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 # ====================================================================
12 # Performance in cycles per processed byte and improvement coefficient
13 # over code generated with "default" compiler:
15 # hardware-assisted software(*)
16 # Apple A7 2.31 4.13 (+14%)
17 # Cortex-A53 2.24 8.03 (+97%)
18 # Cortex-A57 2.35 7.88 (+74%)
19 # Denver 2.13 3.97 (+0%)(**)
22 # (*) Software results are presented mostly for reference purposes.
23 # (**) Keep in mind that Denver relies on binary translation, which
24 # optimizes compiler output at run-time.
29 $0 =~ m/(.*[\/\\])[^\/\\]+$/; $dir=$1;
30 ( $xlate="${dir}arm-xlate.pl" and -f $xlate ) or
31 ( $xlate="${dir}../../perlasm/arm-xlate.pl" and -f $xlate) or
32 die "can't locate arm-xlate.pl";
34 open OUT,"| \"$^X\" $xlate $flavour $output";
37 ($ctx,$inp,$num)=("x0","x1","x2");
38 @Xw=map("w$_",(3..17,19));
39 @Xx=map("x$_",(3..17,19));
40 @V=($A,$B,$C,$D,$E)=map("w$_",(20..24));
41 ($t0,$t1,$t2,$K)=map("w$_",(25..28));
45 my ($i,$a,$b,$c,$d,$e)=@_;
48 $code.=<<___ if ($i<15 && !($i&1));
49 lsr @Xx[$i+1],@Xx[$i],#32
51 $code.=<<___ if ($i<14 && !($i&1));
52 ldr @Xx[$i+2],[$inp,#`($i+2)*4-64`]
54 $code.=<<___ if ($i<14 && ($i&1));
56 ror @Xx[$i+1],@Xx[$i+1],#32
58 rev32 @Xx[$i+1],@Xx[$i+1]
61 $code.=<<___ if ($i<14);
65 add $d,$d,$K // future e+=K
67 add $e,$e,$t2 // e+=rot(a,5)
69 add $d,$d,@Xw[($i+1)&15] // future e+=X[i]
70 add $e,$e,$t0 // e+=F(b,c,d)
72 $code.=<<___ if ($i==19);
74 movk $K,#0x6ed9,lsl#16
76 $code.=<<___ if ($i>=14);
77 eor @Xw[$j],@Xw[$j],@Xw[($j+2)&15]
81 eor @Xw[$j],@Xw[$j],@Xw[($j+8)&15]
82 add $d,$d,$K // future e+=K
84 add $e,$e,$t2 // e+=rot(a,5)
85 eor @Xw[$j],@Xw[$j],@Xw[($j+13)&15]
87 add $d,$d,@Xw[($i+1)&15] // future e+=X[i]
88 add $e,$e,$t0 // e+=F(b,c,d)
89 ror @Xw[$j],@Xw[$j],#31
94 my ($i,$a,$b,$c,$d,$e)=@_;
97 $code.=<<___ if ($i==59);
99 movk $K,#0xca62,lsl#16
104 eor @Xw[$j],@Xw[$j],@Xw[($j+2)&15]
107 add $d,$d,$K // future e+=K
108 eor @Xw[$j],@Xw[$j],@Xw[($j+8)&15]
109 add $e,$e,$t2 // e+=rot(a,5)
112 eor @Xw[$j],@Xw[$j],@Xw[($j+13)&15]
113 add $d,$d,@Xw[($i+1)&15] // future e+=X[i]
114 add $e,$e,$t0 // e+=F(b,c,d)
115 ror @Xw[$j],@Xw[$j],#31
120 my ($i,$a,$b,$c,$d,$e)=@_;
123 $code.=<<___ if ($i==39);
125 movk $K,#0x8f1b,lsl#16
127 $code.=<<___ if ($i<78);
128 eor @Xw[$j],@Xw[$j],@Xw[($j+2)&15]
131 add $d,$d,$K // future e+=K
132 eor @Xw[$j],@Xw[$j],@Xw[($j+8)&15]
134 add $e,$e,$t2 // e+=rot(a,5)
136 eor @Xw[$j],@Xw[$j],@Xw[($j+13)&15]
137 add $d,$d,@Xw[($i+1)&15] // future e+=X[i]
138 add $e,$e,$t0 // e+=F(b,c,d)
139 ror @Xw[$j],@Xw[$j],#31
141 $code.=<<___ if ($i==78);
142 ldp @Xw[1],@Xw[2],[$ctx]
145 add $d,$d,$K // future e+=K
147 add $e,$e,$t2 // e+=rot(a,5)
149 add $d,$d,@Xw[($i+1)&15] // future e+=X[i]
150 add $e,$e,$t0 // e+=F(b,c,d)
152 $code.=<<___ if ($i==79);
153 ldp @Xw[3],@Xw[4],[$ctx,#8]
157 add $e,$e,$t2 // e+=rot(a,5)
159 ldr @Xw[5],[$ctx,#16]
160 add $e,$e,$t0 // e+=F(b,c,d)
165 #include "arm_arch.h"
169 .extern OPENSSL_armcap_P
170 .globl sha1_block_data_order
171 .type sha1_block_data_order,%function
173 sha1_block_data_order:
174 ldr x16,.LOPENSSL_armcap_P
175 adr x17,.LOPENSSL_armcap_P
181 stp x29,x30,[sp,#-96]!
194 ldr @Xx[0],[$inp],#64
197 movk $K,#0x5a82,lsl#16
199 ror $Xx[0],@Xx[0],#32
203 add $E,$E,$K // warm it up
206 for($i=0;$i<20;$i++) { &BODY_00_19($i,@V); unshift(@V,pop(@V)); }
207 for(;$i<40;$i++) { &BODY_20_39($i,@V); unshift(@V,pop(@V)); }
208 for(;$i<60;$i++) { &BODY_40_59($i,@V); unshift(@V,pop(@V)); }
209 for(;$i<80;$i++) { &BODY_20_39($i,@V); unshift(@V,pop(@V)); }
228 .size sha1_block_data_order,.-sha1_block_data_order
231 my ($ABCD,$E,$E0,$E1)=map("v$_.16b",(0..3));
232 my @MSG=map("v$_.16b",(4..7));
233 my @Kxx=map("v$_.4s",(16..19));
234 my ($W0,$W1)=("v20.4s","v21.4s");
235 my $ABCD_SAVE="v22.16b";
238 .type sha1_block_armv8,%function
242 stp x29,x30,[sp,#-16]!
247 ld1.32 {$ABCD},[$ctx],#16
248 ld1.32 {$E}[0],[$ctx]
250 ld1.32 {@Kxx[0]-@Kxx[3]},[x4]
253 ld1 {@MSG[0]-@MSG[3]},[$inp],#64
255 rev32 @MSG[0],@MSG[0]
256 rev32 @MSG[1],@MSG[1]
258 add.i32 $W0,@Kxx[0],@MSG[0]
259 rev32 @MSG[2],@MSG[2]
260 orr $ABCD_SAVE,$ABCD,$ABCD // offload
262 add.i32 $W1,@Kxx[0],@MSG[1]
263 rev32 @MSG[3],@MSG[3]
265 sha1c $ABCD,$E,$W0 // 0
266 add.i32 $W0,@Kxx[$j],@MSG[2]
267 sha1su0 @MSG[0],@MSG[1],@MSG[2]
269 for ($j=0,$i=1;$i<20-3;$i++) {
270 my $f=("c","p","m","p")[$i/5];
272 sha1h $E0,$ABCD // $i
274 add.i32 $W1,@Kxx[$j],@MSG[3]
275 sha1su1 @MSG[0],@MSG[3]
277 $code.=<<___ if ($i<20-4);
278 sha1su0 @MSG[1],@MSG[2],@MSG[3]
280 ($E0,$E1)=($E1,$E0); ($W0,$W1)=($W1,$W0);
281 push(@MSG,shift(@MSG)); $j++ if ((($i+3)%5)==0);
284 sha1h $E0,$ABCD // $i
286 add.i32 $W1,@Kxx[$j],@MSG[3]
288 sha1h $E1,$ABCD // 18
291 sha1h $E0,$ABCD // 19
295 add.i32 $ABCD,$ABCD,$ABCD_SAVE
299 st1.32 {$ABCD},[$ctx],#16
300 st1.32 {$E}[0],[$ctx]
304 .size sha1_block_armv8,.-sha1_block_armv8
307 .long 0x5a827999,0x5a827999,0x5a827999,0x5a827999 //K_00_19
308 .long 0x6ed9eba1,0x6ed9eba1,0x6ed9eba1,0x6ed9eba1 //K_20_39
309 .long 0x8f1bbcdc,0x8f1bbcdc,0x8f1bbcdc,0x8f1bbcdc //K_40_59
310 .long 0xca62c1d6,0xca62c1d6,0xca62c1d6,0xca62c1d6 //K_60_79
312 .quad OPENSSL_armcap_P-.
313 .asciz "SHA1 block transform for ARMv8, CRYPTOGAMS by <appro\@openssl.org>"
315 .comm OPENSSL_armcap_P,4,4
320 "sha1c" => 0x5e000000, "sha1p" => 0x5e001000,
321 "sha1m" => 0x5e002000, "sha1su0" => 0x5e003000,
322 "sha1h" => 0x5e280800, "sha1su1" => 0x5e281800 );
325 my ($mnemonic,$arg)=@_;
327 $arg =~ m/[qv]([0-9]+)[^,]*,\s*[qv]([0-9]+)[^,]*(?:,\s*[qv]([0-9]+))?/o
329 sprintf ".inst\t0x%08x\t//%s %s",
330 $opcode{$mnemonic}|$1|($2<<5)|($3<<16),
335 foreach(split("\n",$code)) {
337 s/\`([^\`]*)\`/eval($1)/geo;
339 s/\b(sha1\w+)\s+([qv].*)/unsha1($1,$2)/geo;
341 s/\.\w?32\b//o and s/\.16b/\.4s/go;
342 m/(ld|st)1[^\[]+\[0\]/o and s/\.4s/\.s/go;