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.
27 open STDOUT,">".shift;
29 ($ctx,$inp,$num)=("x0","x1","x2");
30 @Xw=map("w$_",(3..17,19));
31 @Xx=map("x$_",(3..17,19));
32 @V=($A,$B,$C,$D,$E)=map("w$_",(20..24));
33 ($t0,$t1,$t2,$K)=map("w$_",(25..28));
37 my ($i,$a,$b,$c,$d,$e)=@_;
40 $code.=<<___ if ($i<15 && !($i&1));
41 lsr @Xx[$i+1],@Xx[$i],#32
43 $code.=<<___ if ($i<14 && !($i&1));
44 ldr @Xx[$i+2],[$inp,#`($i+2)*4-64`]
46 $code.=<<___ if ($i<14 && ($i&1));
48 ror @Xx[$i+1],@Xx[$i+1],#32
50 rev32 @Xx[$i+1],@Xx[$i+1]
53 $code.=<<___ if ($i<14);
57 add $d,$d,$K // future e+=K
59 add $e,$e,$t2 // e+=rot(a,5)
61 add $d,$d,@Xw[($i+1)&15] // future e+=X[i]
62 add $e,$e,$t0 // e+=F(b,c,d)
64 $code.=<<___ if ($i==19);
66 movk $K,#0x6ed9,lsl#16
68 $code.=<<___ if ($i>=14);
69 eor @Xw[$j],@Xw[$j],@Xw[($j+2)&15]
73 eor @Xw[$j],@Xw[$j],@Xw[($j+8)&15]
74 add $d,$d,$K // future e+=K
76 add $e,$e,$t2 // e+=rot(a,5)
77 eor @Xw[$j],@Xw[$j],@Xw[($j+13)&15]
79 add $d,$d,@Xw[($i+1)&15] // future e+=X[i]
80 add $e,$e,$t0 // e+=F(b,c,d)
81 ror @Xw[$j],@Xw[$j],#31
86 my ($i,$a,$b,$c,$d,$e)=@_;
89 $code.=<<___ if ($i==59);
91 movk $K,#0xca62,lsl#16
96 eor @Xw[$j],@Xw[$j],@Xw[($j+2)&15]
99 add $d,$d,$K // future e+=K
100 eor @Xw[$j],@Xw[$j],@Xw[($j+8)&15]
101 add $e,$e,$t2 // e+=rot(a,5)
104 eor @Xw[$j],@Xw[$j],@Xw[($j+13)&15]
105 add $d,$d,@Xw[($i+1)&15] // future e+=X[i]
106 add $e,$e,$t0 // e+=F(b,c,d)
107 ror @Xw[$j],@Xw[$j],#31
112 my ($i,$a,$b,$c,$d,$e)=@_;
115 $code.=<<___ if ($i==39);
117 movk $K,#0x8f1b,lsl#16
119 $code.=<<___ if ($i<78);
120 eor @Xw[$j],@Xw[$j],@Xw[($j+2)&15]
123 add $d,$d,$K // future e+=K
124 eor @Xw[$j],@Xw[$j],@Xw[($j+8)&15]
126 add $e,$e,$t2 // e+=rot(a,5)
128 eor @Xw[$j],@Xw[$j],@Xw[($j+13)&15]
129 add $d,$d,@Xw[($i+1)&15] // future e+=X[i]
130 add $e,$e,$t0 // e+=F(b,c,d)
131 ror @Xw[$j],@Xw[$j],#31
133 $code.=<<___ if ($i==78);
134 ldp @Xw[1],@Xw[2],[$ctx]
137 add $d,$d,$K // future e+=K
139 add $e,$e,$t2 // e+=rot(a,5)
141 add $d,$d,@Xw[($i+1)&15] // future e+=X[i]
142 add $e,$e,$t0 // e+=F(b,c,d)
144 $code.=<<___ if ($i==79);
145 ldp @Xw[3],@Xw[4],[$ctx,#8]
149 add $e,$e,$t2 // e+=rot(a,5)
151 ldr @Xw[5],[$ctx,#16]
152 add $e,$e,$t0 // e+=F(b,c,d)
157 #include "arm_arch.h"
161 .globl sha1_block_data_order
162 .type sha1_block_data_order,%function
164 sha1_block_data_order:
165 ldr x16,.LOPENSSL_armcap_P
166 adr x17,.LOPENSSL_armcap_P
172 stp x29,x30,[sp,#-96]!
185 ldr @Xx[0],[$inp],#64
188 movk $K,#0x5a82,lsl#16
190 ror $Xx[0],@Xx[0],#32
194 add $E,$E,$K // warm it up
197 for($i=0;$i<20;$i++) { &BODY_00_19($i,@V); unshift(@V,pop(@V)); }
198 for(;$i<40;$i++) { &BODY_20_39($i,@V); unshift(@V,pop(@V)); }
199 for(;$i<60;$i++) { &BODY_40_59($i,@V); unshift(@V,pop(@V)); }
200 for(;$i<80;$i++) { &BODY_20_39($i,@V); unshift(@V,pop(@V)); }
219 .size sha1_block_data_order,.-sha1_block_data_order
222 my ($ABCD,$E,$E0,$E1)=map("v$_.16b",(0..3));
223 my @MSG=map("v$_.16b",(4..7));
224 my @Kxx=map("v$_.4s",(16..19));
225 my ($W0,$W1)=("v20.4s","v21.4s");
226 my $ABCD_SAVE="v22.16b";
229 .type sha1_block_armv8,%function
233 stp x29,x30,[sp,#-16]!
238 ld1.32 {$ABCD},[$ctx],#16
239 ld1.32 {$E}[0],[$ctx]
241 ld1.32 {@Kxx[0]-@Kxx[3]},[x4]
244 ld1 {@MSG[0]-@MSG[3]},[$inp],#64
246 rev32 @MSG[0],@MSG[0]
247 rev32 @MSG[1],@MSG[1]
249 add.i32 $W0,@Kxx[0],@MSG[0]
250 rev32 @MSG[2],@MSG[2]
251 orr $ABCD_SAVE,$ABCD,$ABCD // offload
253 add.i32 $W1,@Kxx[0],@MSG[1]
254 rev32 @MSG[3],@MSG[3]
256 sha1c $ABCD,$E,$W0 // 0
257 add.i32 $W0,@Kxx[$j],@MSG[2]
258 sha1su0 @MSG[0],@MSG[1],@MSG[2]
260 for ($j=0,$i=1;$i<20-3;$i++) {
261 my $f=("c","p","m","p")[$i/5];
263 sha1h $E0,$ABCD // $i
265 add.i32 $W1,@Kxx[$j],@MSG[3]
266 sha1su1 @MSG[0],@MSG[3]
268 $code.=<<___ if ($i<20-4);
269 sha1su0 @MSG[1],@MSG[2],@MSG[3]
271 ($E0,$E1)=($E1,$E0); ($W0,$W1)=($W1,$W0);
272 push(@MSG,shift(@MSG)); $j++ if ((($i+3)%5)==0);
275 sha1h $E0,$ABCD // $i
277 add.i32 $W1,@Kxx[$j],@MSG[3]
279 sha1h $E1,$ABCD // 18
282 sha1h $E0,$ABCD // 19
286 add.i32 $ABCD,$ABCD,$ABCD_SAVE
290 st1.32 {$ABCD},[$ctx],#16
291 st1.32 {$E}[0],[$ctx]
295 .size sha1_block_armv8,.-sha1_block_armv8
298 .long 0x5a827999,0x5a827999,0x5a827999,0x5a827999 //K_00_19
299 .long 0x6ed9eba1,0x6ed9eba1,0x6ed9eba1,0x6ed9eba1 //K_20_39
300 .long 0x8f1bbcdc,0x8f1bbcdc,0x8f1bbcdc,0x8f1bbcdc //K_40_59
301 .long 0xca62c1d6,0xca62c1d6,0xca62c1d6,0xca62c1d6 //K_60_79
303 .quad OPENSSL_armcap_P-.
304 .asciz "SHA1 block transform for ARMv8, CRYPTOGAMS by <appro\@openssl.org>"
306 .comm OPENSSL_armcap_P,4,4
311 "sha1c" => 0x5e000000, "sha1p" => 0x5e001000,
312 "sha1m" => 0x5e002000, "sha1su0" => 0x5e003000,
313 "sha1h" => 0x5e280800, "sha1su1" => 0x5e281800 );
316 my ($mnemonic,$arg)=@_;
318 $arg =~ m/[qv]([0-9]+)[^,]*,\s*[qv]([0-9]+)[^,]*(?:,\s*[qv]([0-9]+))?/o
320 sprintf ".inst\t0x%08x\t//%s %s",
321 $opcode{$mnemonic}|$1|($2<<5)|($3<<16),
326 foreach(split("\n",$code)) {
328 s/\`([^\`]*)\`/eval($1)/geo;
330 s/\b(sha1\w+)\s+([qv].*)/unsha1($1,$2)/geo;
332 s/\.\w?32\b//o and s/\.16b/\.4s/go;
333 m/(ld|st)1[^\[]+\[0\]/o and s/\.4s/\.s/go;