1#! /usr/bin/env perl
2# Copyright 2014-2016 The OpenSSL Project Authors. All Rights Reserved.
3#
4# Licensed under the OpenSSL license (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 <appro@openssl.org> 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#
31# (*)	Software results are presented mostly for reference purposes.
32# (**)	Keep in mind that Denver relies on binary translation, which
33#	optimizes compiler output at run-time.
34
35$flavour = shift;
36$output  = shift;
37
38$0 =~ m/(.*[\/\\])[^\/\\]+$/; $dir=$1;
39( $xlate="${dir}arm-xlate.pl" and -f $xlate ) or
40( $xlate="${dir}../../../perlasm/arm-xlate.pl" and -f $xlate) or
41die "can't locate arm-xlate.pl";
42
43open OUT,"| \"$^X\" $xlate $flavour $output";
44*STDOUT=*OUT;
45
46($ctx,$inp,$num)=("x0","x1","x2");
47@Xw=map("w$_",(3..17,19));
48@Xx=map("x$_",(3..17,19));
49@V=($A,$B,$C,$D,$E)=map("w$_",(20..24));
50($t0,$t1,$t2,$K)=map("w$_",(25..28));
51
52
53sub BODY_00_19 {
54my ($i,$a,$b,$c,$d,$e)=@_;
55my $j=($i+2)&15;
56
57$code.=<<___ if ($i<15 && !($i&1));
58	lsr	@Xx[$i+1],@Xx[$i],#32
59___
60$code.=<<___ if ($i<14 && !($i&1));
61	ldr	@Xx[$i+2],[$inp,#`($i+2)*4-64`]
62___
63$code.=<<___ if ($i<14 && ($i&1));
64#ifdef	__ARMEB__
65	ror	@Xx[$i+1],@Xx[$i+1],#32
66#else
67	rev32	@Xx[$i+1],@Xx[$i+1]
68#endif
69___
70$code.=<<___ if ($i<14);
71	bic	$t0,$d,$b
72	and	$t1,$c,$b
73	ror	$t2,$a,#27
74	add	$d,$d,$K		// future e+=K
75	orr	$t0,$t0,$t1
76	add	$e,$e,$t2		// e+=rot(a,5)
77	ror	$b,$b,#2
78	add	$d,$d,@Xw[($i+1)&15]	// future e+=X[i]
79	add	$e,$e,$t0		// e+=F(b,c,d)
80___
81$code.=<<___ if ($i==19);
82	movz	$K,#0xeba1
83	movk	$K,#0x6ed9,lsl#16
84___
85$code.=<<___ if ($i>=14);
86	 eor	@Xw[$j],@Xw[$j],@Xw[($j+2)&15]
87	bic	$t0,$d,$b
88	and	$t1,$c,$b
89	ror	$t2,$a,#27
90	 eor	@Xw[$j],@Xw[$j],@Xw[($j+8)&15]
91	add	$d,$d,$K		// future e+=K
92	orr	$t0,$t0,$t1
93	add	$e,$e,$t2		// e+=rot(a,5)
94	 eor	@Xw[$j],@Xw[$j],@Xw[($j+13)&15]
95	ror	$b,$b,#2
96	add	$d,$d,@Xw[($i+1)&15]	// future e+=X[i]
97	add	$e,$e,$t0		// e+=F(b,c,d)
98	 ror	@Xw[$j],@Xw[$j],#31
99___
100}
101
102sub BODY_40_59 {
103my ($i,$a,$b,$c,$d,$e)=@_;
104my $j=($i+2)&15;
105
106$code.=<<___ if ($i==59);
107	movz	$K,#0xc1d6
108	movk	$K,#0xca62,lsl#16
109___
110$code.=<<___;
111	orr	$t0,$b,$c
112	and	$t1,$b,$c
113	 eor	@Xw[$j],@Xw[$j],@Xw[($j+2)&15]
114	ror	$t2,$a,#27
115	and	$t0,$t0,$d
116	add	$d,$d,$K		// future e+=K
117	 eor	@Xw[$j],@Xw[$j],@Xw[($j+8)&15]
118	add	$e,$e,$t2		// e+=rot(a,5)
119	orr	$t0,$t0,$t1
120	ror	$b,$b,#2
121	 eor	@Xw[$j],@Xw[$j],@Xw[($j+13)&15]
122	add	$d,$d,@Xw[($i+1)&15]	// future e+=X[i]
123	add	$e,$e,$t0		// e+=F(b,c,d)
124	 ror	@Xw[$j],@Xw[$j],#31
125___
126}
127
128sub BODY_20_39 {
129my ($i,$a,$b,$c,$d,$e)=@_;
130my $j=($i+2)&15;
131
132$code.=<<___ if ($i==39);
133	movz	$K,#0xbcdc
134	movk	$K,#0x8f1b,lsl#16
135___
136$code.=<<___ if ($i<78);
137	 eor	@Xw[$j],@Xw[$j],@Xw[($j+2)&15]
138	eor	$t0,$d,$b
139	ror	$t2,$a,#27
140	add	$d,$d,$K		// future e+=K
141	 eor	@Xw[$j],@Xw[$j],@Xw[($j+8)&15]
142	eor	$t0,$t0,$c
143	add	$e,$e,$t2		// e+=rot(a,5)
144	ror	$b,$b,#2
145	 eor	@Xw[$j],@Xw[$j],@Xw[($j+13)&15]
146	add	$d,$d,@Xw[($i+1)&15]	// future e+=X[i]
147	add	$e,$e,$t0		// e+=F(b,c,d)
148	 ror	@Xw[$j],@Xw[$j],#31
149___
150$code.=<<___ if ($i==78);
151	ldp	@Xw[1],@Xw[2],[$ctx]
152	eor	$t0,$d,$b
153	ror	$t2,$a,#27
154	add	$d,$d,$K		// future e+=K
155	eor	$t0,$t0,$c
156	add	$e,$e,$t2		// e+=rot(a,5)
157	ror	$b,$b,#2
158	add	$d,$d,@Xw[($i+1)&15]	// future e+=X[i]
159	add	$e,$e,$t0		// e+=F(b,c,d)
160___
161$code.=<<___ if ($i==79);
162	ldp	@Xw[3],@Xw[4],[$ctx,#8]
163	eor	$t0,$d,$b
164	ror	$t2,$a,#27
165	eor	$t0,$t0,$c
166	add	$e,$e,$t2		// e+=rot(a,5)
167	ror	$b,$b,#2
168	ldr	@Xw[5],[$ctx,#16]
169	add	$e,$e,$t0		// e+=F(b,c,d)
170___
171}
172
173$code.=<<___;
174#include <openssl/arm_arch.h>
175
176.text
177
178.extern	OPENSSL_armcap_P
179.globl	sha1_block_data_order
180.type	sha1_block_data_order,%function
181.align	6
182sha1_block_data_order:
183	adrp	x16,:pg_hi21:OPENSSL_armcap_P
184	add	x16,x16,:lo12:OPENSSL_armcap_P
185	ldr	w16,[x16]
186	tst	w16,#ARMV8_SHA1
187	b.ne	.Lv8_entry
188
189	stp	x29,x30,[sp,#-96]!
190	add	x29,sp,#0
191	stp	x19,x20,[sp,#16]
192	stp	x21,x22,[sp,#32]
193	stp	x23,x24,[sp,#48]
194	stp	x25,x26,[sp,#64]
195	stp	x27,x28,[sp,#80]
196
197	ldp	$A,$B,[$ctx]
198	ldp	$C,$D,[$ctx,#8]
199	ldr	$E,[$ctx,#16]
200
201.Loop:
202	ldr	@Xx[0],[$inp],#64
203	movz	$K,#0x7999
204	sub	$num,$num,#1
205	movk	$K,#0x5a82,lsl#16
206#ifdef	__ARMEB__
207	ror	$Xx[0],@Xx[0],#32
208#else
209	rev32	@Xx[0],@Xx[0]
210#endif
211	add	$E,$E,$K		// warm it up
212	add	$E,$E,@Xw[0]
213___
214for($i=0;$i<20;$i++)	{ &BODY_00_19($i,@V); unshift(@V,pop(@V)); }
215for(;$i<40;$i++)	{ &BODY_20_39($i,@V); unshift(@V,pop(@V)); }
216for(;$i<60;$i++)	{ &BODY_40_59($i,@V); unshift(@V,pop(@V)); }
217for(;$i<80;$i++)	{ &BODY_20_39($i,@V); unshift(@V,pop(@V)); }
218$code.=<<___;
219	add	$B,$B,@Xw[2]
220	add	$C,$C,@Xw[3]
221	add	$A,$A,@Xw[1]
222	add	$D,$D,@Xw[4]
223	add	$E,$E,@Xw[5]
224	stp	$A,$B,[$ctx]
225	stp	$C,$D,[$ctx,#8]
226	str	$E,[$ctx,#16]
227	cbnz	$num,.Loop
228
229	ldp	x19,x20,[sp,#16]
230	ldp	x21,x22,[sp,#32]
231	ldp	x23,x24,[sp,#48]
232	ldp	x25,x26,[sp,#64]
233	ldp	x27,x28,[sp,#80]
234	ldr	x29,[sp],#96
235	ret
236.size	sha1_block_data_order,.-sha1_block_data_order
237___
238{{{
239my ($ABCD,$E,$E0,$E1)=map("v$_.16b",(0..3));
240my @MSG=map("v$_.16b",(4..7));
241my @Kxx=map("v$_.4s",(16..19));
242my ($W0,$W1)=("v20.4s","v21.4s");
243my $ABCD_SAVE="v22.16b";
244
245$code.=<<___;
246.type	sha1_block_armv8,%function
247.align	6
248sha1_block_armv8:
249.Lv8_entry:
250	stp	x29,x30,[sp,#-16]!
251	add	x29,sp,#0
252
253	adrp	x4,:pg_hi21:.Lconst
254	add	x4,x4,:lo12:.Lconst
255	eor	$E,$E,$E
256	ld1.32	{$ABCD},[$ctx],#16
257	ld1.32	{$E}[0],[$ctx]
258	sub	$ctx,$ctx,#16
259	ld1.32	{@Kxx[0]-@Kxx[3]},[x4]
260
261.Loop_hw:
262	ld1	{@MSG[0]-@MSG[3]},[$inp],#64
263	sub	$num,$num,#1
264	rev32	@MSG[0],@MSG[0]
265	rev32	@MSG[1],@MSG[1]
266
267	add.i32	$W0,@Kxx[0],@MSG[0]
268	rev32	@MSG[2],@MSG[2]
269	orr	$ABCD_SAVE,$ABCD,$ABCD	// offload
270
271	add.i32	$W1,@Kxx[0],@MSG[1]
272	rev32	@MSG[3],@MSG[3]
273	sha1h	$E1,$ABCD
274	sha1c	$ABCD,$E,$W0		// 0
275	add.i32	$W0,@Kxx[$j],@MSG[2]
276	sha1su0	@MSG[0],@MSG[1],@MSG[2]
277___
278for ($j=0,$i=1;$i<20-3;$i++) {
279my $f=("c","p","m","p")[$i/5];
280$code.=<<___;
281	sha1h	$E0,$ABCD		// $i
282	sha1$f	$ABCD,$E1,$W1
283	add.i32	$W1,@Kxx[$j],@MSG[3]
284	sha1su1	@MSG[0],@MSG[3]
285___
286$code.=<<___ if ($i<20-4);
287	sha1su0	@MSG[1],@MSG[2],@MSG[3]
288___
289	($E0,$E1)=($E1,$E0);		($W0,$W1)=($W1,$W0);
290	push(@MSG,shift(@MSG));		$j++ if ((($i+3)%5)==0);
291}
292$code.=<<___;
293	sha1h	$E0,$ABCD		// $i
294	sha1p	$ABCD,$E1,$W1
295	add.i32	$W1,@Kxx[$j],@MSG[3]
296
297	sha1h	$E1,$ABCD		// 18
298	sha1p	$ABCD,$E0,$W0
299
300	sha1h	$E0,$ABCD		// 19
301	sha1p	$ABCD,$E1,$W1
302
303	add.i32	$E,$E,$E0
304	add.i32	$ABCD,$ABCD,$ABCD_SAVE
305
306	cbnz	$num,.Loop_hw
307
308	st1.32	{$ABCD},[$ctx],#16
309	st1.32	{$E}[0],[$ctx]
310
311	ldr	x29,[sp],#16
312	ret
313.size	sha1_block_armv8,.-sha1_block_armv8
314.section .rodata
315.align	6
316.Lconst:
317.long	0x5a827999,0x5a827999,0x5a827999,0x5a827999	//K_00_19
318.long	0x6ed9eba1,0x6ed9eba1,0x6ed9eba1,0x6ed9eba1	//K_20_39
319.long	0x8f1bbcdc,0x8f1bbcdc,0x8f1bbcdc,0x8f1bbcdc	//K_40_59
320.long	0xca62c1d6,0xca62c1d6,0xca62c1d6,0xca62c1d6	//K_60_79
321.asciz	"SHA1 block transform for ARMv8, CRYPTOGAMS by <appro\@openssl.org>"
322.align	2
323.comm	OPENSSL_armcap_P,4,4
324.hidden	OPENSSL_armcap_P
325___
326}}}
327
328{   my	%opcode = (
329	"sha1c"		=> 0x5e000000,	"sha1p"		=> 0x5e001000,
330	"sha1m"		=> 0x5e002000,	"sha1su0"	=> 0x5e003000,
331	"sha1h"		=> 0x5e280800,	"sha1su1"	=> 0x5e281800	);
332
333    sub unsha1 {
334	my ($mnemonic,$arg)=@_;
335
336	$arg =~ m/[qv]([0-9]+)[^,]*,\s*[qv]([0-9]+)[^,]*(?:,\s*[qv]([0-9]+))?/o
337	&&
338	sprintf ".inst\t0x%08x\t//%s %s",
339			$opcode{$mnemonic}|$1|($2<<5)|($3<<16),
340			$mnemonic,$arg;
341    }
342}
343
344foreach(split("\n",$code)) {
345
346	s/\`([^\`]*)\`/eval($1)/geo;
347
348	s/\b(sha1\w+)\s+([qv].*)/unsha1($1,$2)/geo;
349
350	s/\.\w?32\b//o		and s/\.16b/\.4s/go;
351	m/(ld|st)1[^\[]+\[0\]/o	and s/\.4s/\.s/go;
352
353	print $_,"\n";
354}
355
356close STDOUT;
357