xref: /aosp_15_r20/external/boringssl/src/crypto/fipsmodule/sha/asm/sha512-armv4.pl (revision 8fb009dc861624b67b6cdb62ea21f0f22d0c584b)
1#! /usr/bin/env perl
2# Copyright 2007-2016 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 <[email protected]> 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# Permission to use under GPL terms is granted.
17# ====================================================================
18
19# SHA512 block procedure for ARMv4. September 2007.
20
21# This code is ~4.5 (four and a half) times faster than code generated
22# by gcc 3.4 and it spends ~72 clock cycles per byte [on single-issue
23# Xscale PXA250 core].
24#
25# July 2010.
26#
27# Rescheduling for dual-issue pipeline resulted in 6% improvement on
28# Cortex A8 core and ~40 cycles per processed byte.
29
30# February 2011.
31#
32# Profiler-assisted and platform-specific optimization resulted in 7%
33# improvement on Coxtex A8 core and ~38 cycles per byte.
34
35# March 2011.
36#
37# Add NEON implementation. On Cortex A8 it was measured to process
38# one byte in 23.3 cycles or ~60% faster than integer-only code.
39
40# August 2012.
41#
42# Improve NEON performance by 12% on Snapdragon S4. In absolute
43# terms it's 22.6 cycles per byte, which is disappointing result.
44# Technical writers asserted that 3-way S4 pipeline can sustain
45# multiple NEON instructions per cycle, but dual NEON issue could
46# not be observed, see http://www.openssl.org/~appro/Snapdragon-S4.html
47# for further details. On side note Cortex-A15 processes one byte in
48# 16 cycles.
49
50# Byte order [in]dependence. =========================================
51#
52# Originally caller was expected to maintain specific *dword* order in
53# h[0-7], namely with most significant dword at *lower* address, which
54# was reflected in below two parameters as 0 and 4. Now caller is
55# expected to maintain native byte order for whole 64-bit values.
56$hi="HI";
57$lo="LO";
58# ====================================================================
59
60$flavour = shift;
61if ($flavour=~/\w[\w\-]*\.\w+$/) { $output=$flavour; undef $flavour; }
62else { while (($output=shift) && ($output!~/\w[\w\-]*\.\w+$/)) {} }
63
64if ($flavour && $flavour ne "void") {
65    $0 =~ m/(.*[\/\\])[^\/\\]+$/; $dir=$1;
66    ( $xlate="${dir}arm-xlate.pl" and -f $xlate ) or
67    ( $xlate="${dir}../../../perlasm/arm-xlate.pl" and -f $xlate) or
68    die "can't locate arm-xlate.pl";
69
70    open OUT,"| \"$^X\" \"$xlate\" $flavour \"$output\"";
71    *STDOUT=*OUT;
72} else {
73    open OUT,">$output";
74    *STDOUT=*OUT;
75}
76
77$ctx="r0";	# parameter block
78$inp="r1";
79$len="r2";
80
81$Tlo="r3";
82$Thi="r4";
83$Alo="r5";
84$Ahi="r6";
85$Elo="r7";
86$Ehi="r8";
87$t0="r9";
88$t1="r10";
89$t2="r11";
90$t3="r12";
91############	r13 is stack pointer
92$Ktbl="r14";
93############	r15 is program counter
94
95$Aoff=8*0;
96$Boff=8*1;
97$Coff=8*2;
98$Doff=8*3;
99$Eoff=8*4;
100$Foff=8*5;
101$Goff=8*6;
102$Hoff=8*7;
103$Xoff=8*8;
104
105sub BODY_00_15() {
106my $magic = shift;
107$code.=<<___;
108	@ Sigma1(x)	(ROTR((x),14) ^ ROTR((x),18)  ^ ROTR((x),41))
109	@ LO		lo>>14^hi<<18 ^ lo>>18^hi<<14 ^ hi>>9^lo<<23
110	@ HI		hi>>14^lo<<18 ^ hi>>18^lo<<14 ^ lo>>9^hi<<23
111	mov	$t0,$Elo,lsr#14
112	str	$Tlo,[sp,#$Xoff+0]
113	mov	$t1,$Ehi,lsr#14
114	str	$Thi,[sp,#$Xoff+4]
115	eor	$t0,$t0,$Ehi,lsl#18
116	ldr	$t2,[sp,#$Hoff+0]	@ h.lo
117	eor	$t1,$t1,$Elo,lsl#18
118	ldr	$t3,[sp,#$Hoff+4]	@ h.hi
119	eor	$t0,$t0,$Elo,lsr#18
120	eor	$t1,$t1,$Ehi,lsr#18
121	eor	$t0,$t0,$Ehi,lsl#14
122	eor	$t1,$t1,$Elo,lsl#14
123	eor	$t0,$t0,$Ehi,lsr#9
124	eor	$t1,$t1,$Elo,lsr#9
125	eor	$t0,$t0,$Elo,lsl#23
126	eor	$t1,$t1,$Ehi,lsl#23	@ Sigma1(e)
127	adds	$Tlo,$Tlo,$t0
128	ldr	$t0,[sp,#$Foff+0]	@ f.lo
129	adc	$Thi,$Thi,$t1		@ T += Sigma1(e)
130	ldr	$t1,[sp,#$Foff+4]	@ f.hi
131	adds	$Tlo,$Tlo,$t2
132	ldr	$t2,[sp,#$Goff+0]	@ g.lo
133	adc	$Thi,$Thi,$t3		@ T += h
134	ldr	$t3,[sp,#$Goff+4]	@ g.hi
135
136	eor	$t0,$t0,$t2
137	str	$Elo,[sp,#$Eoff+0]
138	eor	$t1,$t1,$t3
139	str	$Ehi,[sp,#$Eoff+4]
140	and	$t0,$t0,$Elo
141	str	$Alo,[sp,#$Aoff+0]
142	and	$t1,$t1,$Ehi
143	str	$Ahi,[sp,#$Aoff+4]
144	eor	$t0,$t0,$t2
145	ldr	$t2,[$Ktbl,#$lo]	@ K[i].lo
146	eor	$t1,$t1,$t3		@ Ch(e,f,g)
147	ldr	$t3,[$Ktbl,#$hi]	@ K[i].hi
148
149	adds	$Tlo,$Tlo,$t0
150	ldr	$Elo,[sp,#$Doff+0]	@ d.lo
151	adc	$Thi,$Thi,$t1		@ T += Ch(e,f,g)
152	ldr	$Ehi,[sp,#$Doff+4]	@ d.hi
153	adds	$Tlo,$Tlo,$t2
154	and	$t0,$t2,#0xff
155	adc	$Thi,$Thi,$t3		@ T += K[i]
156	adds	$Elo,$Elo,$Tlo
157	ldr	$t2,[sp,#$Boff+0]	@ b.lo
158	adc	$Ehi,$Ehi,$Thi		@ d += T
159	teq	$t0,#$magic
160
161	ldr	$t3,[sp,#$Coff+0]	@ c.lo
162#if __ARM_ARCH>=7
163	it	eq			@ Thumb2 thing, sanity check in ARM
164#endif
165	orreq	$Ktbl,$Ktbl,#1
166	@ Sigma0(x)	(ROTR((x),28) ^ ROTR((x),34) ^ ROTR((x),39))
167	@ LO		lo>>28^hi<<4  ^ hi>>2^lo<<30 ^ hi>>7^lo<<25
168	@ HI		hi>>28^lo<<4  ^ lo>>2^hi<<30 ^ lo>>7^hi<<25
169	mov	$t0,$Alo,lsr#28
170	mov	$t1,$Ahi,lsr#28
171	eor	$t0,$t0,$Ahi,lsl#4
172	eor	$t1,$t1,$Alo,lsl#4
173	eor	$t0,$t0,$Ahi,lsr#2
174	eor	$t1,$t1,$Alo,lsr#2
175	eor	$t0,$t0,$Alo,lsl#30
176	eor	$t1,$t1,$Ahi,lsl#30
177	eor	$t0,$t0,$Ahi,lsr#7
178	eor	$t1,$t1,$Alo,lsr#7
179	eor	$t0,$t0,$Alo,lsl#25
180	eor	$t1,$t1,$Ahi,lsl#25	@ Sigma0(a)
181	adds	$Tlo,$Tlo,$t0
182	and	$t0,$Alo,$t2
183	adc	$Thi,$Thi,$t1		@ T += Sigma0(a)
184
185	ldr	$t1,[sp,#$Boff+4]	@ b.hi
186	orr	$Alo,$Alo,$t2
187	ldr	$t2,[sp,#$Coff+4]	@ c.hi
188	and	$Alo,$Alo,$t3
189	and	$t3,$Ahi,$t1
190	orr	$Ahi,$Ahi,$t1
191	orr	$Alo,$Alo,$t0		@ Maj(a,b,c).lo
192	and	$Ahi,$Ahi,$t2
193	adds	$Alo,$Alo,$Tlo
194	orr	$Ahi,$Ahi,$t3		@ Maj(a,b,c).hi
195	sub	sp,sp,#8
196	adc	$Ahi,$Ahi,$Thi		@ h += T
197	tst	$Ktbl,#1
198	add	$Ktbl,$Ktbl,#8
199___
200}
201$code=<<___;
202#ifndef __KERNEL__
203# include <openssl/arm_arch.h>
204# define VFP_ABI_PUSH	vstmdb	sp!,{d8-d15}
205# define VFP_ABI_POP	vldmia	sp!,{d8-d15}
206#else
207# define __ARM_MAX_ARCH__ 7
208# define VFP_ABI_PUSH
209# define VFP_ABI_POP
210#endif
211
212@ Silence ARMv8 deprecated IT instruction warnings. This file is used by both
213@ ARMv7 and ARMv8 processors and does not use ARMv8 instructions.
214.arch  armv7-a
215
216#ifdef __ARMEL__
217# define LO 0
218# define HI 4
219# define WORD64(hi0,lo0,hi1,lo1)	.word	lo0,hi0, lo1,hi1
220#else
221# define HI 0
222# define LO 4
223# define WORD64(hi0,lo0,hi1,lo1)	.word	hi0,lo0, hi1,lo1
224#endif
225
226.text
227#if defined(__thumb2__)
228.syntax unified
229.thumb
230# define adrl adr
231#else
232.code	32
233#endif
234
235.type	K512,%object
236.align	5
237K512:
238WORD64(0x428a2f98,0xd728ae22, 0x71374491,0x23ef65cd)
239WORD64(0xb5c0fbcf,0xec4d3b2f, 0xe9b5dba5,0x8189dbbc)
240WORD64(0x3956c25b,0xf348b538, 0x59f111f1,0xb605d019)
241WORD64(0x923f82a4,0xaf194f9b, 0xab1c5ed5,0xda6d8118)
242WORD64(0xd807aa98,0xa3030242, 0x12835b01,0x45706fbe)
243WORD64(0x243185be,0x4ee4b28c, 0x550c7dc3,0xd5ffb4e2)
244WORD64(0x72be5d74,0xf27b896f, 0x80deb1fe,0x3b1696b1)
245WORD64(0x9bdc06a7,0x25c71235, 0xc19bf174,0xcf692694)
246WORD64(0xe49b69c1,0x9ef14ad2, 0xefbe4786,0x384f25e3)
247WORD64(0x0fc19dc6,0x8b8cd5b5, 0x240ca1cc,0x77ac9c65)
248WORD64(0x2de92c6f,0x592b0275, 0x4a7484aa,0x6ea6e483)
249WORD64(0x5cb0a9dc,0xbd41fbd4, 0x76f988da,0x831153b5)
250WORD64(0x983e5152,0xee66dfab, 0xa831c66d,0x2db43210)
251WORD64(0xb00327c8,0x98fb213f, 0xbf597fc7,0xbeef0ee4)
252WORD64(0xc6e00bf3,0x3da88fc2, 0xd5a79147,0x930aa725)
253WORD64(0x06ca6351,0xe003826f, 0x14292967,0x0a0e6e70)
254WORD64(0x27b70a85,0x46d22ffc, 0x2e1b2138,0x5c26c926)
255WORD64(0x4d2c6dfc,0x5ac42aed, 0x53380d13,0x9d95b3df)
256WORD64(0x650a7354,0x8baf63de, 0x766a0abb,0x3c77b2a8)
257WORD64(0x81c2c92e,0x47edaee6, 0x92722c85,0x1482353b)
258WORD64(0xa2bfe8a1,0x4cf10364, 0xa81a664b,0xbc423001)
259WORD64(0xc24b8b70,0xd0f89791, 0xc76c51a3,0x0654be30)
260WORD64(0xd192e819,0xd6ef5218, 0xd6990624,0x5565a910)
261WORD64(0xf40e3585,0x5771202a, 0x106aa070,0x32bbd1b8)
262WORD64(0x19a4c116,0xb8d2d0c8, 0x1e376c08,0x5141ab53)
263WORD64(0x2748774c,0xdf8eeb99, 0x34b0bcb5,0xe19b48a8)
264WORD64(0x391c0cb3,0xc5c95a63, 0x4ed8aa4a,0xe3418acb)
265WORD64(0x5b9cca4f,0x7763e373, 0x682e6ff3,0xd6b2b8a3)
266WORD64(0x748f82ee,0x5defb2fc, 0x78a5636f,0x43172f60)
267WORD64(0x84c87814,0xa1f0ab72, 0x8cc70208,0x1a6439ec)
268WORD64(0x90befffa,0x23631e28, 0xa4506ceb,0xde82bde9)
269WORD64(0xbef9a3f7,0xb2c67915, 0xc67178f2,0xe372532b)
270WORD64(0xca273ece,0xea26619c, 0xd186b8c7,0x21c0c207)
271WORD64(0xeada7dd6,0xcde0eb1e, 0xf57d4f7f,0xee6ed178)
272WORD64(0x06f067aa,0x72176fba, 0x0a637dc5,0xa2c898a6)
273WORD64(0x113f9804,0xbef90dae, 0x1b710b35,0x131c471b)
274WORD64(0x28db77f5,0x23047d84, 0x32caab7b,0x40c72493)
275WORD64(0x3c9ebe0a,0x15c9bebc, 0x431d67c4,0x9c100d4c)
276WORD64(0x4cc5d4be,0xcb3e42b6, 0x597f299c,0xfc657e2a)
277WORD64(0x5fcb6fab,0x3ad6faec, 0x6c44198c,0x4a475817)
278.size	K512,.-K512
279
280.global	sha512_block_data_order_nohw
281.type	sha512_block_data_order_nohw,%function
282sha512_block_data_order_nohw:
283	add	$len,$inp,$len,lsl#7	@ len to point at the end of inp
284	stmdb	sp!,{r4-r12,lr}
285	adr	$Ktbl,K512
286	sub	sp,sp,#9*8
287
288	ldr	$Elo,[$ctx,#$Eoff+$lo]
289	ldr	$Ehi,[$ctx,#$Eoff+$hi]
290	ldr	$t0, [$ctx,#$Goff+$lo]
291	ldr	$t1, [$ctx,#$Goff+$hi]
292	ldr	$t2, [$ctx,#$Hoff+$lo]
293	ldr	$t3, [$ctx,#$Hoff+$hi]
294.Loop:
295	str	$t0, [sp,#$Goff+0]
296	str	$t1, [sp,#$Goff+4]
297	str	$t2, [sp,#$Hoff+0]
298	str	$t3, [sp,#$Hoff+4]
299	ldr	$Alo,[$ctx,#$Aoff+$lo]
300	ldr	$Ahi,[$ctx,#$Aoff+$hi]
301	ldr	$Tlo,[$ctx,#$Boff+$lo]
302	ldr	$Thi,[$ctx,#$Boff+$hi]
303	ldr	$t0, [$ctx,#$Coff+$lo]
304	ldr	$t1, [$ctx,#$Coff+$hi]
305	ldr	$t2, [$ctx,#$Doff+$lo]
306	ldr	$t3, [$ctx,#$Doff+$hi]
307	str	$Tlo,[sp,#$Boff+0]
308	str	$Thi,[sp,#$Boff+4]
309	str	$t0, [sp,#$Coff+0]
310	str	$t1, [sp,#$Coff+4]
311	str	$t2, [sp,#$Doff+0]
312	str	$t3, [sp,#$Doff+4]
313	ldr	$Tlo,[$ctx,#$Foff+$lo]
314	ldr	$Thi,[$ctx,#$Foff+$hi]
315	str	$Tlo,[sp,#$Foff+0]
316	str	$Thi,[sp,#$Foff+4]
317
318.L00_15:
319#if __ARM_ARCH<7
320	ldrb	$Tlo,[$inp,#7]
321	ldrb	$t0, [$inp,#6]
322	ldrb	$t1, [$inp,#5]
323	ldrb	$t2, [$inp,#4]
324	ldrb	$Thi,[$inp,#3]
325	ldrb	$t3, [$inp,#2]
326	orr	$Tlo,$Tlo,$t0,lsl#8
327	ldrb	$t0, [$inp,#1]
328	orr	$Tlo,$Tlo,$t1,lsl#16
329	ldrb	$t1, [$inp],#8
330	orr	$Tlo,$Tlo,$t2,lsl#24
331	orr	$Thi,$Thi,$t3,lsl#8
332	orr	$Thi,$Thi,$t0,lsl#16
333	orr	$Thi,$Thi,$t1,lsl#24
334#else
335	ldr	$Tlo,[$inp,#4]
336	ldr	$Thi,[$inp],#8
337#ifdef __ARMEL__
338	rev	$Tlo,$Tlo
339	rev	$Thi,$Thi
340#endif
341#endif
342___
343	&BODY_00_15(0x94);
344$code.=<<___;
345	tst	$Ktbl,#1
346	beq	.L00_15
347	ldr	$t0,[sp,#`$Xoff+8*(16-1)`+0]
348	ldr	$t1,[sp,#`$Xoff+8*(16-1)`+4]
349	bic	$Ktbl,$Ktbl,#1
350.L16_79:
351	@ sigma0(x)	(ROTR((x),1)  ^ ROTR((x),8)  ^ ((x)>>7))
352	@ LO		lo>>1^hi<<31  ^ lo>>8^hi<<24 ^ lo>>7^hi<<25
353	@ HI		hi>>1^lo<<31  ^ hi>>8^lo<<24 ^ hi>>7
354	mov	$Tlo,$t0,lsr#1
355	ldr	$t2,[sp,#`$Xoff+8*(16-14)`+0]
356	mov	$Thi,$t1,lsr#1
357	ldr	$t3,[sp,#`$Xoff+8*(16-14)`+4]
358	eor	$Tlo,$Tlo,$t1,lsl#31
359	eor	$Thi,$Thi,$t0,lsl#31
360	eor	$Tlo,$Tlo,$t0,lsr#8
361	eor	$Thi,$Thi,$t1,lsr#8
362	eor	$Tlo,$Tlo,$t1,lsl#24
363	eor	$Thi,$Thi,$t0,lsl#24
364	eor	$Tlo,$Tlo,$t0,lsr#7
365	eor	$Thi,$Thi,$t1,lsr#7
366	eor	$Tlo,$Tlo,$t1,lsl#25
367
368	@ sigma1(x)	(ROTR((x),19) ^ ROTR((x),61) ^ ((x)>>6))
369	@ LO		lo>>19^hi<<13 ^ hi>>29^lo<<3 ^ lo>>6^hi<<26
370	@ HI		hi>>19^lo<<13 ^ lo>>29^hi<<3 ^ hi>>6
371	mov	$t0,$t2,lsr#19
372	mov	$t1,$t3,lsr#19
373	eor	$t0,$t0,$t3,lsl#13
374	eor	$t1,$t1,$t2,lsl#13
375	eor	$t0,$t0,$t3,lsr#29
376	eor	$t1,$t1,$t2,lsr#29
377	eor	$t0,$t0,$t2,lsl#3
378	eor	$t1,$t1,$t3,lsl#3
379	eor	$t0,$t0,$t2,lsr#6
380	eor	$t1,$t1,$t3,lsr#6
381	ldr	$t2,[sp,#`$Xoff+8*(16-9)`+0]
382	eor	$t0,$t0,$t3,lsl#26
383
384	ldr	$t3,[sp,#`$Xoff+8*(16-9)`+4]
385	adds	$Tlo,$Tlo,$t0
386	ldr	$t0,[sp,#`$Xoff+8*16`+0]
387	adc	$Thi,$Thi,$t1
388
389	ldr	$t1,[sp,#`$Xoff+8*16`+4]
390	adds	$Tlo,$Tlo,$t2
391	adc	$Thi,$Thi,$t3
392	adds	$Tlo,$Tlo,$t0
393	adc	$Thi,$Thi,$t1
394___
395	&BODY_00_15(0x17);
396$code.=<<___;
397#if __ARM_ARCH>=7
398	ittt	eq			@ Thumb2 thing, sanity check in ARM
399#endif
400	ldreq	$t0,[sp,#`$Xoff+8*(16-1)`+0]
401	ldreq	$t1,[sp,#`$Xoff+8*(16-1)`+4]
402	beq	.L16_79
403	bic	$Ktbl,$Ktbl,#1
404
405	ldr	$Tlo,[sp,#$Boff+0]
406	ldr	$Thi,[sp,#$Boff+4]
407	ldr	$t0, [$ctx,#$Aoff+$lo]
408	ldr	$t1, [$ctx,#$Aoff+$hi]
409	ldr	$t2, [$ctx,#$Boff+$lo]
410	ldr	$t3, [$ctx,#$Boff+$hi]
411	adds	$t0,$Alo,$t0
412	str	$t0, [$ctx,#$Aoff+$lo]
413	adc	$t1,$Ahi,$t1
414	str	$t1, [$ctx,#$Aoff+$hi]
415	adds	$t2,$Tlo,$t2
416	str	$t2, [$ctx,#$Boff+$lo]
417	adc	$t3,$Thi,$t3
418	str	$t3, [$ctx,#$Boff+$hi]
419
420	ldr	$Alo,[sp,#$Coff+0]
421	ldr	$Ahi,[sp,#$Coff+4]
422	ldr	$Tlo,[sp,#$Doff+0]
423	ldr	$Thi,[sp,#$Doff+4]
424	ldr	$t0, [$ctx,#$Coff+$lo]
425	ldr	$t1, [$ctx,#$Coff+$hi]
426	ldr	$t2, [$ctx,#$Doff+$lo]
427	ldr	$t3, [$ctx,#$Doff+$hi]
428	adds	$t0,$Alo,$t0
429	str	$t0, [$ctx,#$Coff+$lo]
430	adc	$t1,$Ahi,$t1
431	str	$t1, [$ctx,#$Coff+$hi]
432	adds	$t2,$Tlo,$t2
433	str	$t2, [$ctx,#$Doff+$lo]
434	adc	$t3,$Thi,$t3
435	str	$t3, [$ctx,#$Doff+$hi]
436
437	ldr	$Tlo,[sp,#$Foff+0]
438	ldr	$Thi,[sp,#$Foff+4]
439	ldr	$t0, [$ctx,#$Eoff+$lo]
440	ldr	$t1, [$ctx,#$Eoff+$hi]
441	ldr	$t2, [$ctx,#$Foff+$lo]
442	ldr	$t3, [$ctx,#$Foff+$hi]
443	adds	$Elo,$Elo,$t0
444	str	$Elo,[$ctx,#$Eoff+$lo]
445	adc	$Ehi,$Ehi,$t1
446	str	$Ehi,[$ctx,#$Eoff+$hi]
447	adds	$t2,$Tlo,$t2
448	str	$t2, [$ctx,#$Foff+$lo]
449	adc	$t3,$Thi,$t3
450	str	$t3, [$ctx,#$Foff+$hi]
451
452	ldr	$Alo,[sp,#$Goff+0]
453	ldr	$Ahi,[sp,#$Goff+4]
454	ldr	$Tlo,[sp,#$Hoff+0]
455	ldr	$Thi,[sp,#$Hoff+4]
456	ldr	$t0, [$ctx,#$Goff+$lo]
457	ldr	$t1, [$ctx,#$Goff+$hi]
458	ldr	$t2, [$ctx,#$Hoff+$lo]
459	ldr	$t3, [$ctx,#$Hoff+$hi]
460	adds	$t0,$Alo,$t0
461	str	$t0, [$ctx,#$Goff+$lo]
462	adc	$t1,$Ahi,$t1
463	str	$t1, [$ctx,#$Goff+$hi]
464	adds	$t2,$Tlo,$t2
465	str	$t2, [$ctx,#$Hoff+$lo]
466	adc	$t3,$Thi,$t3
467	str	$t3, [$ctx,#$Hoff+$hi]
468
469	add	sp,sp,#640
470	sub	$Ktbl,$Ktbl,#640
471
472	teq	$inp,$len
473	bne	.Loop
474
475	add	sp,sp,#8*9		@ destroy frame
476#if __ARM_ARCH>=5
477	ldmia	sp!,{r4-r12,pc}
478#else
479	ldmia	sp!,{r4-r12,lr}
480	tst	lr,#1
481	moveq	pc,lr			@ be binary compatible with V4, yet
482	bx	lr			@ interoperable with Thumb ISA:-)
483#endif
484.size	sha512_block_data_order_nohw,.-sha512_block_data_order_nohw
485___
486
487{
488my @Sigma0=(28,34,39);
489my @Sigma1=(14,18,41);
490my @sigma0=(1, 8, 7);
491my @sigma1=(19,61,6);
492
493my $Ktbl="r3";
494my $cnt="r12";	# volatile register known as ip, intra-procedure-call scratch
495
496my @X=map("d$_",(0..15));
497my @V=($A,$B,$C,$D,$E,$F,$G,$H)=map("d$_",(16..23));
498
499sub NEON_00_15() {
500my $i=shift;
501my ($a,$b,$c,$d,$e,$f,$g,$h)=@_;
502my ($t0,$t1,$t2,$T1,$K,$Ch,$Maj)=map("d$_",(24..31));	# temps
503
504$code.=<<___ if ($i<16 || $i&1);
505	vshr.u64	$t0,$e,#@Sigma1[0]	@ $i
506#if $i<16
507	vld1.64		{@X[$i%16]},[$inp]!	@ handles unaligned
508#endif
509	vshr.u64	$t1,$e,#@Sigma1[1]
510#if $i>0
511	 vadd.i64	$a,$Maj			@ h+=Maj from the past
512#endif
513	vshr.u64	$t2,$e,#@Sigma1[2]
514___
515$code.=<<___;
516	vld1.64		{$K},[$Ktbl,:64]!	@ K[i++]
517	vsli.64		$t0,$e,#`64-@Sigma1[0]`
518	vsli.64		$t1,$e,#`64-@Sigma1[1]`
519	vmov		$Ch,$e
520	vsli.64		$t2,$e,#`64-@Sigma1[2]`
521#if $i<16 && defined(__ARMEL__)
522	vrev64.8	@X[$i],@X[$i]
523#endif
524	veor		$t1,$t0
525	vbsl		$Ch,$f,$g		@ Ch(e,f,g)
526	vshr.u64	$t0,$a,#@Sigma0[0]
527	veor		$t2,$t1			@ Sigma1(e)
528	vadd.i64	$T1,$Ch,$h
529	vshr.u64	$t1,$a,#@Sigma0[1]
530	vsli.64		$t0,$a,#`64-@Sigma0[0]`
531	vadd.i64	$T1,$t2
532	vshr.u64	$t2,$a,#@Sigma0[2]
533	vadd.i64	$K,@X[$i%16]
534	vsli.64		$t1,$a,#`64-@Sigma0[1]`
535	veor		$Maj,$a,$b
536	vsli.64		$t2,$a,#`64-@Sigma0[2]`
537	veor		$h,$t0,$t1
538	vadd.i64	$T1,$K
539	vbsl		$Maj,$c,$b		@ Maj(a,b,c)
540	veor		$h,$t2			@ Sigma0(a)
541	vadd.i64	$d,$T1
542	vadd.i64	$Maj,$T1
543	@ vadd.i64	$h,$Maj
544___
545}
546
547sub NEON_16_79() {
548my $i=shift;
549
550if ($i&1)	{ &NEON_00_15($i,@_); return; }
551
552# 2x-vectorized, therefore runs every 2nd round
553my @X=map("q$_",(0..7));			# view @X as 128-bit vector
554my ($t0,$t1,$s0,$s1) = map("q$_",(12..15));	# temps
555my ($d0,$d1,$d2) = map("d$_",(24..26));		# temps from NEON_00_15
556my $e=@_[4];					# $e from NEON_00_15
557$i /= 2;
558$code.=<<___;
559	vshr.u64	$t0,@X[($i+7)%8],#@sigma1[0]
560	vshr.u64	$t1,@X[($i+7)%8],#@sigma1[1]
561	 vadd.i64	@_[0],d30			@ h+=Maj from the past
562	vshr.u64	$s1,@X[($i+7)%8],#@sigma1[2]
563	vsli.64		$t0,@X[($i+7)%8],#`64-@sigma1[0]`
564	vext.8		$s0,@X[$i%8],@X[($i+1)%8],#8	@ X[i+1]
565	vsli.64		$t1,@X[($i+7)%8],#`64-@sigma1[1]`
566	veor		$s1,$t0
567	vshr.u64	$t0,$s0,#@sigma0[0]
568	veor		$s1,$t1				@ sigma1(X[i+14])
569	vshr.u64	$t1,$s0,#@sigma0[1]
570	vadd.i64	@X[$i%8],$s1
571	vshr.u64	$s1,$s0,#@sigma0[2]
572	vsli.64		$t0,$s0,#`64-@sigma0[0]`
573	vsli.64		$t1,$s0,#`64-@sigma0[1]`
574	vext.8		$s0,@X[($i+4)%8],@X[($i+5)%8],#8	@ X[i+9]
575	veor		$s1,$t0
576	vshr.u64	$d0,$e,#@Sigma1[0]		@ from NEON_00_15
577	vadd.i64	@X[$i%8],$s0
578	vshr.u64	$d1,$e,#@Sigma1[1]		@ from NEON_00_15
579	veor		$s1,$t1				@ sigma0(X[i+1])
580	vshr.u64	$d2,$e,#@Sigma1[2]		@ from NEON_00_15
581	vadd.i64	@X[$i%8],$s1
582___
583	&NEON_00_15(2*$i,@_);
584}
585
586$code.=<<___;
587#if __ARM_MAX_ARCH__>=7
588.arch	armv7-a
589.fpu	neon
590
591.global	sha512_block_data_order_neon
592.type	sha512_block_data_order_neon,%function
593.align	4
594sha512_block_data_order_neon:
595	dmb				@ errata #451034 on early Cortex A8
596	add	$len,$inp,$len,lsl#7	@ len to point at the end of inp
597	adr	$Ktbl,K512
598	VFP_ABI_PUSH
599	vldmia	$ctx,{$A-$H}		@ load context
600.Loop_neon:
601___
602for($i=0;$i<16;$i++)	{ &NEON_00_15($i,@V); unshift(@V,pop(@V)); }
603$code.=<<___;
604	mov		$cnt,#4
605.L16_79_neon:
606	subs		$cnt,#1
607___
608for(;$i<32;$i++)	{ &NEON_16_79($i,@V); unshift(@V,pop(@V)); }
609$code.=<<___;
610	bne		.L16_79_neon
611
612	 vadd.i64	$A,d30		@ h+=Maj from the past
613	vldmia		$ctx,{d24-d31}	@ load context to temp
614	vadd.i64	q8,q12		@ vectorized accumulate
615	vadd.i64	q9,q13
616	vadd.i64	q10,q14
617	vadd.i64	q11,q15
618	vstmia		$ctx,{$A-$H}	@ save context
619	teq		$inp,$len
620	sub		$Ktbl,#640	@ rewind K512
621	bne		.Loop_neon
622
623	VFP_ABI_POP
624	ret				@ bx lr
625.size	sha512_block_data_order_neon,.-sha512_block_data_order_neon
626#endif
627___
628}
629$code.=<<___;
630.asciz	"SHA512 block transform for ARMv4/NEON, CRYPTOGAMS by <appro\@openssl.org>"
631.align	2
632___
633
634$code =~ s/\`([^\`]*)\`/eval $1/gem;
635$code =~ s/\bbx\s+lr\b/.word\t0xe12fff1e/gm;	# make it possible to compile with -march=armv4
636$code =~ s/\bret\b/bx	lr/gm;
637
638open SELF,$0;
639while(<SELF>) {
640	next if (/^#!/);
641	last if (!s/^#/@/ and !/^$/);
642	print;
643}
644close SELF;
645
646print $code;
647close STDOUT or die "error closing STDOUT: $!"; # enforce flush
648