inp,%r12 # borrow $a4 | |
| 559 | mov $_out,%r13 # borrow $a0 | |
| 560 | mov $_ctx,%r15 # borrow $a2 | |
| 561 | mov $_in0,%rsi # borrow $a3 | |
| 562 | ||
| 563 | vpand $mask14,$temp,$temp | |
| 564 | mov $a1,$A | |
| 565 | vpor $temp,$iv,$iv | |
| 566 | vmovdqu $iv,(%r13,%r12) # write output | |
| 567 | lea 16(%r12),%r12 # inp++ | |
| 568 | ||
| 569 | add $SZ*0(%r15),$A | |
| 570 | add $SZ*1(%r15),$B | |
| 571 | add $SZ*2(%r15),$C | |
| 572 | add $SZ*3(%r15),$D | |
| 573 | add $SZ*4(%r15),$E | |
| 574 | add $SZ*5(%r15),$F | |
| 575 | add $SZ*6(%r15),$G | |
| 576 | add $SZ*7(%r15),$H | |
| 577 | ||
| 578 | cmp $_end,%r12 | |
| 579 | ||
| 580 | mov $A,$SZ*0(%r15) | |
| 581 | mov $B,$SZ*1(%r15) | |
| 582 | mov $C,$SZ*2(%r15) | |
| 583 | mov $D,$SZ*3(%r15) | |
| 584 | mov $E,$SZ*4(%r15) | |
| 585 | mov $F,$SZ*5(%r15) | |
| 586 | mov $G,$SZ*6(%r15) | |
| 587 | mov $H,$SZ*7(%r15) | |
| 588 | ||
| 589 | jb .Lloop_xop | |
| 590 | ||
| 591 | mov $_ivp,$ivp | |
| 592 | mov $_rsp,%rsi | |
| 593 | vmovdqu $iv,($ivp) # output IV | |
| 594 | vzeroall | |
| 595 | ___ | |
| 596 | $code.=<<___ if ($win64); | |
| 597 | movaps `$framesz+16*0`(%rsp),%xmm6 | |
| 598 | movaps `$framesz+16*1`(%rsp),%xmm7 | |
| 599 | movaps `$framesz+16*2`(%rsp),%xmm8 | |
| 600 | movaps `$framesz+16*3`(%rsp),%xmm9 | |
| 601 | movaps `$framesz+16*4`(%rsp),%xmm10 | |
| 602 | movaps `$framesz+16*5`(%rsp),%xmm11 | |
| 603 | movaps `$framesz+16*6`(%rsp),%xmm12 | |
| 604 | movaps `$framesz+16*7`(%rsp),%xmm13 | |
| 605 | movaps `$framesz+16*8`(%rsp),%xmm14 | |
| 606 | movaps `$framesz+16*9`(%rsp),%xmm15 | |
| 607 | ___ | |
| 608 | $code.=<<___; | |
| 609 | mov (%rsi),%r15 | |
| 610 | mov 8(%rsi),%r14 | |
| 611 | mov 16(%rsi),%r13 | |
| 612 | mov 24(%rsi),%r12 | |
| 613 | mov 32(%rsi),%rbp | |
| 614 | mov 40(%rsi),%rbx | |
| 615 | lea 48(%rsi),%rsp | |
| 616 | .Lepilogue_xop: | |
| 617 | ret | |
| 618 | .size ${func}_xop,.-${func}_xop | |
| 619 | ___ | |
| 620 | ###################################################################### | |
| 621 | # AVX+shrd code path | |
| 622 | # | |
| 623 | local *ror = sub { &shrd(@_[0],@_) }; | |
| 624 | ||
| 625 | $code.=<<___; | |
| 626 | .type ${func}_avx,\@function,6 | |
| 627 | .align 64 | |
| 628 | ${func}_avx: | |
| 629 | .Lavx_shortcut: | |
| 630 | mov `($win64?56:8)`(%rsp),$in0 # load 7th parameter | |
| 631 | push %rbx | |
| 632 | push %rbp | |
| 633 | push %r12 | |
| 634 | push %r13 | |
| 635 | push %r14 | |
| 636 | push %r15 | |
| 637 | mov %rsp,%r11 # copy %rsp | |
| 638 | sub \$`$framesz+$win64*16*10`,%rsp | |
| 639 | and \$-64,%rsp # align stack frame | |
| 640 | ||
| 641 | shl \$6,$len | |
| 642 | sub $inp,$out # re-bias | |
| 643 | sub $inp,$in0 | |
| 644 | add $inp,$len # end of input | |
| 645 | ||
| 646 | #mov $inp,$_inp # saved later | |
| 647 | mov $out,$_out | |
| 648 | mov $len,$_end | |
| 649 | #mov $key,$_key # remains resident in $inp register | |
| 650 | mov $ivp,$_ivp | |
| 651 | mov $ctx,$_ctx | |
| 652 | mov $in0,$_in0 | |
| 653 | mov %r11,$_rsp | |
| 654 | ___ | |
| 655 | $code.=<<___ if ($win64); | |
| 656 | movaps %xmm6,`$framesz+16*0`(%rsp) | |
| 657 | movaps %xmm7,`$framesz+16*1`(%rsp) | |
| 658 | movaps %xmm8,`$framesz+16*2`(%rsp) | |
| 659 | movaps %xmm9,`$framesz+16*3`(%rsp) | |
| 660 | movaps %xmm10,`$framesz+16*4`(%rsp) | |
| 661 | movaps %xmm11,`$framesz+16*5`(%rsp) | |
| 662 | movaps %xmm12,`$framesz+16*6`(%rsp) | |
| 663 | movaps %xmm13,`$framesz+16*7`(%rsp) | |
| 664 | movaps %xmm14,`$framesz+16*8`(%rsp) | |
| 665 | movaps %xmm15,`$framesz+16*9`(%rsp) | |
| 666 | ___ | |
| 667 | $code.=<<___; | |
| 668 | .Lprologue_avx: | |
| 669 | vzeroall | |
| 670 | ||
| 671 | mov $inp,%r12 # borrow $a4 | |
| 672 | lea 0x80($key),$inp # size optimization, reassign | |
| 673 | lea $TABLE+`$SZ*2*$rounds+32`(%rip),%r13 # borrow $a0 | |
| 674 | mov 0xf0-0x80($inp),%r14d # rounds, borrow $a1 | |
| 675 | mov $ctx,%r15 # borrow $a2 | |
| 676 | mov $in0,%rsi # borrow $a3 | |
| 677 | vmovdqu ($ivp),$iv # load IV | |
| 678 | sub \$9,%r14 | |
| 679 | ||
| 680 | mov $SZ*0(%r15),$A | |
| 681 | mov $SZ*1(%r15),$B | |
| 682 | mov $SZ*2(%r15),$C | |
| 683 | mov $SZ*3(%r15),$D | |
| 684 | mov $SZ*4(%r15),$E | |
| 685 | mov $SZ*5(%r15),$F | |
| 686 | mov $SZ*6(%r15),$G | |
| 687 | mov $SZ*7(%r15),$H | |
| 688 | ||
| 689 | vmovdqa 0x00(%r13,%r14,8),$mask14 | |
| 690 | vmovdqa 0x10(%r13,%r14,8),$mask12 | |
| 691 | vmovdqa 0x20(%r13,%r14,8),$mask10 | |
| 692 | vmovdqu 0x00-0x80($inp),$roundkey | |
| 693 | ___ | |
| 694 | if ($SZ==4) { # SHA256 | |
| 695 | my @X = map("%xmm$_",(0..3)); | |
| 696 | my ($t0,$t1,$t2,$t3) = map("%xmm$_",(4..7)); | |
| 697 | ||
| 698 | $code.=<<___; | |
| 699 | jmp .Lloop_avx | |
| 700 | .align 16 | |
| 701 | .Lloop_avx: | |
| 702 | vmovdqa $TABLE+`$SZ*2*$rounds`(%rip),$t3 | |
| 703 | vmovdqu 0x00(%rsi,%r12),@X[0] | |
| 704 | vmovdqu 0x10(%rsi,%r12),@X[1] | |
| 705 | vmovdqu 0x20(%rsi,%r12),@X[2] | |
| 706 | vmovdqu 0x30(%rsi,%r12),@X[3] | |
| 707 | vpshufb $t3,@X[0],@X[0] | |
| 708 | lea $TABLE(%rip),$Tbl | |
| 709 | vpshufb $t3,@X[1],@X[1] | |
| 710 | vpshufb $t3,@X[2],@X[2] | |
| 711 | vpaddd 0x00($Tbl),@X[0],$t0 | |
| 712 | vpshufb $t3,@X[3],@X[3] | |
| 713 | vpaddd 0x20($Tbl),@X[1],$t1 | |
| 714 | vpaddd 0x40($Tbl),@X[2],$t2 | |
| 715 | vpaddd 0x60($Tbl),@X[3],$t3 | |
| 716 | vmovdqa $t0,0x00(%rsp) | |
| 717 | mov $A,$a1 | |
| 718 | vmovdqa $t1,0x10(%rsp) | |
| 719 | mov $B,$a3 | |
| 720 | vmovdqa $t2,0x20(%rsp) | |
| 721 | xor $C,$a3 # magic | |
| 722 | vmovdqa $t3,0x30(%rsp) | |
| 723 | mov $E,$a0 | |
| 724 | jmp .Lavx_00_47 | |
| 725 | ||
| 726 | .align 16 | |
| 727 | .Lavx_00_47: | |
| 728 | sub \$-16*2*$SZ,$Tbl # size optimization | |
| 729 | vmovdqu (%r12),$inout # $a4 | |
| 730 | mov %r12,$_inp # $a4 | |
| 731 | ___ | |
| 732 | sub Xupdate_256_AVX () { | |
| 733 | ( | |
| 734 | '&vpalignr ($t0,@X[1],@X[0],$SZ)', # X[1..4] | |
| 735 | '&vpalignr ($t3,@X[3],@X[2],$SZ)', # X[9..12] | |
| 736 | '&vpsrld ($t2,$t0,$sigma0[0]);', | |
| 737 | '&vpaddd (@X[0],@X[0],$t3)', # X[0..3] += X[9..12] | |
| 738 | '&vpsrld ($t3,$t0,$sigma0[2])', | |
| 739 | '&vpslld ($t1,$t0,8*$SZ-$sigma0[1]);', | |
| 740 | '&vpxor ($t0,$t3,$t2)', | |
| 741 | '&vpshufd ($t3,@X[3],0b11111010)',# X[14..15] | |
| 742 | '&vpsrld ($t2,$t2,$sigma0[1]-$sigma0[0]);', | |
| 743 | '&vpxor ($t0,$t0,$t1)', | |
| 744 | '&vpslld ($t1,$t1,$sigma0[1]-$sigma0[0]);', | |
| 745 | '&vpxor ($t0,$t0,$t2)', | |
| 746 | '&vpsrld ($t2,$t3,$sigma1[2]);', | |
| 747 | '&vpxor ($t0,$t0,$t1)', # sigma0(X[1..4]) | |
| 748 | '&vpsrlq ($t3,$t3,$sigma1[0]);', | |
| 749 | '&vpaddd (@X[0],@X[0],$t0)', # X[0..3] += sigma0(X[1..4]) | |
| 750 | '&vpxor ($t2,$t2,$t3);', | |
| 751 | '&vpsrlq ($t3,$t3,$sigma1[1]-$sigma1[0])', | |
| 752 | '&vpxor ($t2,$t2,$t3)', # sigma1(X[14..15]) | |
| 753 | '&vpshufd ($t2,$t2,0b10000100)', | |
| 754 | '&vpsrldq ($t2,$t2,8)', | |
| 755 | '&vpaddd (@X[0],@X[0],$t2)', # X[0..1] += sigma1(X[14..15]) | |
| 756 | '&vpshufd ($t3,@X[0],0b01010000)',# X[16..17] | |
| 757 | '&vpsrld ($t2,$t3,$sigma1[2])', | |
| 758 | '&vpsrlq ($t3,$t3,$sigma1[0])', | |
| 759 | '&vpxor ($t2,$t2,$t3);', | |
| 760 | '&vpsrlq ($t3,$t3,$sigma1[1]-$sigma1[0])', | |
| 761 | '&vpxor ($t2,$t2,$t3)', | |
| 762 | '&vpshufd ($t2,$t2,0b11101000)', | |
| 763 | '&vpslldq ($t2,$t2,8)', | |
| 764 | '&vpaddd (@X[0],@X[0],$t2)' # X[2..3] += sigma1(X[16..17]) | |
| 765 | ); | |
| 766 | } | |
| 767 | ||
| 768 | sub AVX_256_00_47 () { | |
| 769 | my $j = shift; | |
| 770 | my $body = shift; | |
| 771 | my @X = @_; | |
| 772 | my @insns = (&$body,&$body,&$body,&$body); # 104 instructions | |
| 773 | ||
| 774 | foreach (Xupdate_256_AVX()) { # 29 instructions | |
| 775 | eval; | |
| 776 | eval(shift(@insns)); | |
| 777 | eval(shift(@insns)); | |
| 778 | eval(shift(@insns)); | |
| 779 | } | |
| 780 | &vpaddd ($t2,@X[0],16*2*$j."($Tbl)"); | |
| 781 | foreach (@insns) { eval; } # remaining instructions | |
| 782 | &vmovdqa (16*$j."(%rsp)",$t2); | |
| 783 | } | |
| 784 | ||
| 785 | $aesni_cbc_idx=0; | |
| 786 | for ($i=0,$j=0; $j<4; $j++) { | |
| 787 | &AVX_256_00_47($j,\&body_00_15,@X); | |
| 788 | push(@X,shift(@X)); # rotate(@X) | |
| 789 | } | |
| 790 | &mov ("%r12",$_inp); # borrow $a4 | |
| 791 | &vpand ($temp,$temp,$mask14); | |
| 792 | &mov ("%r15",$_out); # borrow $a2 | |
| 793 | &vpor ($iv,$iv,$temp); | |
| 794 | &vmovdqu ("(%r15,%r12)",$iv); # write output | |
| 795 | &lea ("%r12","16(%r12)"); # inp++ | |
| 796 | ||
| 797 | &cmpb ($SZ-1+16*2*$SZ."($Tbl)",0); | |
| 798 | &jne (".Lavx_00_47"); | |
| 799 | ||
| 800 | &vmovdqu ($inout,"(%r12)"); | |
| 801 | &mov ($_inp,"%r12"); | |
| 802 | ||
| 803 | $aesni_cbc_idx=0; | |
| 804 | for ($i=0; $i<16; ) { | |
| 805 | foreach(body_00_15()) { eval; } | |
| 806 | } | |
| 807 | ||
| 808 | } | |
| 809 | $code.=<<___; | |
| 810 | mov $_inp,%r12 # borrow $a4 | |
| 811 | mov $_out,%r13 # borrow $a0 | |
| 812 | mov $_ctx,%r15 # borrow $a2 | |
| 813 | mov $_in0,%rsi # borrow $a3 | |
| 814 | ||
| 815 | vpand $mask14,$temp,$temp | |
| 816 | mov $a1,$A | |
| 817 | vpor $temp,$iv,$iv | |
| 818 | vmovdqu $iv,(%r13,%r12) # write output | |
| 819 | lea 16(%r12),%r12 # inp++ | |
| 820 | ||
| 821 | add $SZ*0(%r15),$A | |
| 822 | add $SZ*1(%r15),$B | |
| 823 | add $SZ*2(%r15),$C | |
| 824 | add $SZ*3(%r15),$D | |
| 825 | add $SZ*4(%r15),$E | |
| 826 | add $SZ*5(%r15),$F | |
| 827 | add $SZ*6(%r15),$G | |
| 828 | add $SZ*7(%r15),$H | |
| 829 | ||
| 830 | cmp $_end,%r12 | |
| 831 | ||
| 832 | mov $A,$SZ*0(%r15) | |
| 833 | mov $B,$SZ*1(%r15) | |
| 834 | mov $C,$SZ*2(%r15) | |
| 835 | mov $D,$SZ*3(%r15) | |
| 836 | mov $E,$SZ*4(%r15) | |
| 837 | mov $F,$SZ*5(%r15) | |
| 838 | mov $G,$SZ*6(%r15) | |
| 839 | mov $H,$SZ*7(%r15) | |
| 840 | jb .Lloop_avx | |
| 841 | ||
| 842 | mov $_ivp,$ivp | |
| 843 | mov $_rsp,%rsi | |
| 844 | vmovdqu $iv,($ivp) # output IV | |
| 845 | vzeroall | |
| 846 | ___ | |
| 847 | $code.=<<___ if ($win64); | |
| 848 | movaps `$framesz+16*0`(%rsp),%xmm6 | |
| 849 | movaps `$framesz+16*1`(%rsp),%xmm7 | |
| 850 | movaps `$framesz+16*2`(%rsp),%xmm8 | |
| 851 | movaps `$framesz+16*3`(%rsp),%xmm9 | |
| 852 | movaps `$framesz+16*4`(%rsp),%xmm10 | |
| 853 | movaps `$framesz+16*5`(%rsp),%xmm11 | |
| 854 | movaps `$framesz+16*6`(%rsp),%xmm12 | |
| 855 | movaps `$framesz+16*7`(%rsp),%xmm13 | |
| 856 | movaps `$framesz+16*8`(%rsp),%xmm14 | |
| 857 | movaps `$framesz+16*9`(%rsp),%xmm15 | |
| 858 | ___ | |
| 859 | $code.=<<___; | |
| 860 | mov (%rsi),%r15 | |
| 861 | mov 8(%rsi),%r14 | |
| 862 | mov 16(%rsi),%r13 | |
| 863 | mov 24(%rsi),%r12 | |
| 864 | mov 32(%rsi),%rbp | |
| 865 | mov 40(%rsi),%rbx | |
| 866 | lea 48(%rsi),%rsp | |
| 867 | .Lepilogue_avx: | |
| 868 | ret | |
| 869 | .size ${func}_avx,.-${func}_avx | |
| 870 | ___ | |
| 871 | ||
| 872 | if ($avx>1) {{ | |
| 873 | ###################################################################### | |
| 874 | # AVX2+BMI code path | |
| 875 | # | |
| 876 | my $a5=$SZ==4?"%esi":"%rsi"; # zap $inp | |
| 877 | my $PUSH8=8*2*$SZ; | |
| 878 | use integer; | |
| 879 | ||
| 880 | sub bodyx_00_15 () { | |
| 881 | # at start $a1 should be zero, $a3 - $b^$c and $a4 copy of $f | |
| 882 | ( | |
| 883 | '($a,$b,$c,$d,$e,$f,$g,$h)=@ROT;'. | |
| 884 | ||
| 885 | '&add ($h,(32*($i/(16/$SZ))+$SZ*($i%(16/$SZ)))%$PUSH8.$base)', # h+=X[i]+K[i] | |
| 886 | '&and ($a4,$e)', # f&e | |
| 887 | '&rorx ($a0,$e,$Sigma1[2])', | |
| 888 | '&rorx ($a2,$e,$Sigma1[1])', | |
| 889 | ||
| 890 | '&lea ($a,"($a,$a1)")', # h+=Sigma0(a) from the past | |
| 891 | '&lea ($h,"($h,$a4)")', | |
| 892 | '&andn ($a4,$e,$g)', # ~e&g | |
| 893 | '&xor ($a0,$a2)', | |
| 894 | ||
| 895 | '&rorx ($a1,$e,$Sigma1[0])', | |
| 896 | '&lea ($h,"($h,$a4)")', # h+=Ch(e,f,g)=(e&f)+(~e&g) | |
| 897 | '&xor ($a0,$a1)', # Sigma1(e) | |
| 898 | '&mov ($a2,$a)', | |
| 899 | ||
| 900 | '&rorx ($a4,$a,$Sigma0[2])', | |
| 901 | '&lea ($h,"($h,$a0)")', # h+=Sigma1(e) | |
| 902 | '&xor ($a2,$b)', # a^b, b^c in next round | |
| 903 | '&rorx ($a1,$a,$Sigma0[1])', | |
| 904 | ||
| 905 | '&rorx ($a0,$a,$Sigma0[0])', | |
| 906 | '&lea ($d,"($d,$h)")', # d+=h | |
| 907 | '&and ($a3,$a2)', # (b^c)&(a^b) | |
| 908 | @aesni_cbc_block[$aesni_cbc_idx++]. | |
| 909 | '&xor ($a1,$a4)', | |
| 910 | ||
| 911 | '&xor ($a3,$b)', # Maj(a,b,c)=Ch(a^b,c,b) | |
| 912 | '&xor ($a1,$a0)', # Sigma0(a) | |
| 913 | '&lea ($h,"($h,$a3)");'. # h+=Maj(a,b,c) | |
| 914 | '&mov ($a4,$e)', # copy of f in future | |
| 915 | ||
| 916 | '($a2,$a3) = ($a3,$a2); unshift(@ROT,pop(@ROT)); $i++;' | |
| 917 | ); | |
| 918 | # and at the finish one has to $a+=$a1 | |
| 919 | } | |
| 920 | ||
| 921 | $code.=<<___; | |
| 922 | .type ${func}_avx2,\@function,6 | |
| 923 | .align 64 | |
| 924 | ${func}_avx2: | |
| 925 | .Lavx2_shortcut: | |
| 926 | mov `($win64?56:8)`(%rsp),$in0 # load 7th parameter | |
| 927 | push %rbx | |
| 928 | push %rbp | |
| 929 | push %r12 | |
| 930 | push %r13 | |
| 931 | push %r14 | |
| 932 | push %r15 | |
| 933 | mov %rsp,%r11 # copy %rsp | |
| 934 | sub \$`2*$SZ*$rounds+8*8+$win64*16*10`,%rsp | |
| 935 | and \$-256*$SZ,%rsp # align stack frame | |
| 936 | add \$`2*$SZ*($rounds-8)`,%rsp | |
| 937 | ||
| 938 | shl \$6,$len | |
| 939 | sub $inp,$out # re-bias | |
| 940 | sub $inp,$in0 | |
| 941 | add $inp,$len # end of input | |
| 942 | ||
| 943 | #mov $inp,$_inp # saved later | |
| 944 | #mov $out,$_out # kept in $offload | |
| 945 | mov $len,$_end | |
| 946 | #mov $key,$_key # remains resident in $inp register | |
| 947 | mov $ivp,$_ivp | |
| 948 | mov $ctx,$_ctx | |
| 949 | mov $in0,$_in0 | |
| 950 | mov %r11,$_rsp | |
| 951 | ___ | |
| 952 | $code.=<<___ if ($win64); | |
| 953 | movaps %xmm6,`$framesz+16*0`(%rsp) | |
| 954 | movaps %xmm7,`$framesz+16*1`(%rsp) | |
| 955 | movaps %xmm8,`$framesz+16*2`(%rsp) | |
| 956 | movaps %xmm9,`$framesz+16*3`(%rsp) | |
| 957 | movaps %xmm10,`$framesz+16*4`(%rsp) | |
| 958 | movaps %xmm11,`$framesz+16*5`(%rsp) | |
| 959 | movaps %xmm12,`$framesz+16*6`(%rsp) | |
| 960 | movaps %xmm13,`$framesz+16*7`(%rsp) | |
| 961 | movaps %xmm14,`$framesz+16*8`(%rsp) | |
| 962 | movaps %xmm15,`$framesz+16*9`(%rsp) | |
| 963 | ___ | |
| 964 | $code.=<<___; | |
| 965 | .Lprologue_avx2: | |
| 966 | vzeroall | |
| 967 | ||
| 968 | mov $inp,%r13 # borrow $a0 | |
| 969 | vpinsrq \$1,$out,$offload,$offload | |
| 970 | lea 0x80($key),$inp # size optimization, reassign | |
| 971 | lea $TABLE+`$SZ*2*$rounds+32`(%rip),%r12 # borrow $a4 | |
| 972 | mov 0xf0-0x80($inp),%r14d # rounds, borrow $a1 | |
| 973 | mov $ctx,%r15 # borrow $a2 | |
| 974 | mov $in0,%rsi # borrow $a3 | |
| 975 | vmovdqu ($ivp),$iv # load IV | |
| 976 | lea -9(%r14),%r14 | |
| 977 | ||
| 978 | vmovdqa 0x00(%r12,%r14,8),$mask14 | |
| 979 | vmovdqa 0x10(%r12,%r14,8),$mask12 | |
| 980 | vmovdqa 0x20(%r12,%r14,8),$mask10 | |
| 981 | ||
| 982 | sub \$-16*$SZ,%r13 # inp++, size optimization | |
| 983 | mov $SZ*0(%r15),$A | |
| 42b9a417 | 984 | lea (%rsi,%r13),%r12 # borrow $a0 |
| 8a97a330 AP |
985 | mov $SZ*1(%r15),$B |
| 986 | cmp $len,%r13 # $_end | |
| 987 | mov $SZ*2(%r15),$C | |
| 42b9a417 | 988 | cmove %rsp,%r12 # next block or random data |
| 8a97a330 AP |
989 | mov $SZ*3(%r15),$D |
| 990 | mov $SZ*4(%r15),$E | |
| 991 | mov $SZ*5(%r15),$F | |
| 992 | mov $SZ*6(%r15),$G | |
| 993 | mov $SZ*7(%r15),$H | |
| 994 | vmovdqu 0x00-0x80($inp),$roundkey | |
| 995 | ___ | |
| 996 | if ($SZ==4) { # SHA256 | |
| 997 | my @X = map("%ymm$_",(0..3)); | |
| 998 | my ($t0,$t1,$t2,$t3) = map("%ymm$_",(4..7)); | |
| 999 | ||
| 1000 | $code.=<<___; | |
| 1001 | jmp .Loop_avx2 | |
| 1002 | .align 16 | |
| 1003 | .Loop_avx2: | |
| 8a97a330 | 1004 | vmovdqa $TABLE+`$SZ*2*$rounds`(%rip),$t3 |
| 8a97a330 | 1005 | vmovdqu -16*$SZ+0(%rsi,%r13),%xmm0 |
| 8a97a330 AP |
1006 | vmovdqu -16*$SZ+16(%rsi,%r13),%xmm1 |
| 1007 | vmovdqu -16*$SZ+32(%rsi,%r13),%xmm2 | |
| 1008 | vmovdqu -16*$SZ+48(%rsi,%r13),%xmm3 | |
| 1009 | ||
| 42b9a417 AP |
1010 | vinserti128 \$1,(%r12),@X[0],@X[0] |
| 1011 | vinserti128 \$1,16(%r12),@X[1],@X[1] | |
| 8a97a330 | 1012 | vpshufb $t3,@X[0],@X[0] |
| 42b9a417 | 1013 | vinserti128 \$1,32(%r12),@X[2],@X[2] |
| 8a97a330 | 1014 | vpshufb $t3,@X[1],@X[1] |
| 42b9a417 | 1015 | vinserti128 \$1,48(%r12),@X[3],@X[3] |
| 8a97a330 AP |
1016 | |
| 1017 | lea $TABLE(%rip),$Tbl | |
| 1018 | vpshufb $t3,@X[2],@X[2] | |
| 1019 | lea -16*$SZ(%r13),%r13 | |
| 1020 | vpaddd 0x00($Tbl),@X[0],$t0 | |
| 1021 | vpshufb $t3,@X[3],@X[3] | |
| 1022 | vpaddd 0x20($Tbl),@X[1],$t1 | |
| 1023 | vpaddd 0x40($Tbl),@X[2],$t2 | |
| 1024 | vpaddd 0x60($Tbl),@X[3],$t3 | |
| 1025 | vmovdqa $t0,0x00(%rsp) | |
| 1026 | xor $a1,$a1 | |
| 1027 | vmovdqa $t1,0x20(%rsp) | |
| 1028 | lea -$PUSH8(%rsp),%rsp | |
| 1029 | mov $B,$a3 | |
| 1030 | vmovdqa $t2,0x00(%rsp) | |
| 1031 | xor $C,$a3 # magic | |
| 1032 | vmovdqa $t3,0x20(%rsp) | |
| 1033 | mov $F,$a4 | |
| 1034 | sub \$-16*2*$SZ,$Tbl # size optimization | |
| 1035 | jmp .Lavx2_00_47 | |
| 1036 | ||
| 1037 | .align 16 | |
| 1038 | .Lavx2_00_47: | |
| 1039 | vmovdqu (%r13),$inout | |
| 1040 | vpinsrq \$0,%r13,$offload,$offload | |
| 1041 | ___ | |
| 1042 | ||
| 1043 | sub AVX2_256_00_47 () { | |
| 1044 | my $j = shift; | |
| 1045 | my $body = shift; | |
| 1046 | my @X = @_; | |
| 1047 | my @insns = (&$body,&$body,&$body,&$body); # 96 instructions | |
| 1048 | my $base = "+2*$PUSH8(%rsp)"; | |
| 1049 | ||
| 1050 | &lea ("%rsp","-$PUSH8(%rsp)") if (($j%2)==0); | |
| 1051 | foreach (Xupdate_256_AVX()) { # 29 instructions | |
| 1052 | eval; | |
| 1053 | eval(shift(@insns)); | |
| 1054 | eval(shift(@insns)); | |
| 1055 | eval(shift(@insns)); | |
| 1056 | } | |
| 1057 | &vpaddd ($t2,@X[0],16*2*$j."($Tbl)"); | |
| 1058 | foreach (@insns) { eval; } # remaining instructions | |
| 1059 | &vmovdqa ((32*$j)%$PUSH8."(%rsp)",$t2); | |
| 1060 | } | |
| 1061 | $aesni_cbc_idx=0; | |
| 1062 | for ($i=0,$j=0; $j<4; $j++) { | |
| 1063 | &AVX2_256_00_47($j,\&bodyx_00_15,@X); | |
| 1064 | push(@X,shift(@X)); # rotate(@X) | |
| 1065 | } | |
| 1066 | &vmovq ("%r13",$offload); # borrow $a0 | |
| 1067 | &vpextrq ("%r15",$offload,1); # borrow $a2 | |
| 1068 | &vpand ($temp,$temp,$mask14); | |
| 1069 | &vpor ($iv,$iv,$temp); | |
| 1070 | &vmovdqu ("(%r15,%r13)",$iv); # write output | |
| 1071 | &lea ("%r13","16(%r13)"); # inp++ | |
| 1072 | ||
| 1073 | &lea ($Tbl,16*2*$SZ."($Tbl)"); | |
| 1074 | &cmpb (($SZ-1)."($Tbl)",0); | |
| 1075 | &jne (".Lavx2_00_47"); | |
| 1076 | ||
| 1077 | &vmovdqu ($inout,"(%r13)"); | |
| 1078 | &vpinsrq ($offload,$offload,"%r13",0); | |
| 1079 | ||
| 1080 | $aesni_cbc_idx=0; | |
| 1081 | for ($i=0; $i<16; ) { | |
| 1082 | my $base=$i<8?"+$PUSH8(%rsp)":"(%rsp)"; | |
| 1083 | foreach(bodyx_00_15()) { eval; } | |
| 1084 | } | |
| 1085 | } | |
| 1086 | $code.=<<___; | |
| 1087 | vpextrq \$1,$offload,%r12 # $_out, borrow $a4 | |
| 1088 | vmovq $offload,%r13 # $_inp, borrow $a0 | |
| 1089 | mov `2*$SZ*$rounds+5*8`(%rsp),%r15 # $_ctx, borrow $a2 | |
| 1090 | add $a1,$A | |
| 1091 | lea `2*$SZ*($rounds-8)`(%rsp),$Tbl | |
| 1092 | ||
| 1093 | vpand $mask14,$temp,$temp | |
| 1094 | vpor $temp,$iv,$iv | |
| 1095 | vmovdqu $iv,(%r12,%r13) # write output | |
| 1096 | lea 16(%r13),%r13 | |
| 1097 | ||
| 1098 | add $SZ*0(%r15),$A | |
| 1099 | add $SZ*1(%r15),$B | |
| 1100 | add $SZ*2(%r15),$C | |
| 1101 | add $SZ*3(%r15),$D | |
| 1102 | add $SZ*4(%r15),$E | |
| 1103 | add $SZ*5(%r15),$F | |
| 1104 | add $SZ*6(%r15),$G | |
| 1105 | add $SZ*7(%r15),$H | |
| 1106 | ||
| 1107 | mov $A,$SZ*0(%r15) | |
| 1108 | mov $B,$SZ*1(%r15) | |
| 1109 | mov $C,$SZ*2(%r15) | |
| 1110 | mov $D,$SZ*3(%r15) | |
| 1111 | mov $E,$SZ*4(%r15) | |
| 1112 | mov $F,$SZ*5(%r15) | |
| 1113 | mov $G,$SZ*6(%r15) | |
| 1114 | mov $H,$SZ*7(%r15) | |
| 1115 | ||
| 1116 | cmp `$PUSH8+2*8`($Tbl),%r13 # $_end | |
| 1117 | je .Ldone_avx2 | |
| 1118 | ||
| 1119 | xor $a1,$a1 | |
| 1120 | mov $B,$a3 | |
| 1121 | mov $F,$a4 | |
| 1122 | xor $C,$a3 # magic | |
| 1123 | jmp .Lower_avx2 | |
| 1124 | .align 16 | |
| 1125 | .Lower_avx2: | |
| 1126 | vmovdqu (%r13),$inout | |
| 1127 | vpinsrq \$0,%r13,$offload,$offload | |
| 1128 | ___ | |
| 1129 | $aesni_cbc_idx=0; | |
| 1130 | for ($i=0; $i<16; ) { | |
| 1131 | my $base="+16($Tbl)"; | |
| 1132 | foreach(bodyx_00_15()) { eval; } | |
| 1133 | &lea ($Tbl,"-$PUSH8($Tbl)") if ($i==8); | |
| 1134 | } | |
| 1135 | $code.=<<___; | |
| 1136 | vmovq $offload,%r13 # borrow $a0 | |
| 1137 | vpextrq \$1,$offload,%r15 # borrow $a2 | |
| 1138 | vpand $mask14,$temp,$temp | |
| 1139 | vpor $temp,$iv,$iv | |
| 1140 | lea -$PUSH8($Tbl),$Tbl | |
| 1141 | vmovdqu $iv,(%r15,%r13) # write output | |
| 1142 | lea 16(%r13),%r13 # inp++ | |
| 1143 | cmp %rsp,$Tbl | |
| 1144 | jae .Lower_avx2 | |
| 1145 | ||
| 1146 | mov `2*$SZ*$rounds+5*8`(%rsp),%r15 # $_ctx, borrow $a2 | |
| 1147 | lea 16*$SZ(%r13),%r13 | |
| 1148 | mov `2*$SZ*$rounds+6*8`(%rsp),%rsi # $_in0, borrow $a3 | |
| 1149 | add $a1,$A | |
| 1150 | lea `2*$SZ*($rounds-8)`(%rsp),%rsp | |
| 1151 | ||
| 1152 | add $SZ*0(%r15),$A | |
| 1153 | add $SZ*1(%r15),$B | |
| 1154 | add $SZ*2(%r15),$C | |
| 1155 | add $SZ*3(%r15),$D | |
| 1156 | add $SZ*4(%r15),$E | |
| 1157 | add $SZ*5(%r15),$F | |
| 1158 | add $SZ*6(%r15),$G | |
| 42b9a417 | 1159 | lea (%rsi,%r13),%r12 |
| 8a97a330 AP |
1160 | add $SZ*7(%r15),$H |
| 1161 | ||
| 1162 | cmp $_end,%r13 | |
| 1163 | ||
| 1164 | mov $A,$SZ*0(%r15) | |
| 42b9a417 | 1165 | cmove %rsp,%r12 # next block or stale data |
| 8a97a330 AP |
1166 | mov $B,$SZ*1(%r15) |
| 1167 | mov $C,$SZ*2(%r15) | |
| 1168 | mov $D,$SZ*3(%r15) | |
| 1169 | mov $E,$SZ*4(%r15) | |
| 1170 | mov $F,$SZ*5(%r15) | |
| 1171 | mov $G,$SZ*6(%r15) | |
| 1172 | mov $H,$SZ*7(%r15) | |
| 1173 | ||
| 8a97a330 AP |
1174 | jbe .Loop_avx2 |
| 1175 | lea (%rsp),$Tbl | |
| 1176 | ||
| 1177 | .Ldone_avx2: | |
| 1178 | lea ($Tbl),%rsp | |
| 1179 | mov $_ivp,$ivp | |
| 1180 | mov $_rsp,%rsi | |
| 1181 | vmovdqu $iv,($ivp) # output IV | |
| 1182 | vzeroall | |
| 1183 | ___ | |
| 1184 | $code.=<<___ if ($win64); | |
| 1185 | movaps `$framesz+16*0`(%rsp),%xmm6 | |
| 1186 | movaps `$framesz+16*1`(%rsp),%xmm7 | |
| 1187 | movaps `$framesz+16*2`(%rsp),%xmm8 | |
| 1188 | movaps `$framesz+16*3`(%rsp),%xmm9 | |
| 1189 | movaps `$framesz+16*4`(%rsp),%xmm10 | |
| 1190 | movaps `$framesz+16*5`(%rsp),%xmm11 | |
| 1191 | movaps `$framesz+16*6`(%rsp),%xmm12 | |
| 1192 | movaps `$framesz+16*7`(%rsp),%xmm13 | |
| 1193 | movaps `$framesz+16*8`(%rsp),%xmm14 | |
| 1194 | movaps `$framesz+16*9`(%rsp),%xmm15 | |
| 1195 | ___ | |
| 1196 | $code.=<<___; | |
| 1197 | mov (%rsi),%r15 | |
| 1198 | mov 8(%rsi),%r14 | |
| 1199 | mov 16(%rsi),%r13 | |
| 1200 | mov 24(%rsi),%r12 | |
| 1201 | mov 32(%rsi),%rbp | |
| 1202 | mov 40(%rsi),%rbx | |
| 1203 | lea 48(%rsi),%rsp | |
| 1204 | .Lepilogue_avx2: | |
| 1205 | ret | |
| 1206 | .size ${func}_avx2,.-${func}_avx2 | |
| 1207 | ___ | |
| 1208 | }} | |
| 619b9466 AP |
1209 | }} |
| 1210 | {{ | |
| 1211 | my ($in0,$out,$len,$key,$ivp,$ctx,$inp)=("%rdi","%rsi","%rdx","%rcx","%r8","%r9","%r10"); | |
| 1212 | ||
| 1213 | my ($rounds,$Tbl)=("%r11d","%rbx"); | |
| 1214 | ||
| 1215 | my ($iv,$in,$rndkey0)=map("%xmm$_",(6,14,15)); | |
| 1216 | my @rndkey=("%xmm4","%xmm5"); | |
| 1217 | my $r=0; | |
| 1218 | my $sn=0; | |
| 1219 | ||
| 1220 | my ($Wi,$ABEF,$CDGH,$TMP,$BSWAP,$ABEF_SAVE,$CDGH_SAVE)=map("%xmm$_",(0..3,7..9)); | |
| 1221 | my @MSG=map("%xmm$_",(10..13)); | |
| 1222 | ||
| 1223 | my $aesenc=sub { | |
| 1224 | use integer; | |
| 1225 | my ($n,$k)=($r/10,$r%10); | |
| 1226 | if ($k==0) { | |
| 1227 | $code.=<<___; | |
| 1228 | movups `16*$n`($in0),$in # load input | |
| 1229 | xorps $rndkey0,$in | |
| 1230 | ___ | |
| 1231 | $code.=<<___ if ($n); | |
| 1232 | movups $iv,`16*($n-1)`($out,$in0) # write output | |
| 1233 | ___ | |
| 1234 | $code.=<<___; | |
| 1235 | xorps $in,$iv | |
| 1236 | movups `32+16*$k-112`($key),$rndkey[1] | |
| 1237 | aesenc $rndkey[0],$iv | |
| 1238 | ___ | |
| 1239 | } elsif ($k==9) { | |
| 1240 | $sn++; | |
| 1241 | $code.=<<___; | |
| 1242 | cmp \$11,$rounds | |
| 1243 | jb .Laesenclast$sn | |
| 1244 | movups `32+16*($k+0)-112`($key),$rndkey[1] | |
| 1245 | aesenc $rndkey[0],$iv | |
| 1246 | movups `32+16*($k+1)-112`($key),$rndkey[0] | |
| 1247 | aesenc $rndkey[1],$iv | |
| 1248 | je .Laesenclast$sn | |
| 1249 | movups `32+16*($k+2)-112`($key),$rndkey[1] | |
| 1250 | aesenc $rndkey[0],$iv | |
| 1251 | movups `32+16*($k+3)-112`($key),$rndkey[0] | |
| 1252 | aesenc $rndkey[1],$iv | |
| 1253 | .Laesenclast$sn: | |
| 1254 | aesenclast $rndkey[0],$iv | |
| 1255 | movups 16-112($key),$rndkey[1] # forward reference | |
| 1256 | nop | |
| 1257 | ___ | |
| 1258 | } else { | |
| 1259 | $code.=<<___; | |
| 1260 | movups `32+16*$k-112`($key),$rndkey[1] | |
| 1261 | aesenc $rndkey[0],$iv | |
| 1262 | ___ | |
| 1263 | } | |
| 1264 | $r++; unshift(@rndkey,pop(@rndkey)); | |
| 1265 | }; | |
| 1266 | ||
| 977f32e8 AP |
1267 | if ($shaext) { |
| 1268 | my $Tbl="%rax"; | |
| 1269 | ||
| 619b9466 AP |
1270 | $code.=<<___; |
| 1271 | .type ${func}_shaext,\@function,6 | |
| 1272 | .align 32 | |
| 1273 | ${func}_shaext: | |
| 619b9466 | 1274 | mov `($win64?56:8)`(%rsp),$inp # load 7th argument |
| 619b9466 AP |
1275 | ___ |
| 1276 | $code.=<<___ if ($win64); | |
| 977f32e8 | 1277 | lea `-8-10*16`(%rsp),%rsp |
| 619b9466 AP |
1278 | movaps %xmm6,-8-10*16(%rax) |
| 1279 | movaps %xmm7,-8-9*16(%rax) | |
| 1280 | movaps %xmm8,-8-8*16(%rax) | |
| 1281 | movaps %xmm9,-8-7*16(%rax) | |
| 1282 | movaps %xmm10,-8-6*16(%rax) | |
| 1283 | movaps %xmm11,-8-5*16(%rax) | |
| 1284 | movaps %xmm12,-8-4*16(%rax) | |
| 1285 | movaps %xmm13,-8-3*16(%rax) | |
| 1286 | movaps %xmm14,-8-2*16(%rax) | |
| 1287 | movaps %xmm15,-8-1*16(%rax) | |
| 1288 | .Lprologue_shaext: | |
| 1289 | ___ | |
| 1290 | $code.=<<___; | |
| 1291 | lea K256+0x80(%rip),$Tbl | |
| 1292 | movdqu ($ctx),$ABEF # DCBA | |
| 1293 | movdqu 16($ctx),$CDGH # HGFE | |
| 1294 | movdqa 0x200-0x80($Tbl),$TMP # byte swap mask | |
| 1295 | ||
| 1296 | mov 240($key),$rounds | |
| 1297 | sub $in0,$out | |
| 1298 | movups ($key),$rndkey0 # $key[0] | |
| 1299 | movups 16($key),$rndkey[0] # forward reference | |
| 1300 | lea 112($key),$key # size optimization | |
| 1301 | ||
| 1302 | pshufd \$0x1b,$ABEF,$Wi # ABCD | |
| 1303 | pshufd \$0xb1,$ABEF,$ABEF # CDAB | |
| 1304 | pshufd \$0x1b,$CDGH,$CDGH # EFGH | |
| 1305 | movdqa $TMP,$BSWAP # offload | |
| 1306 | palignr \$8,$CDGH,$ABEF # ABEF | |
| 1307 | punpcklqdq $Wi,$CDGH # CDGH | |
| 1308 | ||
| 1309 | jmp .Loop_shaext | |
| 1310 | ||
| 1311 | .align 16 | |
| 1312 | .Loop_shaext: | |
| 1313 | movdqu ($inp),@MSG[0] | |
| 1314 | movdqu 0x10($inp),@MSG[1] | |
| 1315 | movdqu 0x20($inp),@MSG[2] | |
| 1316 | pshufb $TMP,@MSG[0] | |
| 1317 | movdqu 0x30($inp),@MSG[3] | |
| 1318 | ||
| 1319 | movdqa 0*32-0x80($Tbl),$Wi | |
| 1320 | paddd @MSG[0],$Wi | |
| 1321 | pshufb $TMP,@MSG[1] | |
| 1322 | movdqa $CDGH,$CDGH_SAVE # offload | |
| 1323 | movdqa $ABEF,$ABEF_SAVE # offload | |
| 1324 | ___ | |
| 1325 | &$aesenc(); | |
| 1326 | $code.=<<___; | |
| 1327 | sha256rnds2 $ABEF,$CDGH # 0-3 | |
| 1328 | pshufd \$0x0e,$Wi,$Wi | |
| 1329 | ___ | |
| 1330 | &$aesenc(); | |
| 1331 | $code.=<<___; | |
| 1332 | sha256rnds2 $CDGH,$ABEF | |
| 1333 | ||
| 1334 | movdqa 1*32-0x80($Tbl),$Wi | |
| 1335 | paddd @MSG[1],$Wi | |
| 1336 | pshufb $TMP,@MSG[2] | |
| 1337 | lea 0x40($inp),$inp | |
| 1338 | ___ | |
| 1339 | &$aesenc(); | |
| 1340 | $code.=<<___; | |
| 1341 | sha256rnds2 $ABEF,$CDGH # 4-7 | |
| 1342 | pshufd \$0x0e,$Wi,$Wi | |
| 1343 | ___ | |
| 1344 | &$aesenc(); | |
| 1345 | $code.=<<___; | |
| 1346 | sha256rnds2 $CDGH,$ABEF | |
| 1347 | ||
| 1348 | movdqa 2*32-0x80($Tbl),$Wi | |
| 1349 | paddd @MSG[2],$Wi | |
| 1350 | pshufb $TMP,@MSG[3] | |
| 1351 | sha256msg1 @MSG[1],@MSG[0] | |
| 1352 | ___ | |
| 1353 | &$aesenc(); | |
| 1354 | $code.=<<___; | |
| 1355 | sha256rnds2 $ABEF,$CDGH # 8-11 | |
| 1356 | pshufd \$0x0e,$Wi,$Wi | |
| 1357 | movdqa @MSG[3],$TMP | |
| 1358 | palignr \$4,@MSG[2],$TMP | |
| 1359 | paddd $TMP,@MSG[0] | |
| 1360 | ___ | |
| 1361 | &$aesenc(); | |
| 1362 | $code.=<<___; | |
| 1363 | sha256rnds2 $CDGH,$ABEF | |
| 1364 | ||
| 1365 | movdqa 3*32-0x80($Tbl),$Wi | |
| 1366 | paddd @MSG[3],$Wi | |
| 1367 | sha256msg2 @MSG[3],@MSG[0] | |
| 1368 | sha256msg1 @MSG[2],@MSG[1] | |
| 1369 | ___ | |
| 1370 | &$aesenc(); | |
| 1371 | $code.=<<___; | |
| 1372 | sha256rnds2 $ABEF,$CDGH # 12-15 | |
| 1373 | pshufd \$0x0e,$Wi,$Wi | |
| 1374 | ___ | |
| 1375 | &$aesenc(); | |
| 1376 | $code.=<<___; | |
| 1377 | movdqa @MSG[0],$TMP | |
| 1378 | palignr \$4,@MSG[3],$TMP | |
| 1379 | paddd $TMP,@MSG[1] | |
| 1380 | sha256rnds2 $CDGH,$ABEF | |
| 1381 | ___ | |
| 1382 | for($i=4;$i<16-3;$i++) { | |
| 1383 | &$aesenc() if (($r%10)==0); | |
| 1384 | $code.=<<___; | |
| 1385 | movdqa $i*32-0x80($Tbl),$Wi | |
| 1386 | paddd @MSG[0],$Wi | |
| 1387 | sha256msg2 @MSG[0],@MSG[1] | |
| 1388 | sha256msg1 @MSG[3],@MSG[2] | |
| 1389 | ___ | |
| 1390 | &$aesenc(); | |
| 1391 | $code.=<<___; | |
| 1392 | sha256rnds2 $ABEF,$CDGH # 16-19... | |
| 1393 | pshufd \$0x0e,$Wi,$Wi | |
| 1394 | movdqa @MSG[1],$TMP | |
| 1395 | palignr \$4,@MSG[0],$TMP | |
| 1396 | paddd $TMP,@MSG[2] | |
| 1397 | ___ | |
| 1398 | &$aesenc(); | |
| 1399 | &$aesenc() if ($r==19); | |
| 1400 | $code.=<<___; | |
| 1401 | sha256rnds2 $CDGH,$ABEF | |
| 1402 | ___ | |
| 1403 | push(@MSG,shift(@MSG)); | |
| 1404 | } | |
| 1405 | $code.=<<___; | |
| 1406 | movdqa 13*32-0x80($Tbl),$Wi | |
| 1407 | paddd @MSG[0],$Wi | |
| 1408 | sha256msg2 @MSG[0],@MSG[1] | |
| 1409 | sha256msg1 @MSG[3],@MSG[2] | |
| 1410 | ___ | |
| 1411 | &$aesenc(); | |
| 1412 | $code.=<<___; | |
| 1413 | sha256rnds2 $ABEF,$CDGH # 52-55 | |
| 1414 | pshufd \$0x0e,$Wi,$Wi | |
| 1415 | movdqa @MSG[1],$TMP | |
| 1416 | palignr \$4,@MSG[0],$TMP | |
| 1417 | paddd $TMP,@MSG[2] | |
| 1418 | ___ | |
| 1419 | &$aesenc(); | |
| 1420 | &$aesenc(); | |
| 1421 | $code.=<<___; | |
| 1422 | sha256rnds2 $CDGH,$ABEF | |
| 1423 | ||
| 1424 | movdqa 14*32-0x80($Tbl),$Wi | |
| 1425 | paddd @MSG[1],$Wi | |
| 1426 | sha256msg2 @MSG[1],@MSG[2] | |
| 1427 | movdqa $BSWAP,$TMP | |
| 1428 | ___ | |
| 1429 | &$aesenc(); | |
| 1430 | $code.=<<___; | |
| 1431 | sha256rnds2 $ABEF,$CDGH # 56-59 | |
| 1432 | pshufd \$0x0e,$Wi,$Wi | |
| 1433 | ___ | |
| 1434 | &$aesenc(); | |
| 1435 | $code.=<<___; | |
| 1436 | sha256rnds2 $CDGH,$ABEF | |
| 1437 | ||
| 1438 | movdqa 15*32-0x80($Tbl),$Wi | |
| 1439 | paddd @MSG[2],$Wi | |
| 1440 | ___ | |
| 1441 | &$aesenc(); | |
| 1442 | &$aesenc(); | |
| 1443 | $code.=<<___; | |
| 1444 | sha256rnds2 $ABEF,$CDGH # 60-63 | |
| 1445 | pshufd \$0x0e,$Wi,$Wi | |
| 1446 | ___ | |
| 1447 | &$aesenc(); | |
| 1448 | $code.=<<___; | |
| 1449 | sha256rnds2 $CDGH,$ABEF | |
| 1450 | #pxor $CDGH,$rndkey0 # black magic | |
| 1451 | ___ | |
| 1452 | while ($r<40) { &$aesenc(); } # remaining aesenc's | |
| 1453 | $code.=<<___; | |
| 1454 | #xorps $CDGH,$rndkey0 # black magic | |
| 1455 | paddd $CDGH_SAVE,$CDGH | |
| 1456 | paddd $ABEF_SAVE,$ABEF | |
| 1457 | ||
| 1458 | dec $len | |
| 1459 | movups $iv,48($out,$in0) # write output | |
| 1460 | lea 64($in0),$in0 | |
| 1461 | jnz .Loop_shaext | |
| 1462 | ||
| 1463 | pshufd \$0xb1,$CDGH,$CDGH # DCHG | |
| 1464 | pshufd \$0x1b,$ABEF,$TMP # FEBA | |
| 1465 | pshufd \$0xb1,$ABEF,$ABEF # BAFE | |
| 1466 | punpckhqdq $CDGH,$ABEF # DCBA | |
| 1467 | palignr \$8,$TMP,$CDGH # HGFE | |
| 1468 | ||
| 1469 | movups $iv,($ivp) # write IV | |
| 1470 | movdqu $ABEF,($ctx) | |
| 1471 | movdqu $CDGH,16($ctx) | |
| 1472 | ___ | |
| 1473 | $code.=<<___ if ($win64); | |
| 977f32e8 AP |
1474 | movaps 0*16(%rsp),%xmm6 |
| 1475 | movaps 1*16(%rsp),%xmm7 | |
| 1476 | movaps 2*16(%rsp),%xmm8 | |
| 1477 | movaps 3*16(%rsp),%xmm9 | |
| 1478 | movaps 4*16(%rsp),%xmm10 | |
| 1479 | movaps 5*16(%rsp),%xmm11 | |
| 1480 | movaps 6*16(%rsp),%xmm12 | |
| 1481 | movaps 7*16(%rsp),%xmm13 | |
| 1482 | movaps 8*16(%rsp),%xmm14 | |
| 1483 | movaps 9*16(%rsp),%xmm15 | |
| 1484 | lea 8+10*16(%rsp),%rsp | |
| 619b9466 AP |
1485 | .Lepilogue_shaext: |
| 1486 | ___ | |
| 1487 | $code.=<<___; | |
| 619b9466 AP |
1488 | ret |
| 1489 | .size ${func}_shaext,.-${func}_shaext | |
| 1490 | ___ | |
| 977f32e8 | 1491 | } |
| 8a97a330 AP |
1492 | }}}}} |
| 1493 | ||
| 1494 | # EXCEPTION_DISPOSITION handler (EXCEPTION_RECORD *rec,ULONG64 frame, | |
| 1495 | # CONTEXT *context,DISPATCHER_CONTEXT *disp) | |
| 1496 | if ($win64) { | |
| 1497 | $rec="%rcx"; | |
| 1498 | $frame="%rdx"; | |
| 1499 | $context="%r8"; | |
| 1500 | $disp="%r9"; | |
| 1501 | ||
| 1502 | $code.=<<___ if ($avx); | |
| 1503 | .extern __imp_RtlVirtualUnwind | |
| 1504 | .type se_handler,\@abi-omnipotent | |
| 1505 | .align 16 | |
| 1506 | se_handler: | |
| 1507 | push %rsi | |
| 1508 | push %rdi | |
| 1509 | push %rbx | |
| 1510 | push %rbp | |
| 1511 | push %r12 | |
| 1512 | push %r13 | |
| 1513 | push %r14 | |
| 1514 | push %r15 | |
| 1515 | pushfq | |
| 1516 | sub \$64,%rsp | |
| 1517 | ||
| 1518 | mov 120($context),%rax # pull context->Rax | |
| 1519 | mov 248($context),%rbx # pull context->Rip | |
| 1520 | ||
| 1521 | mov 8($disp),%rsi # disp->ImageBase | |
| 1522 | mov 56($disp),%r11 # disp->HanderlData | |
| 1523 | ||
| 1524 | mov 0(%r11),%r10d # HandlerData[0] | |
| 1525 | lea (%rsi,%r10),%r10 # prologue label | |
| 1526 | cmp %r10,%rbx # context->Rip<prologue label | |
| 1527 | jb .Lin_prologue | |
| 1528 | ||
| 1529 | mov 152($context),%rax # pull context->Rsp | |
| 1530 | ||
| 1531 | mov 4(%r11),%r10d # HandlerData[1] | |
| 1532 | lea (%rsi,%r10),%r10 # epilogue label | |
| 1533 | cmp %r10,%rbx # context->Rip>=epilogue label | |
| 1534 | jae .Lin_prologue | |
| 1535 | ___ | |
| 977f32e8 AP |
1536 | $code.=<<___ if ($shaext); |
| 1537 | lea aesni_cbc_sha256_enc_shaext(%rip),%r10 | |
| 1538 | cmp %r10,%rbx | |
| 1539 | jb .Lnot_in_shaext | |
| 1540 | ||
| 1541 | lea (%rax),%rsi | |
| 1542 | lea 512($context),%rdi # &context.Xmm6 | |
| 1543 | mov \$20,%ecx | |
| 1544 | .long 0xa548f3fc # cld; rep movsq | |
| 1545 | lea 168(%rax),%rax # adjust stack pointer | |
| 1546 | jmp .Lin_prologue | |
| 1547 | .Lnot_in_shaext: | |
| 1548 | ___ | |
| 8a97a330 AP |
1549 | $code.=<<___ if ($avx>1); |
| 1550 | lea .Lavx2_shortcut(%rip),%r10 | |
| 1551 | cmp %r10,%rbx # context->Rip<avx2_shortcut | |
| 1552 | jb .Lnot_in_avx2 | |
| 1553 | ||
| 1554 | and \$-256*$SZ,%rax | |
| 1555 | add \$`2*$SZ*($rounds-8)`,%rax | |
| 1556 | .Lnot_in_avx2: | |
| 1557 | ___ | |
| 1558 | $code.=<<___; | |
| 1559 | mov %rax,%rsi # put aside Rsp | |
| 1560 | mov 16*$SZ+7*8(%rax),%rax # pull $_rsp | |
| 1561 | lea 48(%rax),%rax | |
| 1562 | ||
| 1563 | mov -8(%rax),%rbx | |
| 1564 | mov -16(%rax),%rbp | |
| 1565 | mov -24(%rax),%r12 | |
| 1566 | mov -32(%rax),%r13 | |
| 1567 | mov -40(%rax),%r14 | |
| 1568 | mov -48(%rax),%r15 | |
| 1569 | mov %rbx,144($context) # restore context->Rbx | |
| 1570 | mov %rbp,160($context) # restore context->Rbp | |
| 1571 | mov %r12,216($context) # restore context->R12 | |
| 1572 | mov %r13,224($context) # restore context->R13 | |
| 1573 | mov %r14,232($context) # restore context->R14 | |
| 1574 | mov %r15,240($context) # restore context->R15 | |
| 1575 | ||
| 8a97a330 AP |
1576 | lea 16*$SZ+8*8(%rsi),%rsi # Xmm6- save area |
| 1577 | lea 512($context),%rdi # &context.Xmm6 | |
| 1578 | mov \$20,%ecx | |
| 1579 | .long 0xa548f3fc # cld; rep movsq | |
| 1580 | ||
| 1581 | .Lin_prologue: | |
| 1582 | mov 8(%rax),%rdi | |
| 1583 | mov 16(%rax),%rsi | |
| 1584 | mov %rax,152($context) # restore context->Rsp | |
| 1585 | mov %rsi,168($context) # restore context->Rsi | |
| 1586 | mov %rdi,176($context) # restore context->Rdi | |
| 1587 | ||
| 1588 | mov 40($disp),%rdi # disp->ContextRecord | |
| 1589 | mov $context,%rsi # context | |
| 1590 | mov \$154,%ecx # sizeof(CONTEXT) | |
| 1591 | .long 0xa548f3fc # cld; rep movsq | |
| 1592 | ||
| 1593 | mov $disp,%rsi | |
| 1594 | xor %rcx,%rcx # arg1, UNW_FLAG_NHANDLER | |
| 1595 | mov 8(%rsi),%rdx # arg2, disp->ImageBase | |
| 1596 | mov 0(%rsi),%r8 # arg3, disp->ControlPc | |
| 1597 | mov 16(%rsi),%r9 # arg4, disp->FunctionEntry | |
| 1598 | mov 40(%rsi),%r10 # disp->ContextRecord | |
| 1599 | lea 56(%rsi),%r11 # &disp->HandlerData | |
| 1600 | lea 24(%rsi),%r12 # &disp->EstablisherFrame | |
| 1601 | mov %r10,32(%rsp) # arg5 | |
| 1602 | mov %r11,40(%rsp) # arg6 | |
| 1603 | mov %r12,48(%rsp) # arg7 | |
| 1604 | mov %rcx,56(%rsp) # arg8, (NULL) | |
| 1605 | call *__imp_RtlVirtualUnwind(%rip) | |
| 1606 | ||
| 1607 | mov \$1,%eax # ExceptionContinueSearch | |
| 1608 | add \$64,%rsp | |
| 1609 | popfq | |
| 1610 | pop %r15 | |
| 1611 | pop %r14 | |
| 1612 | pop %r13 | |
| 1613 | pop %r12 | |
| 1614 | pop %rbp | |
| 1615 | pop %rbx | |
| 1616 | pop %rdi | |
| 1617 | pop %rsi | |
| 1618 | ret | |
| 1619 | .size se_handler,.-se_handler | |
| 1620 | ||
| 1621 | .section .pdata | |
| 1622 | .rva .LSEH_begin_${func}_xop | |
| 1623 | .rva .LSEH_end_${func}_xop | |
| 1624 | .rva .LSEH_info_${func}_xop | |
| 1625 | ||
| 1626 | .rva .LSEH_begin_${func}_avx | |
| 1627 | .rva .LSEH_end_${func}_avx | |
| 1628 | .rva .LSEH_info_${func}_avx | |
| 1629 | ___ | |
| 1630 | $code.=<<___ if ($avx>1); | |
| 1631 | .rva .LSEH_begin_${func}_avx2 | |
| 1632 | .rva .LSEH_end_${func}_avx2 | |
| 1633 | .rva .LSEH_info_${func}_avx2 | |
| 1634 | ___ | |
| 977f32e8 AP |
1635 | $code.=<<___ if ($shaext); |
| 1636 | .rva .LSEH_begin_${func}_shaext | |
| 1637 | .rva .LSEH_end_${func}_shaext | |
| 1638 | .rva .LSEH_info_${func}_shaext | |
| 1639 | ___ | |
| 8a97a330 AP |
1640 | $code.=<<___ if ($avx); |
| 1641 | .section .xdata | |
| 1642 | .align 8 | |
| 1643 | .LSEH_info_${func}_xop: | |
| 1644 | .byte 9,0,0,0 | |
| 1645 | .rva se_handler | |
| 1646 | .rva .Lprologue_xop,.Lepilogue_xop # HandlerData[] | |
| 1647 | ||
| 1648 | .LSEH_info_${func}_avx: | |
| 1649 | .byte 9,0,0,0 | |
| 1650 | .rva se_handler | |
| 1651 | .rva .Lprologue_avx,.Lepilogue_avx # HandlerData[] | |
| 1652 | ___ | |
| 1653 | $code.=<<___ if ($avx>1); | |
| 1654 | .LSEH_info_${func}_avx2: | |
| 1655 | .byte 9,0,0,0 | |
| 1656 | .rva se_handler | |
| 1657 | .rva .Lprologue_avx2,.Lepilogue_avx2 # HandlerData[] | |
| 1658 | ___ | |
| 977f32e8 AP |
1659 | $code.=<<___ if ($shaext); |
| 1660 | .LSEH_info_${func}_shaext: | |
| 1661 | .byte 9,0,0,0 | |
| 1662 | .rva se_handler | |
| 1663 | .rva .Lprologue_shaext,.Lepilogue_shaext # HandlerData[] | |
| 1664 | ___ | |
| 8a97a330 AP |
1665 | } |
| 1666 | ||
| 619b9466 AP |
1667 | #################################################################### |
| 1668 | sub rex { | |
| 1669 | local *opcode=shift; | |
| 1670 | my ($dst,$src)=@_; | |
| 1671 | my $rex=0; | |
| 1672 | ||
| 1673 | $rex|=0x04 if($dst>=8); | |
| 1674 | $rex|=0x01 if($src>=8); | |
| 1675 | unshift @opcode,$rex|0x40 if($rex); | |
| 1676 | } | |
| 1677 | ||
| 1678 | { | |
| 1679 | my %opcodelet = ( | |
| 1680 | "sha256rnds2" => 0xcb, | |
| 1681 | "sha256msg1" => 0xcc, | |
| 1682 | "sha256msg2" => 0xcd ); | |
| 1683 | ||
| 1684 | sub sha256op38 { | |
| 1685 | my $instr = shift; | |
| 1686 | ||
| 1687 | if (defined($opcodelet{$instr}) && @_[0] =~ /%xmm([0-7]),\s*%xmm([0-7])/) { | |
| 1688 | my @opcode=(0x0f,0x38); | |
| 1689 | rex(\@opcode,$2,$1); | |
| 1690 | push @opcode,$opcodelet{$instr}; | |
| 1691 | push @opcode,0xc0|($1&7)|(($2&7)<<3); # ModR/M | |
| 1692 | return ".byte\t".join(',',@opcode); | |
| 1693 | } else { | |
| 1694 | return $instr."\t".@_[0]; | |
| 1695 | } | |
| 1696 | } | |
| 1697 | } | |
| 1698 | ||
| 8a97a330 | 1699 | $code =~ s/\`([^\`]*)\`/eval $1/gem; |
| 619b9466 | 1700 | $code =~ s/\b(sha256[^\s]*)\s+(.*)/sha256op38($1,$2)/gem; |
| 8a97a330 AP |
1701 | print $code; |
| 1702 | close STDOUT; |