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 2.19 8.73 (+108%)
18 # Cortex-A57 2.35 7.88 (+74%)
20 # (*) Software results are presented mostly for reference purposes.
23 open STDOUT,">".shift;
25 ($ctx,$inp,$num)=("x0","x1","x2");
26 @Xw=map("w$_",(3..17,19));
27 @Xx=map("x$_",(3..17,19));
28 @V=($A,$B,$C,$D,$E)=map("w$_",(20..24));
29 ($t0,$t1,$t2,$K)=map("w$_",(25..28));
33 my ($i,$a,$b,$c,$d,$e)=@_;
36 $code.=<<___ if ($i<15 && !($i&1));
37 lsr @Xx[$i+1],@Xx[$i],#32
39 $code.=<<___ if ($i<14 && !($i&1));
40 ldr @Xx[$i+2],[$inp,#`($i+2)*4-64`]
42 $code.=<<___ if ($i<14 && ($i&1));
44 ror @Xx[$i+1],@Xx[$i+1],#32
46 rev32 @Xx[$i+1],@Xx[$i+1]
49 $code.=<<___ if ($i<14);
53 add $d,$d,$K // future e+=K
55 add $e,$e,$t2 // e+=rot(a,5)
57 add $d,$d,@Xw[($i+1)&15] // future e+=X[i]
58 add $e,$e,$t0 // e+=F(b,c,d)
60 $code.=<<___ if ($i==19);
62 movk $K,#0x6ed9,lsl#16
64 $code.=<<___ if ($i>=14);
65 eor @Xw[$j],@Xw[$j],@Xw[($j+2)&15]
69 eor @Xw[$j],@Xw[$j],@Xw[($j+8)&15]
70 add $d,$d,$K // future e+=K
72 add $e,$e,$t2 // e+=rot(a,5)
73 eor @Xw[$j],@Xw[$j],@Xw[($j+13)&15]
75 add $d,$d,@Xw[($i+1)&15] // future e+=X[i]
76 add $e,$e,$t0 // e+=F(b,c,d)
77 ror @Xw[$j],@Xw[$j],#31
82 my ($i,$a,$b,$c,$d,$e)=@_;
85 $code.=<<___ if ($i==59);
87 movk $K,#0xca62,lsl#16
92 eor @Xw[$j],@Xw[$j],@Xw[($j+2)&15]
95 add $d,$d,$K // future e+=K
96 eor @Xw[$j],@Xw[$j],@Xw[($j+8)&15]
97 add $e,$e,$t2 // e+=rot(a,5)
100 eor @Xw[$j],@Xw[$j],@Xw[($j+13)&15]
101 add $d,$d,@Xw[($i+1)&15] // future e+=X[i]
102 add $e,$e,$t0 // e+=F(b,c,d)
103 ror @Xw[$j],@Xw[$j],#31
108 my ($i,$a,$b,$c,$d,$e)=@_;
111 $code.=<<___ if ($i==39);
113 movk $K,#0x8f1b,lsl#16
115 $code.=<<___ if ($i<78);
116 eor @Xw[$j],@Xw[$j],@Xw[($j+2)&15]
119 add $d,$d,$K // future e+=K
120 eor @Xw[$j],@Xw[$j],@Xw[($j+8)&15]
122 add $e,$e,$t2 // e+=rot(a,5)
124 eor @Xw[$j],@Xw[$j],@Xw[($j+13)&15]
125 add $d,$d,@Xw[($i+1)&15] // future e+=X[i]
126 add $e,$e,$t0 // e+=F(b,c,d)
127 ror @Xw[$j],@Xw[$j],#31
129 $code.=<<___ if ($i==78);
130 ldp @Xw[1],@Xw[2],[$ctx]
133 add $d,$d,$K // future e+=K
135 add $e,$e,$t2 // e+=rot(a,5)
137 add $d,$d,@Xw[($i+1)&15] // future e+=X[i]
138 add $e,$e,$t0 // e+=F(b,c,d)
140 $code.=<<___ if ($i==79);
141 ldp @Xw[3],@Xw[4],[$ctx,#8]
145 add $e,$e,$t2 // e+=rot(a,5)
147 ldr @Xw[5],[$ctx,#16]
148 add $e,$e,$t0 // e+=F(b,c,d)
153 #include "arm_arch.h"
157 .globl sha1_block_data_order
158 .type sha1_block_data_order,%function
160 sha1_block_data_order:
161 ldr x16,.LOPENSSL_armcap_P
162 adr x17,.LOPENSSL_armcap_P
168 stp x29,x30,[sp,#-96]!
181 ldr @Xx[0],[$inp],#64
184 movk $K,#0x5a82,lsl#16
186 ror $Xx[0],@Xx[0],#32
190 add $E,$E,$K // warm it up
193 for($i=0;$i<20;$i++) { &BODY_00_19($i,@V); unshift(@V,pop(@V)); }
194 for(;$i<40;$i++) { &BODY_20_39($i,@V); unshift(@V,pop(@V)); }
195 for(;$i<60;$i++) { &BODY_40_59($i,@V); unshift(@V,pop(@V)); }
196 for(;$i<80;$i++) { &BODY_20_39($i,@V); unshift(@V,pop(@V)); }
215 .size sha1_block_data_order,.-sha1_block_data_order
218 my ($ABCD,$E,$E0,$E1)=map("v$_.16b",(0..3));
219 my @MSG=map("v$_.16b",(4..7));
220 my @Kxx=map("v$_.4s",(16..19));
221 my ($W0,$W1)=("v20.4s","v21.4s");
222 my $ABCD_SAVE="v22.16b";
225 .type sha1_block_armv8,%function
229 stp x29,x30,[sp,#-16]!
234 ld1.32 {$ABCD},[$ctx],#16
235 ld1.32 {$E}[0],[$ctx]
237 ld1.32 {@Kxx[0]-@Kxx[3]},[x4]
240 ld1 {@MSG[0]-@MSG[3]},[$inp],#64
242 rev32 @MSG[0],@MSG[0]
243 rev32 @MSG[1],@MSG[1]
245 add.i32 $W0,@Kxx[0],@MSG[0]
246 rev32 @MSG[2],@MSG[2]
247 orr $ABCD_SAVE,$ABCD,$ABCD // offload
249 add.i32 $W1,@Kxx[0],@MSG[1]
250 rev32 @MSG[3],@MSG[3]
252 sha1c $ABCD,$E,$W0 // 0
253 add.i32 $W0,@Kxx[$j],@MSG[2]
254 sha1su0 @MSG[0],@MSG[1],@MSG[2]
256 for ($j=0,$i=1;$i<20-3;$i++) {
257 my $f=("c","p","m","p")[$i/5];
259 sha1h $E0,$ABCD // $i
261 add.i32 $W1,@Kxx[$j],@MSG[3]
262 sha1su1 @MSG[0],@MSG[3]
264 $code.=<<___ if ($i<20-4);
265 sha1su0 @MSG[1],@MSG[2],@MSG[3]
267 ($E0,$E1)=($E1,$E0); ($W0,$W1)=($W1,$W0);
268 push(@MSG,shift(@MSG)); $j++ if ((($i+3)%5)==0);
271 sha1h $E0,$ABCD // $i
273 add.i32 $W1,@Kxx[$j],@MSG[3]
275 sha1h $E1,$ABCD // 18
278 sha1h $E0,$ABCD // 19
282 add.i32 $ABCD,$ABCD,$ABCD_SAVE
286 st1.32 {$ABCD},[$ctx],#16
287 st1.32 {$E}[0],[$ctx]
291 .size sha1_block_armv8,.-sha1_block_armv8
294 .long 0x5a827999,0x5a827999,0x5a827999,0x5a827999 //K_00_19
295 .long 0x6ed9eba1,0x6ed9eba1,0x6ed9eba1,0x6ed9eba1 //K_20_39
296 .long 0x8f1bbcdc,0x8f1bbcdc,0x8f1bbcdc,0x8f1bbcdc //K_40_59
297 .long 0xca62c1d6,0xca62c1d6,0xca62c1d6,0xca62c1d6 //K_60_79
299 .quad OPENSSL_armcap_P-.
300 .asciz "SHA1 block transform for ARMv8, CRYPTOGAMS by <appro\@openssl.org>"
302 .comm OPENSSL_armcap_P,4,4
307 "sha1c" => 0x5e000000, "sha1p" => 0x5e001000,
308 "sha1m" => 0x5e002000, "sha1su0" => 0x5e003000,
309 "sha1h" => 0x5e280800, "sha1su1" => 0x5e281800 );
312 my ($mnemonic,$arg)=@_;
314 $arg =~ m/[qv]([0-9]+)[^,]*,\s*[qv]([0-9]+)[^,]*(?:,\s*[qv]([0-9]+))?/o
316 sprintf ".inst\t0x%08x\t//%s %s",
317 $opcode{$mnemonic}|$1|($2<<5)|($3<<16),
322 foreach(split("\n",$code)) {
324 s/\`([^\`]*)\`/eval($1)/geo;
326 s/\b(sha1\w+)\s+([qv].*)/unsha1($1,$2)/geo;
328 s/\.\w?32\b//o and s/\.16b/\.4s/go;
329 m/(ld|st)1[^\[]+\[0\]/o and s/\.4s/\.s/go;