VirtualBox

source: vbox/trunk/src/libs/openssl-3.0.3/crypto/sha/asm/sha1-armv8.pl@ 96159

最後變更 在這個檔案從96159是 94082,由 vboxsync 提交於 3 年 前

libs/openssl-3.0.1: started applying and adjusting our OpenSSL changes to 3.0.1. bugref:10128

檔案大小: 8.3 KB
 
1#! /usr/bin/env perl
2# Copyright 2014-2020 The OpenSSL Project Authors. All Rights Reserved.
3#
4# Licensed under the Apache License 2.0 (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# ====================================================================
11# Written by Andy Polyakov <[email protected]> for the OpenSSL
12# project. The module is, however, dual licensed under OpenSSL and
13# CRYPTOGAMS licenses depending on where you obtain it. For further
14# details see http://www.openssl.org/~appro/cryptogams/.
15# ====================================================================
16#
17# SHA1 for ARMv8.
18#
19# Performance in cycles per processed byte and improvement coefficient
20# over code generated with "default" compiler:
21#
22# hardware-assisted software(*)
23# Apple A7 2.31 4.13 (+14%)
24# Cortex-A53 2.24 8.03 (+97%)
25# Cortex-A57 2.35 7.88 (+74%)
26# Denver 2.13 3.97 (+0%)(**)
27# X-Gene 8.80 (+200%)
28# Mongoose 2.05 6.50 (+160%)
29# Kryo 1.88 8.00 (+90%)
30# ThunderX2 2.64 6.36 (+150%)
31#
32# (*) Software results are presented mostly for reference purposes.
33# (**) Keep in mind that Denver relies on binary translation, which
34# optimizes compiler output at run-time.
35
36# $output is the last argument if it looks like a file (it has an extension)
37# $flavour is the first argument if it doesn't look like a file
38$output = $#ARGV >= 0 && $ARGV[$#ARGV] =~ m|\.\w+$| ? pop : undef;
39$flavour = $#ARGV >= 0 && $ARGV[0] !~ m|\.| ? shift : undef;
40
41$0 =~ m/(.*[\/\\])[^\/\\]+$/; $dir=$1;
42( $xlate="${dir}arm-xlate.pl" and -f $xlate ) or
43( $xlate="${dir}../../perlasm/arm-xlate.pl" and -f $xlate) or
44die "can't locate arm-xlate.pl";
45
46open OUT,"| \"$^X\" $xlate $flavour \"$output\""
47 or die "can't call $xlate: $1";
48*STDOUT=*OUT;
49
50($ctx,$inp,$num)=("x0","x1","x2");
51@Xw=map("w$_",(3..17,19));
52@Xx=map("x$_",(3..17,19));
53@V=($A,$B,$C,$D,$E)=map("w$_",(20..24));
54($t0,$t1,$t2,$K)=map("w$_",(25..28));
55
56
57sub BODY_00_19 {
58my ($i,$a,$b,$c,$d,$e)=@_;
59my $j=($i+2)&15;
60
61$code.=<<___ if ($i<15 && !($i&1));
62 lsr @Xx[$i+1],@Xx[$i],#32
63___
64$code.=<<___ if ($i<14 && !($i&1));
65 ldur @Xx[$i+2],[$inp,#`($i+2)*4-64`]
66___
67$code.=<<___ if ($i<14 && ($i&1));
68#ifdef __AARCH64EB__
69 ror @Xx[$i+1],@Xx[$i+1],#32
70#else
71 rev32 @Xx[$i+1],@Xx[$i+1]
72#endif
73___
74$code.=<<___ if ($i<14);
75 bic $t0,$d,$b
76 and $t1,$c,$b
77 ror $t2,$a,#27
78 add $d,$d,$K // future e+=K
79 orr $t0,$t0,$t1
80 add $e,$e,$t2 // e+=rot(a,5)
81 ror $b,$b,#2
82 add $d,$d,@Xw[($i+1)&15] // future e+=X[i]
83 add $e,$e,$t0 // e+=F(b,c,d)
84___
85$code.=<<___ if ($i==19);
86 movz $K,#0xeba1
87 movk $K,#0x6ed9,lsl#16
88___
89$code.=<<___ if ($i>=14);
90 eor @Xw[$j],@Xw[$j],@Xw[($j+2)&15]
91 bic $t0,$d,$b
92 and $t1,$c,$b
93 ror $t2,$a,#27
94 eor @Xw[$j],@Xw[$j],@Xw[($j+8)&15]
95 add $d,$d,$K // future e+=K
96 orr $t0,$t0,$t1
97 add $e,$e,$t2 // e+=rot(a,5)
98 eor @Xw[$j],@Xw[$j],@Xw[($j+13)&15]
99 ror $b,$b,#2
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[$j],#31
103___
104}
105
106sub BODY_40_59 {
107my ($i,$a,$b,$c,$d,$e)=@_;
108my $j=($i+2)&15;
109
110$code.=<<___ if ($i==59);
111 movz $K,#0xc1d6
112 movk $K,#0xca62,lsl#16
113___
114$code.=<<___;
115 orr $t0,$b,$c
116 and $t1,$b,$c
117 eor @Xw[$j],@Xw[$j],@Xw[($j+2)&15]
118 ror $t2,$a,#27
119 and $t0,$t0,$d
120 add $d,$d,$K // future e+=K
121 eor @Xw[$j],@Xw[$j],@Xw[($j+8)&15]
122 add $e,$e,$t2 // e+=rot(a,5)
123 orr $t0,$t0,$t1
124 ror $b,$b,#2
125 eor @Xw[$j],@Xw[$j],@Xw[($j+13)&15]
126 add $d,$d,@Xw[($i+1)&15] // future e+=X[i]
127 add $e,$e,$t0 // e+=F(b,c,d)
128 ror @Xw[$j],@Xw[$j],#31
129___
130}
131
132sub BODY_20_39 {
133my ($i,$a,$b,$c,$d,$e)=@_;
134my $j=($i+2)&15;
135
136$code.=<<___ if ($i==39);
137 movz $K,#0xbcdc
138 movk $K,#0x8f1b,lsl#16
139___
140$code.=<<___ if ($i<78);
141 eor @Xw[$j],@Xw[$j],@Xw[($j+2)&15]
142 eor $t0,$d,$b
143 ror $t2,$a,#27
144 add $d,$d,$K // future e+=K
145 eor @Xw[$j],@Xw[$j],@Xw[($j+8)&15]
146 eor $t0,$t0,$c
147 add $e,$e,$t2 // e+=rot(a,5)
148 ror $b,$b,#2
149 eor @Xw[$j],@Xw[$j],@Xw[($j+13)&15]
150 add $d,$d,@Xw[($i+1)&15] // future e+=X[i]
151 add $e,$e,$t0 // e+=F(b,c,d)
152 ror @Xw[$j],@Xw[$j],#31
153___
154$code.=<<___ if ($i==78);
155 ldp @Xw[1],@Xw[2],[$ctx]
156 eor $t0,$d,$b
157 ror $t2,$a,#27
158 add $d,$d,$K // future e+=K
159 eor $t0,$t0,$c
160 add $e,$e,$t2 // e+=rot(a,5)
161 ror $b,$b,#2
162 add $d,$d,@Xw[($i+1)&15] // future e+=X[i]
163 add $e,$e,$t0 // e+=F(b,c,d)
164___
165$code.=<<___ if ($i==79);
166 ldp @Xw[3],@Xw[4],[$ctx,#8]
167 eor $t0,$d,$b
168 ror $t2,$a,#27
169 eor $t0,$t0,$c
170 add $e,$e,$t2 // e+=rot(a,5)
171 ror $b,$b,#2
172 ldr @Xw[5],[$ctx,#16]
173 add $e,$e,$t0 // e+=F(b,c,d)
174___
175}
176
177$code.=<<___;
178#ifndef __KERNEL__
179# include "arm_arch.h"
180.extern OPENSSL_armcap_P
181.hidden OPENSSL_armcap_P
182#endif
183
184.text
185
186.globl sha1_block_data_order
187.type sha1_block_data_order,%function
188.align 6
189sha1_block_data_order:
190 adrp x16,OPENSSL_armcap_P
191 ldr w16,[x16,#:lo12:OPENSSL_armcap_P]
192 tst w16,#ARMV8_SHA1
193 b.ne .Lv8_entry
194
195 stp x29,x30,[sp,#-96]!
196 add x29,sp,#0
197 stp x19,x20,[sp,#16]
198 stp x21,x22,[sp,#32]
199 stp x23,x24,[sp,#48]
200 stp x25,x26,[sp,#64]
201 stp x27,x28,[sp,#80]
202
203 ldp $A,$B,[$ctx]
204 ldp $C,$D,[$ctx,#8]
205 ldr $E,[$ctx,#16]
206
207.Loop:
208 ldr @Xx[0],[$inp],#64
209 movz $K,#0x7999
210 sub $num,$num,#1
211 movk $K,#0x5a82,lsl#16
212#ifdef __AARCH64EB__
213 ror $Xx[0],@Xx[0],#32
214#else
215 rev32 @Xx[0],@Xx[0]
216#endif
217 add $E,$E,$K // warm it up
218 add $E,$E,@Xw[0]
219___
220for($i=0;$i<20;$i++) { &BODY_00_19($i,@V); unshift(@V,pop(@V)); }
221for(;$i<40;$i++) { &BODY_20_39($i,@V); unshift(@V,pop(@V)); }
222for(;$i<60;$i++) { &BODY_40_59($i,@V); unshift(@V,pop(@V)); }
223for(;$i<80;$i++) { &BODY_20_39($i,@V); unshift(@V,pop(@V)); }
224$code.=<<___;
225 add $B,$B,@Xw[2]
226 add $C,$C,@Xw[3]
227 add $A,$A,@Xw[1]
228 add $D,$D,@Xw[4]
229 add $E,$E,@Xw[5]
230 stp $A,$B,[$ctx]
231 stp $C,$D,[$ctx,#8]
232 str $E,[$ctx,#16]
233 cbnz $num,.Loop
234
235 ldp x19,x20,[sp,#16]
236 ldp x21,x22,[sp,#32]
237 ldp x23,x24,[sp,#48]
238 ldp x25,x26,[sp,#64]
239 ldp x27,x28,[sp,#80]
240 ldr x29,[sp],#96
241 ret
242.size sha1_block_data_order,.-sha1_block_data_order
243___
244{{{
245my ($ABCD,$E,$E0,$E1)=map("v$_.16b",(0..3));
246my @MSG=map("v$_.16b",(4..7));
247my @Kxx=map("v$_.4s",(16..19));
248my ($W0,$W1)=("v20.4s","v21.4s");
249my $ABCD_SAVE="v22.16b";
250
251$code.=<<___;
252.type sha1_block_armv8,%function
253.align 6
254sha1_block_armv8:
255.Lv8_entry:
256 stp x29,x30,[sp,#-16]!
257 add x29,sp,#0
258
259 adr x4,.Lconst
260 eor $E,$E,$E
261 ld1.32 {$ABCD},[$ctx],#16
262 ld1.32 {$E}[0],[$ctx]
263 sub $ctx,$ctx,#16
264 ld1.32 {@Kxx[0]-@Kxx[3]},[x4]
265
266.Loop_hw:
267 ld1 {@MSG[0]-@MSG[3]},[$inp],#64
268 sub $num,$num,#1
269 rev32 @MSG[0],@MSG[0]
270 rev32 @MSG[1],@MSG[1]
271
272 add.i32 $W0,@Kxx[0],@MSG[0]
273 rev32 @MSG[2],@MSG[2]
274 orr $ABCD_SAVE,$ABCD,$ABCD // offload
275
276 add.i32 $W1,@Kxx[0],@MSG[1]
277 rev32 @MSG[3],@MSG[3]
278 sha1h $E1,$ABCD
279 sha1c $ABCD,$E,$W0 // 0
280 add.i32 $W0,@Kxx[$j],@MSG[2]
281 sha1su0 @MSG[0],@MSG[1],@MSG[2]
282___
283for ($j=0,$i=1;$i<20-3;$i++) {
284my $f=("c","p","m","p")[$i/5];
285$code.=<<___;
286 sha1h $E0,$ABCD // $i
287 sha1$f $ABCD,$E1,$W1
288 add.i32 $W1,@Kxx[$j],@MSG[3]
289 sha1su1 @MSG[0],@MSG[3]
290___
291$code.=<<___ if ($i<20-4);
292 sha1su0 @MSG[1],@MSG[2],@MSG[3]
293___
294 ($E0,$E1)=($E1,$E0); ($W0,$W1)=($W1,$W0);
295 push(@MSG,shift(@MSG)); $j++ if ((($i+3)%5)==0);
296}
297$code.=<<___;
298 sha1h $E0,$ABCD // $i
299 sha1p $ABCD,$E1,$W1
300 add.i32 $W1,@Kxx[$j],@MSG[3]
301
302 sha1h $E1,$ABCD // 18
303 sha1p $ABCD,$E0,$W0
304
305 sha1h $E0,$ABCD // 19
306 sha1p $ABCD,$E1,$W1
307
308 add.i32 $E,$E,$E0
309 add.i32 $ABCD,$ABCD,$ABCD_SAVE
310
311 cbnz $num,.Loop_hw
312
313 st1.32 {$ABCD},[$ctx],#16
314 st1.32 {$E}[0],[$ctx]
315
316 ldr x29,[sp],#16
317 ret
318.size sha1_block_armv8,.-sha1_block_armv8
319.align 6
320.Lconst:
321.long 0x5a827999,0x5a827999,0x5a827999,0x5a827999 //K_00_19
322.long 0x6ed9eba1,0x6ed9eba1,0x6ed9eba1,0x6ed9eba1 //K_20_39
323.long 0x8f1bbcdc,0x8f1bbcdc,0x8f1bbcdc,0x8f1bbcdc //K_40_59
324.long 0xca62c1d6,0xca62c1d6,0xca62c1d6,0xca62c1d6 //K_60_79
325.asciz "SHA1 block transform for ARMv8, CRYPTOGAMS by <appro\@openssl.org>"
326.align 2
327___
328}}}
329
330{ my %opcode = (
331 "sha1c" => 0x5e000000, "sha1p" => 0x5e001000,
332 "sha1m" => 0x5e002000, "sha1su0" => 0x5e003000,
333 "sha1h" => 0x5e280800, "sha1su1" => 0x5e281800 );
334
335 sub unsha1 {
336 my ($mnemonic,$arg)=@_;
337
338 $arg =~ m/[qv]([0-9]+)[^,]*,\s*[qv]([0-9]+)[^,]*(?:,\s*[qv]([0-9]+))?/o
339 &&
340 sprintf ".inst\t0x%08x\t//%s %s",
341 $opcode{$mnemonic}|$1|($2<<5)|($3<<16),
342 $mnemonic,$arg;
343 }
344}
345
346foreach(split("\n",$code)) {
347
348 s/\`([^\`]*)\`/eval($1)/geo;
349
350 s/\b(sha1\w+)\s+([qv].*)/unsha1($1,$2)/geo;
351
352 s/\.\w?32\b//o and s/\.16b/\.4s/go;
353 m/(ld|st)1[^\[]+\[0\]/o and s/\.4s/\.s/go;
354
355 print $_,"\n";
356}
357
358close STDOUT or die "error closing STDOUT: $!";
注意: 瀏覽 TracBrowser 來幫助您使用儲存庫瀏覽器

© 2025 Oracle Support Privacy / Do Not Sell My Info Terms of Use Trademark Policy Automated Access Etiquette