sha1-armv8.pl revision 3f9e6ada2c9f7183a41081263585e6a70bbd9f59
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# SHA1 for ARMv8. 11# 12# Performance in cycles per processed byte and improvement coefficient 13# over code generated with "default" compiler: 14# 15# hardware-assisted software(*) 16# Apple A7 2.31 4.13 (+14%) 17# Cortex-A5x n/a n/a 18# 19# (*) Software results are presented mostly for reference purposes. 20 21$flavour = shift; 22open STDOUT,">".shift; 23 24($ctx,$inp,$num)=("x0","x1","x2"); 25@Xw=map("w$_",(3..17,19)); 26@Xx=map("x$_",(3..17,19)); 27@V=($A,$B,$C,$D,$E)=map("w$_",(20..24)); 28($t0,$t1,$t2,$K)=map("w$_",(25..28)); 29 30 31sub BODY_00_19 { 32my ($i,$a,$b,$c,$d,$e)=@_; 33my $j=($i+2)&15; 34 35$code.=<<___ if ($i<15 && !($i&1)); 36 lsr @Xx[$i+1],@Xx[$i],#32 37___ 38$code.=<<___ if ($i<14 && !($i&1)); 39 ldr @Xx[$i+2],[$inp,#`($i+2)*4-64`] 40___ 41$code.=<<___ if ($i<14 && ($i&1)); 42#ifdef __ARMEB__ 43 ror @Xx[$i+1],@Xx[$i+1],#32 44#else 45 rev32 @Xx[$i+1],@Xx[$i+1] 46#endif 47___ 48$code.=<<___ if ($i<14); 49 bic $t0,$d,$b 50 and $t1,$c,$b 51 ror $t2,$a,#27 52 add $d,$d,$K // future e+=K 53 orr $t0,$t0,$t1 54 add $e,$e,$t2 // e+=rot(a,5) 55 ror $b,$b,#2 56 add $d,$d,@Xw[($i+1)&15] // future e+=X[i] 57 add $e,$e,$t0 // e+=F(b,c,d) 58___ 59$code.=<<___ if ($i==19); 60 movz $K,#0xeba1 61 movk $K,#0x6ed9,lsl#16 62___ 63$code.=<<___ if ($i>=14); 64 eor @Xw[$j],@Xw[$j],@Xw[($j+2)&15] 65 bic $t0,$d,$b 66 and $t1,$c,$b 67 ror $t2,$a,#27 68 eor @Xw[$j],@Xw[$j],@Xw[($j+8)&15] 69 add $d,$d,$K // future e+=K 70 orr $t0,$t0,$t1 71 add $e,$e,$t2 // e+=rot(a,5) 72 eor @Xw[$j],@Xw[$j],@Xw[($j+13)&15] 73 ror $b,$b,#2 74 add $d,$d,@Xw[($i+1)&15] // future e+=X[i] 75 add $e,$e,$t0 // e+=F(b,c,d) 76 ror @Xw[$j],@Xw[$j],#31 77___ 78} 79 80sub BODY_40_59 { 81my ($i,$a,$b,$c,$d,$e)=@_; 82my $j=($i+2)&15; 83 84$code.=<<___ if ($i==59); 85 movz $K,#0xc1d6 86 movk $K,#0xca62,lsl#16 87___ 88$code.=<<___; 89 orr $t0,$b,$c 90 and $t1,$b,$c 91 eor @Xw[$j],@Xw[$j],@Xw[($j+2)&15] 92 ror $t2,$a,#27 93 and $t0,$t0,$d 94 add $d,$d,$K // future e+=K 95 eor @Xw[$j],@Xw[$j],@Xw[($j+8)&15] 96 add $e,$e,$t2 // e+=rot(a,5) 97 orr $t0,$t0,$t1 98 ror $b,$b,#2 99 eor @Xw[$j],@Xw[$j],@Xw[($j+13)&15] 100 add $d,$d,@Xw[($i+1)&15] // future e+=X[i] 101 add $e,$e,$t0 // e+=F(b,c,d) 102 ror @Xw[$j],@Xw[$j],#31 103___ 104} 105 106sub BODY_20_39 { 107my ($i,$a,$b,$c,$d,$e)=@_; 108my $j=($i+2)&15; 109 110$code.=<<___ if ($i==39); 111 movz $K,#0xbcdc 112 movk $K,#0x8f1b,lsl#16 113___ 114$code.=<<___ if ($i<78); 115 eor @Xw[$j],@Xw[$j],@Xw[($j+2)&15] 116 eor $t0,$d,$b 117 ror $t2,$a,#27 118 add $d,$d,$K // future e+=K 119 eor @Xw[$j],@Xw[$j],@Xw[($j+8)&15] 120 eor $t0,$t0,$c 121 add $e,$e,$t2 // e+=rot(a,5) 122 ror $b,$b,#2 123 eor @Xw[$j],@Xw[$j],@Xw[($j+13)&15] 124 add $d,$d,@Xw[($i+1)&15] // future e+=X[i] 125 add $e,$e,$t0 // e+=F(b,c,d) 126 ror @Xw[$j],@Xw[$j],#31 127___ 128$code.=<<___ if ($i==78); 129 ldp @Xw[1],@Xw[2],[$ctx] 130 eor $t0,$d,$b 131 ror $t2,$a,#27 132 add $d,$d,$K // future e+=K 133 eor $t0,$t0,$c 134 add $e,$e,$t2 // e+=rot(a,5) 135 ror $b,$b,#2 136 add $d,$d,@Xw[($i+1)&15] // future e+=X[i] 137 add $e,$e,$t0 // e+=F(b,c,d) 138___ 139$code.=<<___ if ($i==79); 140 ldp @Xw[3],@Xw[4],[$ctx,#8] 141 eor $t0,$d,$b 142 ror $t2,$a,#27 143 eor $t0,$t0,$c 144 add $e,$e,$t2 // e+=rot(a,5) 145 ror $b,$b,#2 146 ldr @Xw[5],[$ctx,#16] 147 add $e,$e,$t0 // e+=F(b,c,d) 148___ 149} 150 151$code.=<<___; 152#include "arm_arch.h" 153 154.text 155 156.globl sha1_block_data_order 157.type sha1_block_data_order,%function 158.align 6 159sha1_block_data_order: 160 ldr x16,.LOPENSSL_armcap_P 161 adr x17,.LOPENSSL_armcap_P 162 add x16,x16,x17 163 ldr w16,[x16] 164 tst w16,#ARMV8_SHA1 165 b.ne .Lv8_entry 166 167 stp x29,x30,[sp,#-96]! 168 add x29,sp,#0 169 stp x19,x20,[sp,#16] 170 stp x21,x22,[sp,#32] 171 stp x23,x24,[sp,#48] 172 stp x25,x26,[sp,#64] 173 stp x27,x28,[sp,#80] 174 175 ldp $A,$B,[$ctx] 176 ldp $C,$D,[$ctx,#8] 177 ldr $E,[$ctx,#16] 178 179.Loop: 180 ldr @Xx[0],[$inp],#64 181 movz $K,#0x7999 182 sub $num,$num,#1 183 movk $K,#0x5a82,lsl#16 184#ifdef __ARMEB__ 185 ror $Xx[0],@Xx[0],#32 186#else 187 rev32 @Xx[0],@Xx[0] 188#endif 189 add $E,$E,$K // warm it up 190 add $E,$E,@Xw[0] 191___ 192for($i=0;$i<20;$i++) { &BODY_00_19($i,@V); unshift(@V,pop(@V)); } 193for(;$i<40;$i++) { &BODY_20_39($i,@V); unshift(@V,pop(@V)); } 194for(;$i<60;$i++) { &BODY_40_59($i,@V); unshift(@V,pop(@V)); } 195for(;$i<80;$i++) { &BODY_20_39($i,@V); unshift(@V,pop(@V)); } 196$code.=<<___; 197 add $B,$B,@Xw[2] 198 add $C,$C,@Xw[3] 199 add $A,$A,@Xw[1] 200 add $D,$D,@Xw[4] 201 add $E,$E,@Xw[5] 202 stp $A,$B,[$ctx] 203 stp $C,$D,[$ctx,#8] 204 str $E,[$ctx,#16] 205 cbnz $num,.Loop 206 207 ldp x19,x20,[sp,#16] 208 ldp x21,x22,[sp,#32] 209 ldp x23,x24,[sp,#48] 210 ldp x25,x26,[sp,#64] 211 ldp x27,x28,[sp,#80] 212 ldr x29,[sp],#96 213 ret 214.size sha1_block_data_order,.-sha1_block_data_order 215___ 216{{{ 217my ($ABCD,$E,$E0,$E1)=map("v$_.16b",(0..3)); 218my @MSG=map("v$_.16b",(4..7)); 219my @Kxx=map("v$_.4s",(16..19)); 220my ($W0,$W1)=("v20.4s","v21.4s"); 221my $ABCD_SAVE="v22.16b"; 222 223$code.=<<___; 224.type sha1_block_armv8,%function 225.align 6 226sha1_block_armv8: 227.Lv8_entry: 228 stp x29,x30,[sp,#-16]! 229 add x29,sp,#0 230 231 adr x4,.Lconst 232 eor $E,$E,$E 233 ld1.32 {$ABCD},[$ctx],#16 234 ld1.32 {$E}[0],[$ctx] 235 sub $ctx,$ctx,#16 236 ld1.32 {@Kxx[0]-@Kxx[3]},[x4] 237 238.Loop_hw: 239 ld1 {@MSG[0]-@MSG[3]},[$inp],#64 240 sub $num,$num,#1 241 rev32 @MSG[0],@MSG[0] 242 rev32 @MSG[1],@MSG[1] 243 244 add.i32 $W0,@Kxx[0],@MSG[0] 245 rev32 @MSG[2],@MSG[2] 246 orr $ABCD_SAVE,$ABCD,$ABCD // offload 247 248 add.i32 $W1,@Kxx[0],@MSG[1] 249 rev32 @MSG[3],@MSG[3] 250 sha1h $E1,$ABCD 251 sha1c $ABCD,$E,$W0 // 0 252 add.i32 $W0,@Kxx[$j],@MSG[2] 253 sha1su0 @MSG[0],@MSG[1],@MSG[2] 254___ 255for ($j=0,$i=1;$i<20-3;$i++) { 256my $f=("c","p","m","p")[$i/5]; 257$code.=<<___; 258 sha1h $E0,$ABCD // $i 259 sha1$f $ABCD,$E1,$W1 260 add.i32 $W1,@Kxx[$j],@MSG[3] 261 sha1su1 @MSG[0],@MSG[3] 262___ 263$code.=<<___ if ($i<20-4); 264 sha1su0 @MSG[1],@MSG[2],@MSG[3] 265___ 266 ($E0,$E1)=($E1,$E0); ($W0,$W1)=($W1,$W0); 267 push(@MSG,shift(@MSG)); $j++ if ((($i+3)%5)==0); 268} 269$code.=<<___; 270 sha1h $E0,$ABCD // $i 271 sha1p $ABCD,$E1,$W1 272 add.i32 $W1,@Kxx[$j],@MSG[3] 273 274 sha1h $E1,$ABCD // 18 275 sha1p $ABCD,$E0,$W0 276 277 sha1h $E0,$ABCD // 19 278 sha1p $ABCD,$E1,$W1 279 280 add.i32 $E,$E,$E0 281 add.i32 $ABCD,$ABCD,$ABCD_SAVE 282 283 cbnz $num,.Loop_hw 284 285 st1.32 {$ABCD},[$ctx],#16 286 st1.32 {$E}[0],[$ctx] 287 288 ldr x29,[sp],#16 289 ret 290.size sha1_block_armv8,.-sha1_block_armv8 291.align 6 292.Lconst: 293.long 0x5a827999,0x5a827999,0x5a827999,0x5a827999 //K_00_19 294.long 0x6ed9eba1,0x6ed9eba1,0x6ed9eba1,0x6ed9eba1 //K_20_39 295.long 0x8f1bbcdc,0x8f1bbcdc,0x8f1bbcdc,0x8f1bbcdc //K_40_59 296.long 0xca62c1d6,0xca62c1d6,0xca62c1d6,0xca62c1d6 //K_60_79 297.LOPENSSL_armcap_P: 298.quad OPENSSL_armcap_P-. 299.asciz "SHA1 block transform for ARMv8, CRYPTOGAMS by <appro\@openssl.org>" 300.align 2 301.comm OPENSSL_armcap_P,4,4 302___ 303}}} 304 305{ my %opcode = ( 306 "sha1c" => 0x5e000000, "sha1p" => 0x5e001000, 307 "sha1m" => 0x5e002000, "sha1su0" => 0x5e003000, 308 "sha1h" => 0x5e280800, "sha1su1" => 0x5e281800 ); 309 310 sub unsha1 { 311 my ($mnemonic,$arg)=@_; 312 313 $arg =~ m/[qv]([0-9]+)[^,]*,\s*[qv]([0-9]+)[^,]*(?:,\s*[qv]([0-9]+))?/o 314 && 315 sprintf ".inst\t0x%08x\t//%s %s", 316 $opcode{$mnemonic}|$1|($2<<5)|($3<<16), 317 $mnemonic,$arg; 318 } 319} 320 321foreach(split("\n",$code)) { 322 323 s/\`([^\`]*)\`/eval($1)/geo; 324 325 s/\b(sha1\w+)\s+([qv].*)/unsha1($1,$2)/geo; 326 327 s/\.\w?32\b//o and s/\.16b/\.4s/go; 328 m/(ld|st)1[^\[]+\[0\]/o and s/\.4s/\.s/go; 329 330 print $_,"\n"; 331} 332 333close STDOUT; 334