xref: /openssl/crypto/sha/asm/sha1-thumb.pl (revision 33388b44)
1#! /usr/bin/env perl
2# Copyright 2007-2020 The OpenSSL Project Authors. All Rights Reserved.
3#
4# Licensed under the Apache License 2.0 (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_block for Thumb.
18#
19# January 2007.
20#
21# The code does not present direct interest to OpenSSL, because of low
22# performance. Its purpose is to establish _size_ benchmark. Pretty
23# useless one I must say, because 30% or 88 bytes larger ARMv4 code
24# [available on demand] is almost _twice_ as fast. It should also be
25# noted that in-lining of .Lcommon and .Lrotate improves performance
26# by over 40%, while code increases by only 10% or 32 bytes. But once
27# again, the goal was to establish _size_ benchmark, not performance.
28
29$output=pop and open STDOUT,">$output";
30
31$inline=0;
32#$cheat_on_binutils=1;
33
34$t0="r0";
35$t1="r1";
36$t2="r2";
37$a="r3";
38$b="r4";
39$c="r5";
40$d="r6";
41$e="r7";
42$K="r8";	# "upper" registers can be used in add/sub and mov insns
43$ctx="r9";
44$inp="r10";
45$len="r11";
46$Xi="r12";
47
48sub common {
49<<___;
50	sub	$t0,#4
51	ldr	$t1,[$t0]
52	add	$e,$K			@ E+=K_xx_xx
53	lsl	$t2,$a,#5
54	add	$t2,$e
55	lsr	$e,$a,#27
56	add	$t2,$e			@ E+=ROR(A,27)
57	add	$t2,$t1			@ E+=X[i]
58___
59}
60sub rotate {
61<<___;
62	mov	$e,$d			@ E=D
63	mov	$d,$c			@ D=C
64	lsl	$c,$b,#30
65	lsr	$b,$b,#2
66	orr	$c,$b			@ C=ROR(B,2)
67	mov	$b,$a			@ B=A
68	add	$a,$t2,$t1		@ A=E+F_xx_xx(B,C,D)
69___
70}
71
72sub BODY_00_19 {
73$code.=$inline?&common():"\tbl	.Lcommon\n";
74$code.=<<___;
75	mov	$t1,$c
76	eor	$t1,$d
77	and	$t1,$b
78	eor	$t1,$d			@ F_00_19(B,C,D)
79___
80$code.=$inline?&rotate():"\tbl	.Lrotate\n";
81}
82
83sub BODY_20_39 {
84$code.=$inline?&common():"\tbl	.Lcommon\n";
85$code.=<<___;
86	mov	$t1,$b
87	eor	$t1,$c
88	eor	$t1,$d			@ F_20_39(B,C,D)
89___
90$code.=$inline?&rotate():"\tbl	.Lrotate\n";
91}
92
93sub BODY_40_59 {
94$code.=$inline?&common():"\tbl	.Lcommon\n";
95$code.=<<___;
96	mov	$t1,$b
97	and	$t1,$c
98	mov	$e,$b
99	orr	$e,$c
100	and	$e,$d
101	orr	$t1,$e			@ F_40_59(B,C,D)
102___
103$code.=$inline?&rotate():"\tbl	.Lrotate\n";
104}
105
106$code=<<___;
107.text
108.code	16
109
110.global	sha1_block_data_order
111.type	sha1_block_data_order,%function
112
113.align	2
114sha1_block_data_order:
115___
116if ($cheat_on_binutils) {
117$code.=<<___;
118.code	32
119	add	r3,pc,#1
120	bx	r3			@ switch to Thumb ISA
121.code	16
122___
123}
124$code.=<<___;
125	push	{r4-r7}
126	mov	r3,r8
127	mov	r4,r9
128	mov	r5,r10
129	mov	r6,r11
130	mov	r7,r12
131	push	{r3-r7,lr}
132	lsl	r2,#6
133	mov	$ctx,r0			@ save context
134	mov	$inp,r1			@ save inp
135	mov	$len,r2			@ save len
136	add	$len,$inp		@ $len to point at inp end
137
138.Lloop:
139	mov	$Xi,sp
140	mov	$t2,sp
141	sub	$t2,#16*4		@ [3]
142.LXload:
143	ldrb	$a,[$t1,#0]		@ $t1 is r1 and holds inp
144	ldrb	$b,[$t1,#1]
145	ldrb	$c,[$t1,#2]
146	ldrb	$d,[$t1,#3]
147	lsl	$a,#24
148	lsl	$b,#16
149	lsl	$c,#8
150	orr	$a,$b
151	orr	$a,$c
152	orr	$a,$d
153	add	$t1,#4
154	push	{$a}
155	cmp	sp,$t2
156	bne	.LXload			@ [+14*16]
157
158	mov	$inp,$t1		@ update $inp
159	sub	$t2,#32*4
160	sub	$t2,#32*4
161	mov	$e,#31			@ [+4]
162.LXupdate:
163	ldr	$a,[sp,#15*4]
164	ldr	$b,[sp,#13*4]
165	ldr	$c,[sp,#7*4]
166	ldr	$d,[sp,#2*4]
167	eor	$a,$b
168	eor	$a,$c
169	eor	$a,$d
170	ror	$a,$e
171	push	{$a}
172	cmp	sp,$t2
173	bne	.LXupdate		@ [+(11+1)*64]
174
175	ldmia	$t0!,{$a,$b,$c,$d,$e}	@ $t0 is r0 and holds ctx
176	mov	$t0,$Xi
177
178	ldr	$t2,.LK_00_19
179	mov	$t1,$t0
180	sub	$t1,#20*4
181	mov	$Xi,$t1
182	mov	$K,$t2			@ [+7+4]
183.L_00_19:
184___
185	&BODY_00_19();
186$code.=<<___;
187	cmp	$Xi,$t0
188	bne	.L_00_19		@ [+(2+9+4+2+8+2)*20]
189
190	ldr	$t2,.LK_20_39
191	mov	$t1,$t0
192	sub	$t1,#20*4
193	mov	$Xi,$t1
194	mov	$K,$t2			@ [+5]
195.L_20_39_or_60_79:
196___
197	&BODY_20_39();
198$code.=<<___;
199	cmp	$Xi,$t0
200	bne	.L_20_39_or_60_79	@ [+(2+9+3+2+8+2)*20*2]
201	cmp	sp,$t0
202	beq	.Ldone			@ [+2]
203
204	ldr	$t2,.LK_40_59
205	mov	$t1,$t0
206	sub	$t1,#20*4
207	mov	$Xi,$t1
208	mov	$K,$t2			@ [+5]
209.L_40_59:
210___
211	&BODY_40_59();
212$code.=<<___;
213	cmp	$Xi,$t0
214	bne	.L_40_59		@ [+(2+9+6+2+8+2)*20]
215
216	ldr	$t2,.LK_60_79
217	mov	$Xi,sp
218	mov	$K,$t2
219	b	.L_20_39_or_60_79	@ [+4]
220.Ldone:
221	mov	$t0,$ctx
222	ldr	$t1,[$t0,#0]
223	ldr	$t2,[$t0,#4]
224	add	$a,$t1
225	ldr	$t1,[$t0,#8]
226	add	$b,$t2
227	ldr	$t2,[$t0,#12]
228	add	$c,$t1
229	ldr	$t1,[$t0,#16]
230	add	$d,$t2
231	add	$e,$t1
232	stmia	$t0!,{$a,$b,$c,$d,$e}	@ [+20]
233
234	add	sp,#80*4		@ deallocate stack frame
235	mov	$t0,$ctx		@ restore ctx
236	mov	$t1,$inp		@ restore inp
237	cmp	$t1,$len
238	beq	.Lexit
239	b	.Lloop			@ [+6] total 3212 cycles
240.Lexit:
241	pop	{r2-r7}
242	mov	r8,r2
243	mov	r9,r3
244	mov	r10,r4
245	mov	r11,r5
246	mov	r12,r6
247	mov	lr,r7
248	pop	{r4-r7}
249	bx	lr
250.align	2
251___
252$code.=".Lcommon:\n".&common()."\tmov	pc,lr\n" if (!$inline);
253$code.=".Lrotate:\n".&rotate()."\tmov	pc,lr\n" if (!$inline);
254$code.=<<___;
255.align	2
256.LK_00_19:	.word	0x5a827999
257.LK_20_39:	.word	0x6ed9eba1
258.LK_40_59:	.word	0x8f1bbcdc
259.LK_60_79:	.word	0xca62c1d6
260.size	sha1_block_data_order,.-sha1_block_data_order
261.asciz	"SHA1 block transform for Thumb, CRYPTOGAMS by <appro\@openssl.org>"
262___
263
264print $code;
265close STDOUT or die "error closing STDOUT: $!"; # enforce flush
266