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#
30# (*)	Software results are presented mostly for reference purposes.
31# (**)	Keep in mind that Denver relies on binary translation, which
32#	optimizes compiler output at run-time.
33
34$flavour = shift;
35$output  = shift;
36
37$0 =~ m/(.*[\/\\])[^\/\\]+$/; $dir=$1;
38( $xlate="${dir}arm-xlate.pl" and -f $xlate ) or
39( $xlate="${dir}../../../perlasm/arm-xlate.pl" and -f $xlate) or
40die "can't locate arm-xlate.pl";
41
42open OUT,"| \"$^X\" $xlate $flavour $output";
43*STDOUT=*OUT;
44
45($ctx,$inp,$num)=("x0","x1","x2");
46@Xw=map("w$_",(3..17,19));
47@Xx=map("x$_",(3..17,19));
48@V=($A,$B,$C,$D,$E)=map("w$_",(20..24));
49($t0,$t1,$t2,$K)=map("w$_",(25..28));
50
51
52sub BODY_00_19 {
53my ($i,$a,$b,$c,$d,$e)=@_;
54my $j=($i+2)&15;
55
56$code.=<<___ if ($i<15 && !($i&1));
57	lsr	@Xx[$i+1],@Xx[$i],#32
58___
59$code.=<<___ if ($i<14 && !($i&1));
60	ldr	@Xx[$i+2],[$inp,#`($i+2)*4-64`]
61___
62$code.=<<___ if ($i<14 && ($i&1));
63#ifdef	__ARMEB__
64	ror	@Xx[$i+1],@Xx[$i+1],#32
65#else
66	rev32	@Xx[$i+1],@Xx[$i+1]
67#endif
68___
69$code.=<<___ if ($i<14);
70	bic	$t0,$d,$b
71	and	$t1,$c,$b
72	ror	$t2,$a,#27
73	add	$d,$d,$K		// future e+=K
74	orr	$t0,$t0,$t1
75	add	$e,$e,$t2		// e+=rot(a,5)
76	ror	$b,$b,#2
77	add	$d,$d,@Xw[($i+1)&15]	// future e+=X[i]
78	add	$e,$e,$t0		// e+=F(b,c,d)
79___
80$code.=<<___ if ($i==19);
81	movz	$K,#0xeba1
82	movk	$K,#0x6ed9,lsl#16
83___
84$code.=<<___ if ($i>=14);
85	 eor	@Xw[$j],@Xw[$j],@Xw[($j+2)&15]
86	bic	$t0,$d,$b
87	and	$t1,$c,$b
88	ror	$t2,$a,#27
89	 eor	@Xw[$j],@Xw[$j],@Xw[($j+8)&15]
90	add	$d,$d,$K		// future e+=K
91	orr	$t0,$t0,$t1
92	add	$e,$e,$t2		// e+=rot(a,5)
93	 eor	@Xw[$j],@Xw[$j],@Xw[($j+13)&15]
94	ror	$b,$b,#2
95	add	$d,$d,@Xw[($i+1)&15]	// future e+=X[i]
96	add	$e,$e,$t0		// e+=F(b,c,d)
97	 ror	@Xw[$j],@Xw[$j],#31
98___
99}
100
101sub BODY_40_59 {
102my ($i,$a,$b,$c,$d,$e)=@_;
103my $j=($i+2)&15;
104
105$code.=<<___ if ($i==59);
106	movz	$K,#0xc1d6
107	movk	$K,#0xca62,lsl#16
108___
109$code.=<<___;
110	orr	$t0,$b,$c
111	and	$t1,$b,$c
112	 eor	@Xw[$j],@Xw[$j],@Xw[($j+2)&15]
113	ror	$t2,$a,#27
114	and	$t0,$t0,$d
115	add	$d,$d,$K		// future e+=K
116	 eor	@Xw[$j],@Xw[$j],@Xw[($j+8)&15]
117	add	$e,$e,$t2		// e+=rot(a,5)
118	orr	$t0,$t0,$t1
119	ror	$b,$b,#2
120	 eor	@Xw[$j],@Xw[$j],@Xw[($j+13)&15]
121	add	$d,$d,@Xw[($i+1)&15]	// future e+=X[i]
122	add	$e,$e,$t0		// e+=F(b,c,d)
123	 ror	@Xw[$j],@Xw[$j],#31
124___
125}
126
127sub BODY_20_39 {
128my ($i,$a,$b,$c,$d,$e)=@_;
129my $j=($i+2)&15;
130
131$code.=<<___ if ($i==39);
132	movz	$K,#0xbcdc
133	movk	$K,#0x8f1b,lsl#16
134___
135$code.=<<___ if ($i<78);
136	 eor	@Xw[$j],@Xw[$j],@Xw[($j+2)&15]
137	eor	$t0,$d,$b
138	ror	$t2,$a,#27
139	add	$d,$d,$K		// future e+=K
140	 eor	@Xw[$j],@Xw[$j],@Xw[($j+8)&15]
141	eor	$t0,$t0,$c
142	add	$e,$e,$t2		// e+=rot(a,5)
143	ror	$b,$b,#2
144	 eor	@Xw[$j],@Xw[$j],@Xw[($j+13)&15]
145	add	$d,$d,@Xw[($i+1)&15]	// future e+=X[i]
146	add	$e,$e,$t0		// e+=F(b,c,d)
147	 ror	@Xw[$j],@Xw[$j],#31
148___
149$code.=<<___ if ($i==78);
150	ldp	@Xw[1],@Xw[2],[$ctx]
151	eor	$t0,$d,$b
152	ror	$t2,$a,#27
153	add	$d,$d,$K		// future e+=K
154	eor	$t0,$t0,$c
155	add	$e,$e,$t2		// e+=rot(a,5)
156	ror	$b,$b,#2
157	add	$d,$d,@Xw[($i+1)&15]	// future e+=X[i]
158	add	$e,$e,$t0		// e+=F(b,c,d)
159___
160$code.=<<___ if ($i==79);
161	ldp	@Xw[3],@Xw[4],[$ctx,#8]
162	eor	$t0,$d,$b
163	ror	$t2,$a,#27
164	eor	$t0,$t0,$c
165	add	$e,$e,$t2		// e+=rot(a,5)
166	ror	$b,$b,#2
167	ldr	@Xw[5],[$ctx,#16]
168	add	$e,$e,$t0		// e+=F(b,c,d)
169___
170}
171
172$code.=<<___;
173#include <openssl/arm_arch.h>
174
175.text
176
177.extern	OPENSSL_armcap_P
178.globl	sha1_block_data_order
179.type	sha1_block_data_order,%function
180.align	6
181sha1_block_data_order:
182#ifdef	__ILP32__
183	ldrsw	x16,.LOPENSSL_armcap_P
184#else
185	ldr	x16,.LOPENSSL_armcap_P
186#endif
187	adr	x17,.LOPENSSL_armcap_P
188	add	x16,x16,x17
189	ldr	w16,[x16]
190	tst	w16,#ARMV8_SHA1
191	b.ne	.Lv8_entry
192
193	stp	x29,x30,[sp,#-96]!
194	add	x29,sp,#0
195	stp	x19,x20,[sp,#16]
196	stp	x21,x22,[sp,#32]
197	stp	x23,x24,[sp,#48]
198	stp	x25,x26,[sp,#64]
199	stp	x27,x28,[sp,#80]
200
201	ldp	$A,$B,[$ctx]
202	ldp	$C,$D,[$ctx,#8]
203	ldr	$E,[$ctx,#16]
204
205.Loop:
206	ldr	@Xx[0],[$inp],#64
207	movz	$K,#0x7999
208	sub	$num,$num,#1
209	movk	$K,#0x5a82,lsl#16
210#ifdef	__ARMEB__
211	ror	$Xx[0],@Xx[0],#32
212#else
213	rev32	@Xx[0],@Xx[0]
214#endif
215	add	$E,$E,$K		// warm it up
216	add	$E,$E,@Xw[0]
217___
218for($i=0;$i<20;$i++)	{ &BODY_00_19($i,@V); unshift(@V,pop(@V)); }
219for(;$i<40;$i++)	{ &BODY_20_39($i,@V); unshift(@V,pop(@V)); }
220for(;$i<60;$i++)	{ &BODY_40_59($i,@V); unshift(@V,pop(@V)); }
221for(;$i<80;$i++)	{ &BODY_20_39($i,@V); unshift(@V,pop(@V)); }
222$code.=<<___;
223	add	$B,$B,@Xw[2]
224	add	$C,$C,@Xw[3]
225	add	$A,$A,@Xw[1]
226	add	$D,$D,@Xw[4]
227	add	$E,$E,@Xw[5]
228	stp	$A,$B,[$ctx]
229	stp	$C,$D,[$ctx,#8]
230	str	$E,[$ctx,#16]
231	cbnz	$num,.Loop
232
233	ldp	x19,x20,[sp,#16]
234	ldp	x21,x22,[sp,#32]
235	ldp	x23,x24,[sp,#48]
236	ldp	x25,x26,[sp,#64]
237	ldp	x27,x28,[sp,#80]
238	ldr	x29,[sp],#96
239	ret
240.size	sha1_block_data_order,.-sha1_block_data_order
241___
242{{{
243my ($ABCD,$E,$E0,$E1)=map("v$_.16b",(0..3));
244my @MSG=map("v$_.16b",(4..7));
245my @Kxx=map("v$_.4s",(16..19));
246my ($W0,$W1)=("v20.4s","v21.4s");
247my $ABCD_SAVE="v22.16b";
248
249$code.=<<___;
250.type	sha1_block_armv8,%function
251.align	6
252sha1_block_armv8:
253.Lv8_entry:
254	stp	x29,x30,[sp,#-16]!
255	add	x29,sp,#0
256
257	adr	x4,.Lconst
258	eor	$E,$E,$E
259	ld1.32	{$ABCD},[$ctx],#16
260	ld1.32	{$E}[0],[$ctx]
261	sub	$ctx,$ctx,#16
262	ld1.32	{@Kxx[0]-@Kxx[3]},[x4]
263
264.Loop_hw:
265	ld1	{@MSG[0]-@MSG[3]},[$inp],#64
266	sub	$num,$num,#1
267	rev32	@MSG[0],@MSG[0]
268	rev32	@MSG[1],@MSG[1]
269
270	add.i32	$W0,@Kxx[0],@MSG[0]
271	rev32	@MSG[2],@MSG[2]
272	orr	$ABCD_SAVE,$ABCD,$ABCD	// offload
273
274	add.i32	$W1,@Kxx[0],@MSG[1]
275	rev32	@MSG[3],@MSG[3]
276	sha1h	$E1,$ABCD
277	sha1c	$ABCD,$E,$W0		// 0
278	add.i32	$W0,@Kxx[$j],@MSG[2]
279	sha1su0	@MSG[0],@MSG[1],@MSG[2]
280___
281for ($j=0,$i=1;$i<20-3;$i++) {
282my $f=("c","p","m","p")[$i/5];
283$code.=<<___;
284	sha1h	$E0,$ABCD		// $i
285	sha1$f	$ABCD,$E1,$W1
286	add.i32	$W1,@Kxx[$j],@MSG[3]
287	sha1su1	@MSG[0],@MSG[3]
288___
289$code.=<<___ if ($i<20-4);
290	sha1su0	@MSG[1],@MSG[2],@MSG[3]
291___
292	($E0,$E1)=($E1,$E0);		($W0,$W1)=($W1,$W0);
293	push(@MSG,shift(@MSG));		$j++ if ((($i+3)%5)==0);
294}
295$code.=<<___;
296	sha1h	$E0,$ABCD		// $i
297	sha1p	$ABCD,$E1,$W1
298	add.i32	$W1,@Kxx[$j],@MSG[3]
299
300	sha1h	$E1,$ABCD		// 18
301	sha1p	$ABCD,$E0,$W0
302
303	sha1h	$E0,$ABCD		// 19
304	sha1p	$ABCD,$E1,$W1
305
306	add.i32	$E,$E,$E0
307	add.i32	$ABCD,$ABCD,$ABCD_SAVE
308
309	cbnz	$num,.Loop_hw
310
311	st1.32	{$ABCD},[$ctx],#16
312	st1.32	{$E}[0],[$ctx]
313
314	ldr	x29,[sp],#16
315	ret
316.size	sha1_block_armv8,.-sha1_block_armv8
317.align	6
318.Lconst:
319.long	0x5a827999,0x5a827999,0x5a827999,0x5a827999	//K_00_19
320.long	0x6ed9eba1,0x6ed9eba1,0x6ed9eba1,0x6ed9eba1	//K_20_39
321.long	0x8f1bbcdc,0x8f1bbcdc,0x8f1bbcdc,0x8f1bbcdc	//K_40_59
322.long	0xca62c1d6,0xca62c1d6,0xca62c1d6,0xca62c1d6	//K_60_79
323.LOPENSSL_armcap_P:
324#ifdef	__ILP32__
325.long	OPENSSL_armcap_P-.
326#else
327.quad	OPENSSL_armcap_P-.
328#endif
329.asciz	"SHA1 block transform for ARMv8, CRYPTOGAMS by <appro\@openssl.org>"
330.align	2
331.comm	OPENSSL_armcap_P,4,4
332___
333}}}
334
335{   my	%opcode = (
336	"sha1c"		=> 0x5e000000,	"sha1p"		=> 0x5e001000,
337	"sha1m"		=> 0x5e002000,	"sha1su0"	=> 0x5e003000,
338	"sha1h"		=> 0x5e280800,	"sha1su1"	=> 0x5e281800	);
339
340    sub unsha1 {
341	my ($mnemonic,$arg)=@_;
342
343	$arg =~ m/[qv]([0-9]+)[^,]*,\s*[qv]([0-9]+)[^,]*(?:,\s*[qv]([0-9]+))?/o
344	&&
345	sprintf ".inst\t0x%08x\t//%s %s",
346			$opcode{$mnemonic}|$1|($2<<5)|($3<<16),
347			$mnemonic,$arg;
348    }
349}
350
351foreach(split("\n",$code)) {
352
353	s/\`([^\`]*)\`/eval($1)/geo;
354
355	s/\b(sha1\w+)\s+([qv].*)/unsha1($1,$2)/geo;
356
357	s/\.\w?32\b//o		and s/\.16b/\.4s/go;
358	m/(ld|st)1[^\[]+\[0\]/o	and s/\.4s/\.s/go;
359
360	print $_,"\n";
361}
362
363close STDOUT;
364