1#! /usr/bin/env perl
2# Copyright 2005-2016 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# Version 2.1.
18#
19# aes-*-cbc benchmarks are improved by >70% [compared to gcc 3.3.2 on
20# Opteron 240 CPU] plus all the bells-n-whistles from 32-bit version
21# [you'll notice a lot of resemblance], such as compressed S-boxes
22# in little-endian byte order, prefetch of these tables in CBC mode,
23# as well as avoiding L1 cache aliasing between stack frame and key
24# schedule and already mentioned tables, compressed Td4...
25#
26# Performance in number of cycles per processed byte for 128-bit key:
27#
28#		ECB encrypt	ECB decrypt	CBC large chunk
29# AMD64		33		43		13.0
30# EM64T		38		56		18.6(*)
31# Core 2	30		42		14.5(*)
32# Atom		65		86		32.1(*)
33#
34# (*) with hyper-threading off
35
36$flavour = shift;
37$output  = shift;
38if ($flavour =~ /\./) { $output = $flavour; undef $flavour; }
39
40$win64=0; $win64=1 if ($flavour =~ /[nm]asm|mingw64/ || $output =~ /\.asm$/);
41
42$0 =~ m/(.*[\/\\])[^\/\\]+$/; $dir=$1;
43( $xlate="${dir}x86_64-xlate.pl" and -f $xlate ) or
44( $xlate="${dir}../../../perlasm/x86_64-xlate.pl" and -f $xlate) or
45die "can't locate x86_64-xlate.pl";
46
47open OUT,"| \"$^X\" \"$xlate\" $flavour \"$output\"";
48*STDOUT=*OUT;
49
50$verticalspin=1;	# unlike 32-bit version $verticalspin performs
51			# ~15% better on both AMD and Intel cores
52$speed_limit=512;	# see aes-586.pl for details
53
54$code=".text\n";
55
56$s0="%eax";
57$s1="%ebx";
58$s2="%ecx";
59$s3="%edx";
60$acc0="%esi";	$mask80="%rsi";
61$acc1="%edi";	$maskfe="%rdi";
62$acc2="%ebp";	$mask1b="%rbp";
63$inp="%r8";
64$out="%r9";
65$t0="%r10d";
66$t1="%r11d";
67$t2="%r12d";
68$rnds="%r13d";
69$sbox="%r14";
70$key="%r15";
71
72sub hi() { my $r=shift;	$r =~ s/%[er]([a-d])x/%\1h/;	$r; }
73sub lo() { my $r=shift;	$r =~ s/%[er]([a-d])x/%\1l/;
74			$r =~ s/%[er]([sd]i)/%\1l/;
75			$r =~ s/%(r[0-9]+)[d]?/%\1b/;	$r; }
76sub LO() { my $r=shift; $r =~ s/%r([a-z]+)/%e\1/;
77			$r =~ s/%r([0-9]+)/%r\1d/;	$r; }
78sub _data_word()
79{ my $i;
80    while(defined($i=shift)) { $code.=sprintf".long\t0x%08x,0x%08x\n",$i,$i; }
81}
82sub data_word()
83{ my $i;
84  my $last=pop(@_);
85    $code.=".long\t";
86    while(defined($i=shift)) { $code.=sprintf"0x%08x,",$i; }
87    $code.=sprintf"0x%08x\n",$last;
88}
89
90sub data_byte()
91{ my $i;
92  my $last=pop(@_);
93    $code.=".byte\t";
94    while(defined($i=shift)) { $code.=sprintf"0x%02x,",$i&0xff; }
95    $code.=sprintf"0x%02x\n",$last&0xff;
96}
97
98sub encvert()
99{ my $t3="%r8d";	# zaps $inp!
100
101$code.=<<___;
102	# favor 3-way issue Opteron pipeline...
103	movzb	`&lo("$s0")`,$acc0
104	movzb	`&lo("$s1")`,$acc1
105	movzb	`&lo("$s2")`,$acc2
106	mov	0($sbox,$acc0,8),$t0
107	mov	0($sbox,$acc1,8),$t1
108	mov	0($sbox,$acc2,8),$t2
109
110	movzb	`&hi("$s1")`,$acc0
111	movzb	`&hi("$s2")`,$acc1
112	movzb	`&lo("$s3")`,$acc2
113	xor	3($sbox,$acc0,8),$t0
114	xor	3($sbox,$acc1,8),$t1
115	mov	0($sbox,$acc2,8),$t3
116
117	movzb	`&hi("$s3")`,$acc0
118	shr	\$16,$s2
119	movzb	`&hi("$s0")`,$acc2
120	xor	3($sbox,$acc0,8),$t2
121	shr	\$16,$s3
122	xor	3($sbox,$acc2,8),$t3
123
124	shr	\$16,$s1
125	lea	16($key),$key
126	shr	\$16,$s0
127
128	movzb	`&lo("$s2")`,$acc0
129	movzb	`&lo("$s3")`,$acc1
130	movzb	`&lo("$s0")`,$acc2
131	xor	2($sbox,$acc0,8),$t0
132	xor	2($sbox,$acc1,8),$t1
133	xor	2($sbox,$acc2,8),$t2
134
135	movzb	`&hi("$s3")`,$acc0
136	movzb	`&hi("$s0")`,$acc1
137	movzb	`&lo("$s1")`,$acc2
138	xor	1($sbox,$acc0,8),$t0
139	xor	1($sbox,$acc1,8),$t1
140	xor	2($sbox,$acc2,8),$t3
141
142	mov	12($key),$s3
143	movzb	`&hi("$s1")`,$acc1
144	movzb	`&hi("$s2")`,$acc2
145	mov	0($key),$s0
146	xor	1($sbox,$acc1,8),$t2
147	xor	1($sbox,$acc2,8),$t3
148
149	mov	4($key),$s1
150	mov	8($key),$s2
151	xor	$t0,$s0
152	xor	$t1,$s1
153	xor	$t2,$s2
154	xor	$t3,$s3
155___
156}
157
158sub enclastvert()
159{ my $t3="%r8d";	# zaps $inp!
160
161$code.=<<___;
162	movzb	`&lo("$s0")`,$acc0
163	movzb	`&lo("$s1")`,$acc1
164	movzb	`&lo("$s2")`,$acc2
165	movzb	2($sbox,$acc0,8),$t0
166	movzb	2($sbox,$acc1,8),$t1
167	movzb	2($sbox,$acc2,8),$t2
168
169	movzb	`&lo("$s3")`,$acc0
170	movzb	`&hi("$s1")`,$acc1
171	movzb	`&hi("$s2")`,$acc2
172	movzb	2($sbox,$acc0,8),$t3
173	mov	0($sbox,$acc1,8),$acc1	#$t0
174	mov	0($sbox,$acc2,8),$acc2	#$t1
175
176	and	\$0x0000ff00,$acc1
177	and	\$0x0000ff00,$acc2
178
179	xor	$acc1,$t0
180	xor	$acc2,$t1
181	shr	\$16,$s2
182
183	movzb	`&hi("$s3")`,$acc0
184	movzb	`&hi("$s0")`,$acc1
185	shr	\$16,$s3
186	mov	0($sbox,$acc0,8),$acc0	#$t2
187	mov	0($sbox,$acc1,8),$acc1	#$t3
188
189	and	\$0x0000ff00,$acc0
190	and	\$0x0000ff00,$acc1
191	shr	\$16,$s1
192	xor	$acc0,$t2
193	xor	$acc1,$t3
194	shr	\$16,$s0
195
196	movzb	`&lo("$s2")`,$acc0
197	movzb	`&lo("$s3")`,$acc1
198	movzb	`&lo("$s0")`,$acc2
199	mov	0($sbox,$acc0,8),$acc0	#$t0
200	mov	0($sbox,$acc1,8),$acc1	#$t1
201	mov	0($sbox,$acc2,8),$acc2	#$t2
202
203	and	\$0x00ff0000,$acc0
204	and	\$0x00ff0000,$acc1
205	and	\$0x00ff0000,$acc2
206
207	xor	$acc0,$t0
208	xor	$acc1,$t1
209	xor	$acc2,$t2
210
211	movzb	`&lo("$s1")`,$acc0
212	movzb	`&hi("$s3")`,$acc1
213	movzb	`&hi("$s0")`,$acc2
214	mov	0($sbox,$acc0,8),$acc0	#$t3
215	mov	2($sbox,$acc1,8),$acc1	#$t0
216	mov	2($sbox,$acc2,8),$acc2	#$t1
217
218	and	\$0x00ff0000,$acc0
219	and	\$0xff000000,$acc1
220	and	\$0xff000000,$acc2
221
222	xor	$acc0,$t3
223	xor	$acc1,$t0
224	xor	$acc2,$t1
225
226	movzb	`&hi("$s1")`,$acc0
227	movzb	`&hi("$s2")`,$acc1
228	mov	16+12($key),$s3
229	mov	2($sbox,$acc0,8),$acc0	#$t2
230	mov	2($sbox,$acc1,8),$acc1	#$t3
231	mov	16+0($key),$s0
232
233	and	\$0xff000000,$acc0
234	and	\$0xff000000,$acc1
235
236	xor	$acc0,$t2
237	xor	$acc1,$t3
238
239	mov	16+4($key),$s1
240	mov	16+8($key),$s2
241	xor	$t0,$s0
242	xor	$t1,$s1
243	xor	$t2,$s2
244	xor	$t3,$s3
245___
246}
247
248sub encstep()
249{ my ($i,@s) = @_;
250  my $tmp0=$acc0;
251  my $tmp1=$acc1;
252  my $tmp2=$acc2;
253  my $out=($t0,$t1,$t2,$s[0])[$i];
254
255	if ($i==3) {
256		$tmp0=$s[1];
257		$tmp1=$s[2];
258		$tmp2=$s[3];
259	}
260	$code.="	movzb	".&lo($s[0]).",$out\n";
261	$code.="	mov	$s[2],$tmp1\n"		if ($i!=3);
262	$code.="	lea	16($key),$key\n"	if ($i==0);
263
264	$code.="	movzb	".&hi($s[1]).",$tmp0\n";
265	$code.="	mov	0($sbox,$out,8),$out\n";
266
267	$code.="	shr	\$16,$tmp1\n";
268	$code.="	mov	$s[3],$tmp2\n"		if ($i!=3);
269	$code.="	xor	3($sbox,$tmp0,8),$out\n";
270
271	$code.="	movzb	".&lo($tmp1).",$tmp1\n";
272	$code.="	shr	\$24,$tmp2\n";
273	$code.="	xor	4*$i($key),$out\n";
274
275	$code.="	xor	2($sbox,$tmp1,8),$out\n";
276	$code.="	xor	1($sbox,$tmp2,8),$out\n";
277
278	$code.="	mov	$t0,$s[1]\n"		if ($i==3);
279	$code.="	mov	$t1,$s[2]\n"		if ($i==3);
280	$code.="	mov	$t2,$s[3]\n"		if ($i==3);
281	$code.="\n";
282}
283
284sub enclast()
285{ my ($i,@s)=@_;
286  my $tmp0=$acc0;
287  my $tmp1=$acc1;
288  my $tmp2=$acc2;
289  my $out=($t0,$t1,$t2,$s[0])[$i];
290
291	if ($i==3) {
292		$tmp0=$s[1];
293		$tmp1=$s[2];
294		$tmp2=$s[3];
295	}
296	$code.="	movzb	".&lo($s[0]).",$out\n";
297	$code.="	mov	$s[2],$tmp1\n"		if ($i!=3);
298
299	$code.="	mov	2($sbox,$out,8),$out\n";
300	$code.="	shr	\$16,$tmp1\n";
301	$code.="	mov	$s[3],$tmp2\n"		if ($i!=3);
302
303	$code.="	and	\$0x000000ff,$out\n";
304	$code.="	movzb	".&hi($s[1]).",$tmp0\n";
305	$code.="	movzb	".&lo($tmp1).",$tmp1\n";
306	$code.="	shr	\$24,$tmp2\n";
307
308	$code.="	mov	0($sbox,$tmp0,8),$tmp0\n";
309	$code.="	mov	0($sbox,$tmp1,8),$tmp1\n";
310	$code.="	mov	2($sbox,$tmp2,8),$tmp2\n";
311
312	$code.="	and	\$0x0000ff00,$tmp0\n";
313	$code.="	and	\$0x00ff0000,$tmp1\n";
314	$code.="	and	\$0xff000000,$tmp2\n";
315
316	$code.="	xor	$tmp0,$out\n";
317	$code.="	mov	$t0,$s[1]\n"		if ($i==3);
318	$code.="	xor	$tmp1,$out\n";
319	$code.="	mov	$t1,$s[2]\n"		if ($i==3);
320	$code.="	xor	$tmp2,$out\n";
321	$code.="	mov	$t2,$s[3]\n"		if ($i==3);
322	$code.="\n";
323}
324
325$code.=<<___;
326.type	_x86_64_AES_encrypt,\@abi-omnipotent
327.align	16
328_x86_64_AES_encrypt:
329	xor	0($key),$s0			# xor with key
330	xor	4($key),$s1
331	xor	8($key),$s2
332	xor	12($key),$s3
333
334	mov	240($key),$rnds			# load key->rounds
335	sub	\$1,$rnds
336	jmp	.Lenc_loop
337.align	16
338.Lenc_loop:
339___
340	if ($verticalspin) { &encvert(); }
341	else {	&encstep(0,$s0,$s1,$s2,$s3);
342		&encstep(1,$s1,$s2,$s3,$s0);
343		&encstep(2,$s2,$s3,$s0,$s1);
344		&encstep(3,$s3,$s0,$s1,$s2);
345	}
346$code.=<<___;
347	sub	\$1,$rnds
348	jnz	.Lenc_loop
349___
350	if ($verticalspin) { &enclastvert(); }
351	else {	&enclast(0,$s0,$s1,$s2,$s3);
352		&enclast(1,$s1,$s2,$s3,$s0);
353		&enclast(2,$s2,$s3,$s0,$s1);
354		&enclast(3,$s3,$s0,$s1,$s2);
355		$code.=<<___;
356		xor	16+0($key),$s0		# xor with key
357		xor	16+4($key),$s1
358		xor	16+8($key),$s2
359		xor	16+12($key),$s3
360___
361	}
362$code.=<<___;
363	.byte	0xf3,0xc3			# rep ret
364.size	_x86_64_AES_encrypt,.-_x86_64_AES_encrypt
365___
366
367# it's possible to implement this by shifting tN by 8, filling least
368# significant byte with byte load and finally bswap-ing at the end,
369# but such partial register load kills Core 2...
370sub enccompactvert()
371{ my ($t3,$t4,$t5)=("%r8d","%r9d","%r13d");
372
373$code.=<<___;
374	movzb	`&lo("$s0")`,$t0
375	movzb	`&lo("$s1")`,$t1
376	movzb	`&lo("$s2")`,$t2
377	movzb	`&lo("$s3")`,$t3
378	movzb	`&hi("$s1")`,$acc0
379	movzb	`&hi("$s2")`,$acc1
380	shr	\$16,$s2
381	movzb	`&hi("$s3")`,$acc2
382	movzb	($sbox,$t0,1),$t0
383	movzb	($sbox,$t1,1),$t1
384	movzb	($sbox,$t2,1),$t2
385	movzb	($sbox,$t3,1),$t3
386
387	movzb	($sbox,$acc0,1),$t4	#$t0
388	movzb	`&hi("$s0")`,$acc0
389	movzb	($sbox,$acc1,1),$t5	#$t1
390	movzb	`&lo("$s2")`,$acc1
391	movzb	($sbox,$acc2,1),$acc2	#$t2
392	movzb	($sbox,$acc0,1),$acc0	#$t3
393
394	shl	\$8,$t4
395	shr	\$16,$s3
396	shl	\$8,$t5
397	xor	$t4,$t0
398	shr	\$16,$s0
399	movzb	`&lo("$s3")`,$t4
400	shr	\$16,$s1
401	xor	$t5,$t1
402	shl	\$8,$acc2
403	movzb	`&lo("$s0")`,$t5
404	movzb	($sbox,$acc1,1),$acc1	#$t0
405	xor	$acc2,$t2
406
407	shl	\$8,$acc0
408	movzb	`&lo("$s1")`,$acc2
409	shl	\$16,$acc1
410	xor	$acc0,$t3
411	movzb	($sbox,$t4,1),$t4	#$t1
412	movzb	`&hi("$s3")`,$acc0
413	movzb	($sbox,$t5,1),$t5	#$t2
414	xor	$acc1,$t0
415
416	shr	\$8,$s2
417	movzb	`&hi("$s0")`,$acc1
418	shl	\$16,$t4
419	shr	\$8,$s1
420	shl	\$16,$t5
421	xor	$t4,$t1
422	movzb	($sbox,$acc2,1),$acc2	#$t3
423	movzb	($sbox,$acc0,1),$acc0	#$t0
424	movzb	($sbox,$acc1,1),$acc1	#$t1
425	movzb	($sbox,$s2,1),$s3	#$t3
426	movzb	($sbox,$s1,1),$s2	#$t2
427
428	shl	\$16,$acc2
429	xor	$t5,$t2
430	shl	\$24,$acc0
431	xor	$acc2,$t3
432	shl	\$24,$acc1
433	xor	$acc0,$t0
434	shl	\$24,$s3
435	xor	$acc1,$t1
436	shl	\$24,$s2
437	mov	$t0,$s0
438	mov	$t1,$s1
439	xor	$t2,$s2
440	xor	$t3,$s3
441___
442}
443
444sub enctransform_ref()
445{ my $sn = shift;
446  my ($acc,$r2,$tmp)=("%r8d","%r9d","%r13d");
447
448$code.=<<___;
449	mov	$sn,$acc
450	and	\$0x80808080,$acc
451	mov	$acc,$tmp
452	shr	\$7,$tmp
453	lea	($sn,$sn),$r2
454	sub	$tmp,$acc
455	and	\$0xfefefefe,$r2
456	and	\$0x1b1b1b1b,$acc
457	mov	$sn,$tmp
458	xor	$acc,$r2
459
460	xor	$r2,$sn
461	rol	\$24,$sn
462	xor	$r2,$sn
463	ror	\$16,$tmp
464	xor	$tmp,$sn
465	ror	\$8,$tmp
466	xor	$tmp,$sn
467___
468}
469
470# unlike decrypt case it does not pay off to parallelize enctransform
471sub enctransform()
472{ my ($t3,$r20,$r21)=($acc2,"%r8d","%r9d");
473
474$code.=<<___;
475	mov	\$0x80808080,$t0
476	mov	\$0x80808080,$t1
477	and	$s0,$t0
478	and	$s1,$t1
479	mov	$t0,$acc0
480	mov	$t1,$acc1
481	shr	\$7,$t0
482	lea	($s0,$s0),$r20
483	shr	\$7,$t1
484	lea	($s1,$s1),$r21
485	sub	$t0,$acc0
486	sub	$t1,$acc1
487	and	\$0xfefefefe,$r20
488	and	\$0xfefefefe,$r21
489	and	\$0x1b1b1b1b,$acc0
490	and	\$0x1b1b1b1b,$acc1
491	mov	$s0,$t0
492	mov	$s1,$t1
493	xor	$acc0,$r20
494	xor	$acc1,$r21
495
496	xor	$r20,$s0
497	xor	$r21,$s1
498	 mov	\$0x80808080,$t2
499	rol	\$24,$s0
500	 mov	\$0x80808080,$t3
501	rol	\$24,$s1
502	 and	$s2,$t2
503	 and	$s3,$t3
504	xor	$r20,$s0
505	xor	$r21,$s1
506	 mov	$t2,$acc0
507	ror	\$16,$t0
508	 mov	$t3,$acc1
509	ror	\$16,$t1
510	 lea	($s2,$s2),$r20
511	 shr	\$7,$t2
512	xor	$t0,$s0
513	 shr	\$7,$t3
514	xor	$t1,$s1
515	ror	\$8,$t0
516	 lea	($s3,$s3),$r21
517	ror	\$8,$t1
518	 sub	$t2,$acc0
519	 sub	$t3,$acc1
520	xor	$t0,$s0
521	xor	$t1,$s1
522
523	and	\$0xfefefefe,$r20
524	and	\$0xfefefefe,$r21
525	and	\$0x1b1b1b1b,$acc0
526	and	\$0x1b1b1b1b,$acc1
527	mov	$s2,$t2
528	mov	$s3,$t3
529	xor	$acc0,$r20
530	xor	$acc1,$r21
531
532	ror	\$16,$t2
533	xor	$r20,$s2
534	ror	\$16,$t3
535	xor	$r21,$s3
536	rol	\$24,$s2
537	mov	0($sbox),$acc0			# prefetch Te4
538	rol	\$24,$s3
539	xor	$r20,$s2
540	mov	64($sbox),$acc1
541	xor	$r21,$s3
542	mov	128($sbox),$r20
543	xor	$t2,$s2
544	ror	\$8,$t2
545	xor	$t3,$s3
546	ror	\$8,$t3
547	xor	$t2,$s2
548	mov	192($sbox),$r21
549	xor	$t3,$s3
550___
551}
552
553$code.=<<___;
554.type	_x86_64_AES_encrypt_compact,\@abi-omnipotent
555.align	16
556_x86_64_AES_encrypt_compact:
557	lea	128($sbox),$inp			# size optimization
558	mov	0-128($inp),$acc1		# prefetch Te4
559	mov	32-128($inp),$acc2
560	mov	64-128($inp),$t0
561	mov	96-128($inp),$t1
562	mov	128-128($inp),$acc1
563	mov	160-128($inp),$acc2
564	mov	192-128($inp),$t0
565	mov	224-128($inp),$t1
566	jmp	.Lenc_loop_compact
567.align	16
568.Lenc_loop_compact:
569		xor	0($key),$s0		# xor with key
570		xor	4($key),$s1
571		xor	8($key),$s2
572		xor	12($key),$s3
573		lea	16($key),$key
574___
575		&enccompactvert();
576$code.=<<___;
577		cmp	16(%rsp),$key
578		je	.Lenc_compact_done
579___
580		&enctransform();
581$code.=<<___;
582	jmp	.Lenc_loop_compact
583.align	16
584.Lenc_compact_done:
585	xor	0($key),$s0
586	xor	4($key),$s1
587	xor	8($key),$s2
588	xor	12($key),$s3
589	.byte	0xf3,0xc3			# rep ret
590.size	_x86_64_AES_encrypt_compact,.-_x86_64_AES_encrypt_compact
591___
592
593# void aes_nohw_encrypt (const void *inp,void *out,const AES_KEY *key);
594$code.=<<___;
595.align	16
596.globl	aes_nohw_encrypt
597.type	aes_nohw_encrypt,\@function,3
598.hidden	aes_nohw_encrypt
599aes_nohw_encrypt:
600.cfi_startproc
601	mov	%rsp,%rax
602.cfi_def_cfa_register	%rax
603	push	%rbx
604.cfi_push	%rbx
605	push	%rbp
606.cfi_push	%rbp
607	push	%r12
608.cfi_push	%r12
609	push	%r13
610.cfi_push	%r13
611	push	%r14
612.cfi_push	%r14
613	push	%r15
614.cfi_push	%r15
615
616	# allocate frame "above" key schedule
617	lea	-63(%rdx),%rcx	# %rdx is key argument
618	and	\$-64,%rsp
619	sub	%rsp,%rcx
620	neg	%rcx
621	and	\$0x3c0,%rcx
622	sub	%rcx,%rsp
623	sub	\$32,%rsp
624
625	mov	%rsi,16(%rsp)	# save out
626	mov	%rax,24(%rsp)	# save original stack pointer
627.cfi_cfa_expression	%rsp+24,deref,+8
628.Lenc_prologue:
629
630	mov	%rdx,$key
631	mov	240($key),$rnds	# load rounds
632
633	mov	0(%rdi),$s0	# load input vector
634	mov	4(%rdi),$s1
635	mov	8(%rdi),$s2
636	mov	12(%rdi),$s3
637
638	shl	\$4,$rnds
639	lea	($key,$rnds),%rbp
640	mov	$key,(%rsp)	# key schedule
641	mov	%rbp,8(%rsp)	# end of key schedule
642
643	# pick Te4 copy which can't "overlap" with stack frame or key schedule
644	lea	.LAES_Te+2048(%rip),$sbox
645	lea	768(%rsp),%rbp
646	sub	$sbox,%rbp
647	and	\$0x300,%rbp
648	lea	($sbox,%rbp),$sbox
649
650	call	_x86_64_AES_encrypt_compact
651
652	mov	16(%rsp),$out	# restore out
653	mov	24(%rsp),%rsi	# restore saved stack pointer
654.cfi_def_cfa	%rsi,8
655	mov	$s0,0($out)	# write output vector
656	mov	$s1,4($out)
657	mov	$s2,8($out)
658	mov	$s3,12($out)
659
660	mov	-48(%rsi),%r15
661.cfi_restore	%r15
662	mov	-40(%rsi),%r14
663.cfi_restore	%r14
664	mov	-32(%rsi),%r13
665.cfi_restore	%r13
666	mov	-24(%rsi),%r12
667.cfi_restore	%r12
668	mov	-16(%rsi),%rbp
669.cfi_restore	%rbp
670	mov	-8(%rsi),%rbx
671.cfi_restore	%rbx
672	lea	(%rsi),%rsp
673.cfi_def_cfa_register	%rsp
674.Lenc_epilogue:
675	ret
676.cfi_endproc
677.size	aes_nohw_encrypt,.-aes_nohw_encrypt
678___
679
680#------------------------------------------------------------------#
681
682sub decvert()
683{ my $t3="%r8d";	# zaps $inp!
684
685$code.=<<___;
686	# favor 3-way issue Opteron pipeline...
687	movzb	`&lo("$s0")`,$acc0
688	movzb	`&lo("$s1")`,$acc1
689	movzb	`&lo("$s2")`,$acc2
690	mov	0($sbox,$acc0,8),$t0
691	mov	0($sbox,$acc1,8),$t1
692	mov	0($sbox,$acc2,8),$t2
693
694	movzb	`&hi("$s3")`,$acc0
695	movzb	`&hi("$s0")`,$acc1
696	movzb	`&lo("$s3")`,$acc2
697	xor	3($sbox,$acc0,8),$t0
698	xor	3($sbox,$acc1,8),$t1
699	mov	0($sbox,$acc2,8),$t3
700
701	movzb	`&hi("$s1")`,$acc0
702	shr	\$16,$s0
703	movzb	`&hi("$s2")`,$acc2
704	xor	3($sbox,$acc0,8),$t2
705	shr	\$16,$s3
706	xor	3($sbox,$acc2,8),$t3
707
708	shr	\$16,$s1
709	lea	16($key),$key
710	shr	\$16,$s2
711
712	movzb	`&lo("$s2")`,$acc0
713	movzb	`&lo("$s3")`,$acc1
714	movzb	`&lo("$s0")`,$acc2
715	xor	2($sbox,$acc0,8),$t0
716	xor	2($sbox,$acc1,8),$t1
717	xor	2($sbox,$acc2,8),$t2
718
719	movzb	`&hi("$s1")`,$acc0
720	movzb	`&hi("$s2")`,$acc1
721	movzb	`&lo("$s1")`,$acc2
722	xor	1($sbox,$acc0,8),$t0
723	xor	1($sbox,$acc1,8),$t1
724	xor	2($sbox,$acc2,8),$t3
725
726	movzb	`&hi("$s3")`,$acc0
727	mov	12($key),$s3
728	movzb	`&hi("$s0")`,$acc2
729	xor	1($sbox,$acc0,8),$t2
730	mov	0($key),$s0
731	xor	1($sbox,$acc2,8),$t3
732
733	xor	$t0,$s0
734	mov	4($key),$s1
735	mov	8($key),$s2
736	xor	$t2,$s2
737	xor	$t1,$s1
738	xor	$t3,$s3
739___
740}
741
742sub declastvert()
743{ my $t3="%r8d";	# zaps $inp!
744
745$code.=<<___;
746	lea	2048($sbox),$sbox	# size optimization
747	movzb	`&lo("$s0")`,$acc0
748	movzb	`&lo("$s1")`,$acc1
749	movzb	`&lo("$s2")`,$acc2
750	movzb	($sbox,$acc0,1),$t0
751	movzb	($sbox,$acc1,1),$t1
752	movzb	($sbox,$acc2,1),$t2
753
754	movzb	`&lo("$s3")`,$acc0
755	movzb	`&hi("$s3")`,$acc1
756	movzb	`&hi("$s0")`,$acc2
757	movzb	($sbox,$acc0,1),$t3
758	movzb	($sbox,$acc1,1),$acc1	#$t0
759	movzb	($sbox,$acc2,1),$acc2	#$t1
760
761	shl	\$8,$acc1
762	shl	\$8,$acc2
763
764	xor	$acc1,$t0
765	xor	$acc2,$t1
766	shr	\$16,$s3
767
768	movzb	`&hi("$s1")`,$acc0
769	movzb	`&hi("$s2")`,$acc1
770	shr	\$16,$s0
771	movzb	($sbox,$acc0,1),$acc0	#$t2
772	movzb	($sbox,$acc1,1),$acc1	#$t3
773
774	shl	\$8,$acc0
775	shl	\$8,$acc1
776	shr	\$16,$s1
777	xor	$acc0,$t2
778	xor	$acc1,$t3
779	shr	\$16,$s2
780
781	movzb	`&lo("$s2")`,$acc0
782	movzb	`&lo("$s3")`,$acc1
783	movzb	`&lo("$s0")`,$acc2
784	movzb	($sbox,$acc0,1),$acc0	#$t0
785	movzb	($sbox,$acc1,1),$acc1	#$t1
786	movzb	($sbox,$acc2,1),$acc2	#$t2
787
788	shl	\$16,$acc0
789	shl	\$16,$acc1
790	shl	\$16,$acc2
791
792	xor	$acc0,$t0
793	xor	$acc1,$t1
794	xor	$acc2,$t2
795
796	movzb	`&lo("$s1")`,$acc0
797	movzb	`&hi("$s1")`,$acc1
798	movzb	`&hi("$s2")`,$acc2
799	movzb	($sbox,$acc0,1),$acc0	#$t3
800	movzb	($sbox,$acc1,1),$acc1	#$t0
801	movzb	($sbox,$acc2,1),$acc2	#$t1
802
803	shl	\$16,$acc0
804	shl	\$24,$acc1
805	shl	\$24,$acc2
806
807	xor	$acc0,$t3
808	xor	$acc1,$t0
809	xor	$acc2,$t1
810
811	movzb	`&hi("$s3")`,$acc0
812	movzb	`&hi("$s0")`,$acc1
813	mov	16+12($key),$s3
814	movzb	($sbox,$acc0,1),$acc0	#$t2
815	movzb	($sbox,$acc1,1),$acc1	#$t3
816	mov	16+0($key),$s0
817
818	shl	\$24,$acc0
819	shl	\$24,$acc1
820
821	xor	$acc0,$t2
822	xor	$acc1,$t3
823
824	mov	16+4($key),$s1
825	mov	16+8($key),$s2
826	lea	-2048($sbox),$sbox
827	xor	$t0,$s0
828	xor	$t1,$s1
829	xor	$t2,$s2
830	xor	$t3,$s3
831___
832}
833
834sub decstep()
835{ my ($i,@s) = @_;
836  my $tmp0=$acc0;
837  my $tmp1=$acc1;
838  my $tmp2=$acc2;
839  my $out=($t0,$t1,$t2,$s[0])[$i];
840
841	$code.="	mov	$s[0],$out\n"		if ($i!=3);
842			$tmp1=$s[2]			if ($i==3);
843	$code.="	mov	$s[2],$tmp1\n"		if ($i!=3);
844	$code.="	and	\$0xFF,$out\n";
845
846	$code.="	mov	0($sbox,$out,8),$out\n";
847	$code.="	shr	\$16,$tmp1\n";
848			$tmp2=$s[3]			if ($i==3);
849	$code.="	mov	$s[3],$tmp2\n"		if ($i!=3);
850
851			$tmp0=$s[1]			if ($i==3);
852	$code.="	movzb	".&hi($s[1]).",$tmp0\n";
853	$code.="	and	\$0xFF,$tmp1\n";
854	$code.="	shr	\$24,$tmp2\n";
855
856	$code.="	xor	3($sbox,$tmp0,8),$out\n";
857	$code.="	xor	2($sbox,$tmp1,8),$out\n";
858	$code.="	xor	1($sbox,$tmp2,8),$out\n";
859
860	$code.="	mov	$t2,$s[1]\n"		if ($i==3);
861	$code.="	mov	$t1,$s[2]\n"		if ($i==3);
862	$code.="	mov	$t0,$s[3]\n"		if ($i==3);
863	$code.="\n";
864}
865
866sub declast()
867{ my ($i,@s)=@_;
868  my $tmp0=$acc0;
869  my $tmp1=$acc1;
870  my $tmp2=$acc2;
871  my $out=($t0,$t1,$t2,$s[0])[$i];
872
873	$code.="	mov	$s[0],$out\n"		if ($i!=3);
874			$tmp1=$s[2]			if ($i==3);
875	$code.="	mov	$s[2],$tmp1\n"		if ($i!=3);
876	$code.="	and	\$0xFF,$out\n";
877
878	$code.="	movzb	2048($sbox,$out,1),$out\n";
879	$code.="	shr	\$16,$tmp1\n";
880			$tmp2=$s[3]			if ($i==3);
881	$code.="	mov	$s[3],$tmp2\n"		if ($i!=3);
882
883			$tmp0=$s[1]			if ($i==3);
884	$code.="	movzb	".&hi($s[1]).",$tmp0\n";
885	$code.="	and	\$0xFF,$tmp1\n";
886	$code.="	shr	\$24,$tmp2\n";
887
888	$code.="	movzb	2048($sbox,$tmp0,1),$tmp0\n";
889	$code.="	movzb	2048($sbox,$tmp1,1),$tmp1\n";
890	$code.="	movzb	2048($sbox,$tmp2,1),$tmp2\n";
891
892	$code.="	shl	\$8,$tmp0\n";
893	$code.="	shl	\$16,$tmp1\n";
894	$code.="	shl	\$24,$tmp2\n";
895
896	$code.="	xor	$tmp0,$out\n";
897	$code.="	mov	$t2,$s[1]\n"		if ($i==3);
898	$code.="	xor	$tmp1,$out\n";
899	$code.="	mov	$t1,$s[2]\n"		if ($i==3);
900	$code.="	xor	$tmp2,$out\n";
901	$code.="	mov	$t0,$s[3]\n"		if ($i==3);
902	$code.="\n";
903}
904
905$code.=<<___;
906.type	_x86_64_AES_decrypt,\@abi-omnipotent
907.align	16
908_x86_64_AES_decrypt:
909	xor	0($key),$s0			# xor with key
910	xor	4($key),$s1
911	xor	8($key),$s2
912	xor	12($key),$s3
913
914	mov	240($key),$rnds			# load key->rounds
915	sub	\$1,$rnds
916	jmp	.Ldec_loop
917.align	16
918.Ldec_loop:
919___
920	if ($verticalspin) { &decvert(); }
921	else {	&decstep(0,$s0,$s3,$s2,$s1);
922		&decstep(1,$s1,$s0,$s3,$s2);
923		&decstep(2,$s2,$s1,$s0,$s3);
924		&decstep(3,$s3,$s2,$s1,$s0);
925		$code.=<<___;
926		lea	16($key),$key
927		xor	0($key),$s0			# xor with key
928		xor	4($key),$s1
929		xor	8($key),$s2
930		xor	12($key),$s3
931___
932	}
933$code.=<<___;
934	sub	\$1,$rnds
935	jnz	.Ldec_loop
936___
937	if ($verticalspin) { &declastvert(); }
938	else {	&declast(0,$s0,$s3,$s2,$s1);
939		&declast(1,$s1,$s0,$s3,$s2);
940		&declast(2,$s2,$s1,$s0,$s3);
941		&declast(3,$s3,$s2,$s1,$s0);
942		$code.=<<___;
943		xor	16+0($key),$s0			# xor with key
944		xor	16+4($key),$s1
945		xor	16+8($key),$s2
946		xor	16+12($key),$s3
947___
948	}
949$code.=<<___;
950	.byte	0xf3,0xc3			# rep ret
951.size	_x86_64_AES_decrypt,.-_x86_64_AES_decrypt
952___
953
954sub deccompactvert()
955{ my ($t3,$t4,$t5)=("%r8d","%r9d","%r13d");
956
957$code.=<<___;
958	movzb	`&lo("$s0")`,$t0
959	movzb	`&lo("$s1")`,$t1
960	movzb	`&lo("$s2")`,$t2
961	movzb	`&lo("$s3")`,$t3
962	movzb	`&hi("$s3")`,$acc0
963	movzb	`&hi("$s0")`,$acc1
964	shr	\$16,$s3
965	movzb	`&hi("$s1")`,$acc2
966	movzb	($sbox,$t0,1),$t0
967	movzb	($sbox,$t1,1),$t1
968	movzb	($sbox,$t2,1),$t2
969	movzb	($sbox,$t3,1),$t3
970
971	movzb	($sbox,$acc0,1),$t4	#$t0
972	movzb	`&hi("$s2")`,$acc0
973	movzb	($sbox,$acc1,1),$t5	#$t1
974	movzb	($sbox,$acc2,1),$acc2	#$t2
975	movzb	($sbox,$acc0,1),$acc0	#$t3
976
977	shr	\$16,$s2
978	shl	\$8,$t5
979	shl	\$8,$t4
980	movzb	`&lo("$s2")`,$acc1
981	shr	\$16,$s0
982	xor	$t4,$t0
983	shr	\$16,$s1
984	movzb	`&lo("$s3")`,$t4
985
986	shl	\$8,$acc2
987	xor	$t5,$t1
988	shl	\$8,$acc0
989	movzb	`&lo("$s0")`,$t5
990	movzb	($sbox,$acc1,1),$acc1	#$t0
991	xor	$acc2,$t2
992	movzb	`&lo("$s1")`,$acc2
993
994	shl	\$16,$acc1
995	xor	$acc0,$t3
996	movzb	($sbox,$t4,1),$t4	#$t1
997	movzb	`&hi("$s1")`,$acc0
998	movzb	($sbox,$acc2,1),$acc2	#$t3
999	xor	$acc1,$t0
1000	movzb	($sbox,$t5,1),$t5	#$t2
1001	movzb	`&hi("$s2")`,$acc1
1002
1003	shl	\$16,$acc2
1004	shl	\$16,$t4
1005	shl	\$16,$t5
1006	xor	$acc2,$t3
1007	movzb	`&hi("$s3")`,$acc2
1008	xor	$t4,$t1
1009	shr	\$8,$s0
1010	xor	$t5,$t2
1011
1012	movzb	($sbox,$acc0,1),$acc0	#$t0
1013	movzb	($sbox,$acc1,1),$s1	#$t1
1014	movzb	($sbox,$acc2,1),$s2	#$t2
1015	movzb	($sbox,$s0,1),$s3	#$t3
1016
1017	mov	$t0,$s0
1018	shl	\$24,$acc0
1019	shl	\$24,$s1
1020	shl	\$24,$s2
1021	xor	$acc0,$s0
1022	shl	\$24,$s3
1023	xor	$t1,$s1
1024	xor	$t2,$s2
1025	xor	$t3,$s3
1026___
1027}
1028
1029# parallelized version! input is pair of 64-bit values: %rax=s1.s0
1030# and %rcx=s3.s2, output is four 32-bit values in %eax=s0, %ebx=s1,
1031# %ecx=s2 and %edx=s3.
1032sub dectransform()
1033{ my ($tp10,$tp20,$tp40,$tp80,$acc0)=("%rax","%r8", "%r9", "%r10","%rbx");
1034  my ($tp18,$tp28,$tp48,$tp88,$acc8)=("%rcx","%r11","%r12","%r13","%rdx");
1035  my $prefetch = shift;
1036
1037$code.=<<___;
1038	mov	$mask80,$tp40
1039	mov	$mask80,$tp48
1040	and	$tp10,$tp40
1041	and	$tp18,$tp48
1042	mov	$tp40,$acc0
1043	mov	$tp48,$acc8
1044	shr	\$7,$tp40
1045	lea	($tp10,$tp10),$tp20
1046	shr	\$7,$tp48
1047	lea	($tp18,$tp18),$tp28
1048	sub	$tp40,$acc0
1049	sub	$tp48,$acc8
1050	and	$maskfe,$tp20
1051	and	$maskfe,$tp28
1052	and	$mask1b,$acc0
1053	and	$mask1b,$acc8
1054	xor	$acc0,$tp20
1055	xor	$acc8,$tp28
1056	mov	$mask80,$tp80
1057	mov	$mask80,$tp88
1058
1059	and	$tp20,$tp80
1060	and	$tp28,$tp88
1061	mov	$tp80,$acc0
1062	mov	$tp88,$acc8
1063	shr	\$7,$tp80
1064	lea	($tp20,$tp20),$tp40
1065	shr	\$7,$tp88
1066	lea	($tp28,$tp28),$tp48
1067	sub	$tp80,$acc0
1068	sub	$tp88,$acc8
1069	and	$maskfe,$tp40
1070	and	$maskfe,$tp48
1071	and	$mask1b,$acc0
1072	and	$mask1b,$acc8
1073	xor	$acc0,$tp40
1074	xor	$acc8,$tp48
1075	mov	$mask80,$tp80
1076	mov	$mask80,$tp88
1077
1078	and	$tp40,$tp80
1079	and	$tp48,$tp88
1080	mov	$tp80,$acc0
1081	mov	$tp88,$acc8
1082	shr	\$7,$tp80
1083	 xor	$tp10,$tp20		# tp2^=tp1
1084	shr	\$7,$tp88
1085	 xor	$tp18,$tp28		# tp2^=tp1
1086	sub	$tp80,$acc0
1087	sub	$tp88,$acc8
1088	lea	($tp40,$tp40),$tp80
1089	lea	($tp48,$tp48),$tp88
1090	 xor	$tp10,$tp40		# tp4^=tp1
1091	 xor	$tp18,$tp48		# tp4^=tp1
1092	and	$maskfe,$tp80
1093	and	$maskfe,$tp88
1094	and	$mask1b,$acc0
1095	and	$mask1b,$acc8
1096	xor	$acc0,$tp80
1097	xor	$acc8,$tp88
1098
1099	xor	$tp80,$tp10		# tp1^=tp8
1100	xor	$tp88,$tp18		# tp1^=tp8
1101	xor	$tp80,$tp20		# tp2^tp1^=tp8
1102	xor	$tp88,$tp28		# tp2^tp1^=tp8
1103	mov	$tp10,$acc0
1104	mov	$tp18,$acc8
1105	xor	$tp80,$tp40		# tp4^tp1^=tp8
1106	shr	\$32,$acc0
1107	xor	$tp88,$tp48		# tp4^tp1^=tp8
1108	shr	\$32,$acc8
1109	xor	$tp20,$tp80		# tp8^=tp8^tp2^tp1=tp2^tp1
1110	rol	\$8,`&LO("$tp10")`	# ROTATE(tp1^tp8,8)
1111	xor	$tp28,$tp88		# tp8^=tp8^tp2^tp1=tp2^tp1
1112	rol	\$8,`&LO("$tp18")`	# ROTATE(tp1^tp8,8)
1113	xor	$tp40,$tp80		# tp2^tp1^=tp8^tp4^tp1=tp8^tp4^tp2
1114	rol	\$8,`&LO("$acc0")`	# ROTATE(tp1^tp8,8)
1115	xor	$tp48,$tp88		# tp2^tp1^=tp8^tp4^tp1=tp8^tp4^tp2
1116
1117	rol	\$8,`&LO("$acc8")`	# ROTATE(tp1^tp8,8)
1118	xor	`&LO("$tp80")`,`&LO("$tp10")`
1119	shr	\$32,$tp80
1120	xor	`&LO("$tp88")`,`&LO("$tp18")`
1121	shr	\$32,$tp88
1122	xor	`&LO("$tp80")`,`&LO("$acc0")`
1123	xor	`&LO("$tp88")`,`&LO("$acc8")`
1124
1125	mov	$tp20,$tp80
1126	rol	\$24,`&LO("$tp20")`	# ROTATE(tp2^tp1^tp8,24)
1127	mov	$tp28,$tp88
1128	rol	\$24,`&LO("$tp28")`	# ROTATE(tp2^tp1^tp8,24)
1129	shr	\$32,$tp80
1130	xor	`&LO("$tp20")`,`&LO("$tp10")`
1131	shr	\$32,$tp88
1132	xor	`&LO("$tp28")`,`&LO("$tp18")`
1133	rol	\$24,`&LO("$tp80")`	# ROTATE(tp2^tp1^tp8,24)
1134	mov	$tp40,$tp20
1135	rol	\$24,`&LO("$tp88")`	# ROTATE(tp2^tp1^tp8,24)
1136	mov	$tp48,$tp28
1137	shr	\$32,$tp20
1138	xor	`&LO("$tp80")`,`&LO("$acc0")`
1139	shr	\$32,$tp28
1140	xor	`&LO("$tp88")`,`&LO("$acc8")`
1141
1142	`"mov	0($sbox),$mask80"	if ($prefetch)`
1143	rol	\$16,`&LO("$tp40")`	# ROTATE(tp4^tp1^tp8,16)
1144	`"mov	64($sbox),$maskfe"	if ($prefetch)`
1145	rol	\$16,`&LO("$tp48")`	# ROTATE(tp4^tp1^tp8,16)
1146	`"mov	128($sbox),$mask1b"	if ($prefetch)`
1147	rol	\$16,`&LO("$tp20")`	# ROTATE(tp4^tp1^tp8,16)
1148	`"mov	192($sbox),$tp80"	if ($prefetch)`
1149	xor	`&LO("$tp40")`,`&LO("$tp10")`
1150	rol	\$16,`&LO("$tp28")`	# ROTATE(tp4^tp1^tp8,16)
1151	xor	`&LO("$tp48")`,`&LO("$tp18")`
1152	`"mov	256($sbox),$tp88"	if ($prefetch)`
1153	xor	`&LO("$tp20")`,`&LO("$acc0")`
1154	xor	`&LO("$tp28")`,`&LO("$acc8")`
1155___
1156}
1157
1158$code.=<<___;
1159.type	_x86_64_AES_decrypt_compact,\@abi-omnipotent
1160.align	16
1161_x86_64_AES_decrypt_compact:
1162	lea	128($sbox),$inp			# size optimization
1163	mov	0-128($inp),$acc1		# prefetch Td4
1164	mov	32-128($inp),$acc2
1165	mov	64-128($inp),$t0
1166	mov	96-128($inp),$t1
1167	mov	128-128($inp),$acc1
1168	mov	160-128($inp),$acc2
1169	mov	192-128($inp),$t0
1170	mov	224-128($inp),$t1
1171	jmp	.Ldec_loop_compact
1172
1173.align	16
1174.Ldec_loop_compact:
1175		xor	0($key),$s0		# xor with key
1176		xor	4($key),$s1
1177		xor	8($key),$s2
1178		xor	12($key),$s3
1179		lea	16($key),$key
1180___
1181		&deccompactvert();
1182$code.=<<___;
1183		cmp	16(%rsp),$key
1184		je	.Ldec_compact_done
1185
1186		mov	256+0($sbox),$mask80
1187		shl	\$32,%rbx
1188		shl	\$32,%rdx
1189		mov	256+8($sbox),$maskfe
1190		or	%rbx,%rax
1191		or	%rdx,%rcx
1192		mov	256+16($sbox),$mask1b
1193___
1194		&dectransform(1);
1195$code.=<<___;
1196	jmp	.Ldec_loop_compact
1197.align	16
1198.Ldec_compact_done:
1199	xor	0($key),$s0
1200	xor	4($key),$s1
1201	xor	8($key),$s2
1202	xor	12($key),$s3
1203	.byte	0xf3,0xc3			# rep ret
1204.size	_x86_64_AES_decrypt_compact,.-_x86_64_AES_decrypt_compact
1205___
1206
1207# void aes_nohw_decrypt (const void *inp,void *out,const AES_KEY *key);
1208$code.=<<___;
1209.align	16
1210.globl	aes_nohw_decrypt
1211.type	aes_nohw_decrypt,\@function,3
1212.hidden	aes_nohw_decrypt
1213aes_nohw_decrypt:
1214.cfi_startproc
1215	mov	%rsp,%rax
1216.cfi_def_cfa_register	%rax
1217	push	%rbx
1218.cfi_push	%rbx
1219	push	%rbp
1220.cfi_push	%rbp
1221	push	%r12
1222.cfi_push	%r12
1223	push	%r13
1224.cfi_push	%r13
1225	push	%r14
1226.cfi_push	%r14
1227	push	%r15
1228.cfi_push	%r15
1229
1230	# allocate frame "above" key schedule
1231	lea	-63(%rdx),%rcx	# %rdx is key argument
1232	and	\$-64,%rsp
1233	sub	%rsp,%rcx
1234	neg	%rcx
1235	and	\$0x3c0,%rcx
1236	sub	%rcx,%rsp
1237	sub	\$32,%rsp
1238
1239	mov	%rsi,16(%rsp)	# save out
1240	mov	%rax,24(%rsp)	# save original stack pointer
1241.cfi_cfa_expression	%rsp+24,deref,+8
1242.Ldec_prologue:
1243
1244	mov	%rdx,$key
1245	mov	240($key),$rnds	# load rounds
1246
1247	mov	0(%rdi),$s0	# load input vector
1248	mov	4(%rdi),$s1
1249	mov	8(%rdi),$s2
1250	mov	12(%rdi),$s3
1251
1252	shl	\$4,$rnds
1253	lea	($key,$rnds),%rbp
1254	mov	$key,(%rsp)	# key schedule
1255	mov	%rbp,8(%rsp)	# end of key schedule
1256
1257	# pick Td4 copy which can't "overlap" with stack frame or key schedule
1258	lea	.LAES_Td+2048(%rip),$sbox
1259	lea	768(%rsp),%rbp
1260	sub	$sbox,%rbp
1261	and	\$0x300,%rbp
1262	lea	($sbox,%rbp),$sbox
1263	shr	\$3,%rbp	# recall "magic" constants!
1264	add	%rbp,$sbox
1265
1266	call	_x86_64_AES_decrypt_compact
1267
1268	mov	16(%rsp),$out	# restore out
1269	mov	24(%rsp),%rsi	# restore saved stack pointer
1270.cfi_def_cfa	%rsi,8
1271	mov	$s0,0($out)	# write output vector
1272	mov	$s1,4($out)
1273	mov	$s2,8($out)
1274	mov	$s3,12($out)
1275
1276	mov	-48(%rsi),%r15
1277.cfi_restore	%r15
1278	mov	-40(%rsi),%r14
1279.cfi_restore	%r14
1280	mov	-32(%rsi),%r13
1281.cfi_restore	%r13
1282	mov	-24(%rsi),%r12
1283.cfi_restore	%r12
1284	mov	-16(%rsi),%rbp
1285.cfi_restore	%rbp
1286	mov	-8(%rsi),%rbx
1287.cfi_restore	%rbx
1288	lea	(%rsi),%rsp
1289.cfi_def_cfa_register	%rsp
1290.Ldec_epilogue:
1291	ret
1292.cfi_endproc
1293.size	aes_nohw_decrypt,.-aes_nohw_decrypt
1294___
1295#------------------------------------------------------------------#
1296
1297sub enckey()
1298{
1299$code.=<<___;
1300	movz	%dl,%esi		# rk[i]>>0
1301	movzb	-128(%rbp,%rsi),%ebx
1302	movz	%dh,%esi		# rk[i]>>8
1303	shl	\$24,%ebx
1304	xor	%ebx,%eax
1305
1306	movzb	-128(%rbp,%rsi),%ebx
1307	shr	\$16,%edx
1308	movz	%dl,%esi		# rk[i]>>16
1309	xor	%ebx,%eax
1310
1311	movzb	-128(%rbp,%rsi),%ebx
1312	movz	%dh,%esi		# rk[i]>>24
1313	shl	\$8,%ebx
1314	xor	%ebx,%eax
1315
1316	movzb	-128(%rbp,%rsi),%ebx
1317	shl	\$16,%ebx
1318	xor	%ebx,%eax
1319
1320	xor	1024-128(%rbp,%rcx,4),%eax		# rcon
1321___
1322}
1323
1324# int aes_nohw_set_encrypt_key(const unsigned char *userKey, const int bits, AES_KEY *key)
1325$code.=<<___;
1326.align	16
1327.globl aes_nohw_set_encrypt_key
1328.type  aes_nohw_set_encrypt_key,\@function,3
1329aes_nohw_set_encrypt_key:
1330.cfi_startproc
1331	push	%rbx
1332.cfi_push	%rbx
1333	push	%rbp
1334.cfi_push	%rbp
1335	push	%r12			# redundant, but allows to share
1336.cfi_push	%r12
1337	push	%r13			# exception handler...
1338.cfi_push	%r13
1339	push	%r14
1340.cfi_push	%r14
1341	push	%r15
1342.cfi_push	%r15
1343	sub	\$8,%rsp
1344.cfi_adjust_cfa_offset	8
1345.Lenc_key_prologue:
1346
1347	call	_x86_64_AES_set_encrypt_key
1348
1349	mov	40(%rsp),%rbp
1350.cfi_restore	%rbp
1351	mov	48(%rsp),%rbx
1352.cfi_restore	%rbx
1353	add	\$56,%rsp
1354.cfi_adjust_cfa_offset	-56
1355.Lenc_key_epilogue:
1356	ret
1357.cfi_endproc
1358.size aes_nohw_set_encrypt_key,.-aes_nohw_set_encrypt_key
1359
1360.type	_x86_64_AES_set_encrypt_key,\@abi-omnipotent
1361.align	16
1362_x86_64_AES_set_encrypt_key:
1363	mov	%esi,%ecx			# %ecx=bits
1364	mov	%rdi,%rsi			# %rsi=userKey
1365	mov	%rdx,%rdi			# %rdi=key
1366
1367	test	\$-1,%rsi
1368	jz	.Lbadpointer
1369	test	\$-1,%rdi
1370	jz	.Lbadpointer
1371
1372	lea	.LAES_Te(%rip),%rbp
1373	lea	2048+128(%rbp),%rbp
1374
1375	# prefetch Te4
1376	mov	0-128(%rbp),%eax
1377	mov	32-128(%rbp),%ebx
1378	mov	64-128(%rbp),%r8d
1379	mov	96-128(%rbp),%edx
1380	mov	128-128(%rbp),%eax
1381	mov	160-128(%rbp),%ebx
1382	mov	192-128(%rbp),%r8d
1383	mov	224-128(%rbp),%edx
1384
1385	cmp	\$128,%ecx
1386	je	.L10rounds
1387	cmp	\$192,%ecx
1388	je	.L12rounds
1389	cmp	\$256,%ecx
1390	je	.L14rounds
1391	mov	\$-2,%rax			# invalid number of bits
1392	jmp	.Lexit
1393
1394.L10rounds:
1395	mov	0(%rsi),%rax			# copy first 4 dwords
1396	mov	8(%rsi),%rdx
1397	mov	%rax,0(%rdi)
1398	mov	%rdx,8(%rdi)
1399
1400	shr	\$32,%rdx
1401	xor	%ecx,%ecx
1402	jmp	.L10shortcut
1403.align	4
1404.L10loop:
1405		mov	0(%rdi),%eax			# rk[0]
1406		mov	12(%rdi),%edx			# rk[3]
1407.L10shortcut:
1408___
1409		&enckey	();
1410$code.=<<___;
1411		mov	%eax,16(%rdi)			# rk[4]
1412		xor	4(%rdi),%eax
1413		mov	%eax,20(%rdi)			# rk[5]
1414		xor	8(%rdi),%eax
1415		mov	%eax,24(%rdi)			# rk[6]
1416		xor	12(%rdi),%eax
1417		mov	%eax,28(%rdi)			# rk[7]
1418		add	\$1,%ecx
1419		lea	16(%rdi),%rdi
1420		cmp	\$10,%ecx
1421	jl	.L10loop
1422
1423	movl	\$10,80(%rdi)			# setup number of rounds
1424	xor	%rax,%rax
1425	jmp	.Lexit
1426
1427.L12rounds:
1428	mov	0(%rsi),%rax			# copy first 6 dwords
1429	mov	8(%rsi),%rbx
1430	mov	16(%rsi),%rdx
1431	mov	%rax,0(%rdi)
1432	mov	%rbx,8(%rdi)
1433	mov	%rdx,16(%rdi)
1434
1435	shr	\$32,%rdx
1436	xor	%ecx,%ecx
1437	jmp	.L12shortcut
1438.align	4
1439.L12loop:
1440		mov	0(%rdi),%eax			# rk[0]
1441		mov	20(%rdi),%edx			# rk[5]
1442.L12shortcut:
1443___
1444		&enckey	();
1445$code.=<<___;
1446		mov	%eax,24(%rdi)			# rk[6]
1447		xor	4(%rdi),%eax
1448		mov	%eax,28(%rdi)			# rk[7]
1449		xor	8(%rdi),%eax
1450		mov	%eax,32(%rdi)			# rk[8]
1451		xor	12(%rdi),%eax
1452		mov	%eax,36(%rdi)			# rk[9]
1453
1454		cmp	\$7,%ecx
1455		je	.L12break
1456		add	\$1,%ecx
1457
1458		xor	16(%rdi),%eax
1459		mov	%eax,40(%rdi)			# rk[10]
1460		xor	20(%rdi),%eax
1461		mov	%eax,44(%rdi)			# rk[11]
1462
1463		lea	24(%rdi),%rdi
1464	jmp	.L12loop
1465.L12break:
1466	movl	\$12,72(%rdi)		# setup number of rounds
1467	xor	%rax,%rax
1468	jmp	.Lexit
1469
1470.L14rounds:
1471	mov	0(%rsi),%rax			# copy first 8 dwords
1472	mov	8(%rsi),%rbx
1473	mov	16(%rsi),%rcx
1474	mov	24(%rsi),%rdx
1475	mov	%rax,0(%rdi)
1476	mov	%rbx,8(%rdi)
1477	mov	%rcx,16(%rdi)
1478	mov	%rdx,24(%rdi)
1479
1480	shr	\$32,%rdx
1481	xor	%ecx,%ecx
1482	jmp	.L14shortcut
1483.align	4
1484.L14loop:
1485		mov	0(%rdi),%eax			# rk[0]
1486		mov	28(%rdi),%edx			# rk[4]
1487.L14shortcut:
1488___
1489		&enckey	();
1490$code.=<<___;
1491		mov	%eax,32(%rdi)			# rk[8]
1492		xor	4(%rdi),%eax
1493		mov	%eax,36(%rdi)			# rk[9]
1494		xor	8(%rdi),%eax
1495		mov	%eax,40(%rdi)			# rk[10]
1496		xor	12(%rdi),%eax
1497		mov	%eax,44(%rdi)			# rk[11]
1498
1499		cmp	\$6,%ecx
1500		je	.L14break
1501		add	\$1,%ecx
1502
1503		mov	%eax,%edx
1504		mov	16(%rdi),%eax			# rk[4]
1505		movz	%dl,%esi			# rk[11]>>0
1506		movzb	-128(%rbp,%rsi),%ebx
1507		movz	%dh,%esi			# rk[11]>>8
1508		xor	%ebx,%eax
1509
1510		movzb	-128(%rbp,%rsi),%ebx
1511		shr	\$16,%edx
1512		shl	\$8,%ebx
1513		movz	%dl,%esi			# rk[11]>>16
1514		xor	%ebx,%eax
1515
1516		movzb	-128(%rbp,%rsi),%ebx
1517		movz	%dh,%esi			# rk[11]>>24
1518		shl	\$16,%ebx
1519		xor	%ebx,%eax
1520
1521		movzb	-128(%rbp,%rsi),%ebx
1522		shl	\$24,%ebx
1523		xor	%ebx,%eax
1524
1525		mov	%eax,48(%rdi)			# rk[12]
1526		xor	20(%rdi),%eax
1527		mov	%eax,52(%rdi)			# rk[13]
1528		xor	24(%rdi),%eax
1529		mov	%eax,56(%rdi)			# rk[14]
1530		xor	28(%rdi),%eax
1531		mov	%eax,60(%rdi)			# rk[15]
1532
1533		lea	32(%rdi),%rdi
1534	jmp	.L14loop
1535.L14break:
1536	movl	\$14,48(%rdi)		# setup number of rounds
1537	xor	%rax,%rax
1538	jmp	.Lexit
1539
1540.Lbadpointer:
1541	mov	\$-1,%rax
1542.Lexit:
1543	.byte	0xf3,0xc3			# rep ret
1544.size	_x86_64_AES_set_encrypt_key,.-_x86_64_AES_set_encrypt_key
1545___
1546
1547sub deckey_ref()
1548{ my ($i,$ptr,$te,$td) = @_;
1549  my ($tp1,$tp2,$tp4,$tp8,$acc)=("%eax","%ebx","%edi","%edx","%r8d");
1550$code.=<<___;
1551	mov	$i($ptr),$tp1
1552	mov	$tp1,$acc
1553	and	\$0x80808080,$acc
1554	mov	$acc,$tp4
1555	shr	\$7,$tp4
1556	lea	0($tp1,$tp1),$tp2
1557	sub	$tp4,$acc
1558	and	\$0xfefefefe,$tp2
1559	and	\$0x1b1b1b1b,$acc
1560	xor	$tp2,$acc
1561	mov	$acc,$tp2
1562
1563	and	\$0x80808080,$acc
1564	mov	$acc,$tp8
1565	shr	\$7,$tp8
1566	lea	0($tp2,$tp2),$tp4
1567	sub	$tp8,$acc
1568	and	\$0xfefefefe,$tp4
1569	and	\$0x1b1b1b1b,$acc
1570	 xor	$tp1,$tp2		# tp2^tp1
1571	xor	$tp4,$acc
1572	mov	$acc,$tp4
1573
1574	and	\$0x80808080,$acc
1575	mov	$acc,$tp8
1576	shr	\$7,$tp8
1577	sub	$tp8,$acc
1578	lea	0($tp4,$tp4),$tp8
1579	 xor	$tp1,$tp4		# tp4^tp1
1580	and	\$0xfefefefe,$tp8
1581	and	\$0x1b1b1b1b,$acc
1582	xor	$acc,$tp8
1583
1584	xor	$tp8,$tp1		# tp1^tp8
1585	rol	\$8,$tp1		# ROTATE(tp1^tp8,8)
1586	xor	$tp8,$tp2		# tp2^tp1^tp8
1587	xor	$tp8,$tp4		# tp4^tp1^tp8
1588	xor	$tp2,$tp8
1589	xor	$tp4,$tp8		# tp8^(tp8^tp4^tp1)^(tp8^tp2^tp1)=tp8^tp4^tp2
1590
1591	xor	$tp8,$tp1
1592	rol	\$24,$tp2		# ROTATE(tp2^tp1^tp8,24)
1593	xor	$tp2,$tp1
1594	rol	\$16,$tp4		# ROTATE(tp4^tp1^tp8,16)
1595	xor	$tp4,$tp1
1596
1597	mov	$tp1,$i($ptr)
1598___
1599}
1600
1601# int aes_nohw_set_decrypt_key(const unsigned char *userKey, const int bits, AES_KEY *key)
1602$code.=<<___;
1603.align	16
1604.globl aes_nohw_set_decrypt_key
1605.type  aes_nohw_set_decrypt_key,\@function,3
1606aes_nohw_set_decrypt_key:
1607.cfi_startproc
1608	push	%rbx
1609.cfi_push	%rbx
1610	push	%rbp
1611.cfi_push	%rbp
1612	push	%r12
1613.cfi_push	%r12
1614	push	%r13
1615.cfi_push	%r13
1616	push	%r14
1617.cfi_push	%r14
1618	push	%r15
1619.cfi_push	%r15
1620	push	%rdx			# save key schedule
1621.cfi_adjust_cfa_offset	8
1622.Ldec_key_prologue:
1623
1624	call	_x86_64_AES_set_encrypt_key
1625	mov	(%rsp),%r8		# restore key schedule
1626	cmp	\$0,%eax
1627	jne	.Labort
1628
1629	mov	240(%r8),%r14d		# pull number of rounds
1630	xor	%rdi,%rdi
1631	lea	(%rdi,%r14d,4),%rcx
1632	mov	%r8,%rsi
1633	lea	(%r8,%rcx,4),%rdi	# pointer to last chunk
1634.align	4
1635.Linvert:
1636		mov	0(%rsi),%rax
1637		mov	8(%rsi),%rbx
1638		mov	0(%rdi),%rcx
1639		mov	8(%rdi),%rdx
1640		mov	%rax,0(%rdi)
1641		mov	%rbx,8(%rdi)
1642		mov	%rcx,0(%rsi)
1643		mov	%rdx,8(%rsi)
1644		lea	16(%rsi),%rsi
1645		lea	-16(%rdi),%rdi
1646		cmp	%rsi,%rdi
1647	jne	.Linvert
1648
1649	lea	.LAES_Te+2048+1024(%rip),%rax	# rcon
1650
1651	mov	40(%rax),$mask80
1652	mov	48(%rax),$maskfe
1653	mov	56(%rax),$mask1b
1654
1655	mov	%r8,$key
1656	sub	\$1,%r14d
1657.align	4
1658.Lpermute:
1659		lea	16($key),$key
1660		mov	0($key),%rax
1661		mov	8($key),%rcx
1662___
1663		&dectransform ();
1664$code.=<<___;
1665		mov	%eax,0($key)
1666		mov	%ebx,4($key)
1667		mov	%ecx,8($key)
1668		mov	%edx,12($key)
1669		sub	\$1,%r14d
1670	jnz	.Lpermute
1671
1672	xor	%rax,%rax
1673.Labort:
1674	mov	8(%rsp),%r15
1675.cfi_restore	%r15
1676	mov	16(%rsp),%r14
1677.cfi_restore	%r14
1678	mov	24(%rsp),%r13
1679.cfi_restore	%r13
1680	mov	32(%rsp),%r12
1681.cfi_restore	%r12
1682	mov	40(%rsp),%rbp
1683.cfi_restore	%rbp
1684	mov	48(%rsp),%rbx
1685.cfi_restore	%rbx
1686	add	\$56,%rsp
1687.cfi_adjust_cfa_offset	-56
1688.Ldec_key_epilogue:
1689	ret
1690.cfi_endproc
1691.size	aes_nohw_set_decrypt_key,.-aes_nohw_set_decrypt_key
1692___
1693
1694# void aes_nohw_cbc_encrypt (const void char *inp, unsigned char *out,
1695#			                       size_t length, const AES_KEY *key,
1696#			                       unsigned char *ivp,const int enc);
1697{
1698# stack frame layout
1699# -8(%rsp)		return address
1700my $keyp="0(%rsp)";		# one to pass as $key
1701my $keyend="8(%rsp)";		# &(keyp->rd_key[4*keyp->rounds])
1702my $_rsp="16(%rsp)";		# saved %rsp
1703my $_inp="24(%rsp)";		# copy of 1st parameter, inp
1704my $_out="32(%rsp)";		# copy of 2nd parameter, out
1705my $_len="40(%rsp)";		# copy of 3rd parameter, length
1706my $_key="48(%rsp)";		# copy of 4th parameter, key
1707my $_ivp="56(%rsp)";		# copy of 5th parameter, ivp
1708my $ivec="64(%rsp)";		# ivec[16]
1709my $aes_key="80(%rsp)";		# copy of aes_key
1710my $mark="80+240(%rsp)";	# copy of aes_key->rounds
1711
1712$code.=<<___;
1713.align	16
1714.globl	aes_nohw_cbc_encrypt
1715.type	aes_nohw_cbc_encrypt,\@function,6
1716.extern	OPENSSL_ia32cap_P
1717.hidden	aes_nohw_cbc_encrypt
1718aes_nohw_cbc_encrypt:
1719.cfi_startproc
1720	cmp	\$0,%rdx	# check length
1721	je	.Lcbc_epilogue
1722	pushfq
1723.cfi_push	49		# %rflags
1724	push	%rbx
1725.cfi_push	%rbx
1726	push	%rbp
1727.cfi_push	%rbp
1728	push	%r12
1729.cfi_push	%r12
1730	push	%r13
1731.cfi_push	%r13
1732	push	%r14
1733.cfi_push	%r14
1734	push	%r15
1735.cfi_push	%r15
1736.Lcbc_prologue:
1737
1738	cld
1739	mov	%r9d,%r9d	# clear upper half of enc
1740
1741	lea	.LAES_Te(%rip),$sbox
1742	lea	.LAES_Td(%rip),%r10
1743	cmp	\$0,%r9
1744	cmoveq	%r10,$sbox
1745
1746	leaq	OPENSSL_ia32cap_P(%rip),%r10
1747	mov	(%r10), %r10d
1748	cmp	\$$speed_limit,%rdx
1749	jb	.Lcbc_slow_prologue
1750	test	\$15,%rdx
1751	jnz	.Lcbc_slow_prologue
1752	bt	\$28,%r10d
1753	jc	.Lcbc_slow_prologue
1754
1755	# allocate aligned stack frame...
1756	lea	-88-248(%rsp),$key
1757	and	\$-64,$key
1758
1759	# ... and make sure it doesn't alias with AES_T[ed] modulo 4096
1760	mov	$sbox,%r10
1761	lea	2304($sbox),%r11
1762	mov	$key,%r12
1763	and	\$0xFFF,%r10	# s = $sbox&0xfff
1764	and	\$0xFFF,%r11	# e = ($sbox+2048)&0xfff
1765	and	\$0xFFF,%r12	# p = %rsp&0xfff
1766
1767	cmp	%r11,%r12	# if (p=>e) %rsp =- (p-e);
1768	jb	.Lcbc_te_break_out
1769	sub	%r11,%r12
1770	sub	%r12,$key
1771	jmp	.Lcbc_te_ok
1772.Lcbc_te_break_out:		# else %rsp -= (p-s)&0xfff + framesz
1773	sub	%r10,%r12
1774	and	\$0xFFF,%r12
1775	add	\$320,%r12
1776	sub	%r12,$key
1777.align	4
1778.Lcbc_te_ok:
1779
1780	xchg	%rsp,$key
1781.cfi_def_cfa_register	$key
1782	#add	\$8,%rsp	# reserve for return address!
1783	mov	$key,$_rsp	# save %rsp
1784.cfi_cfa_expression	$_rsp,deref,+64
1785.Lcbc_fast_body:
1786	mov	%rdi,$_inp	# save copy of inp
1787	mov	%rsi,$_out	# save copy of out
1788	mov	%rdx,$_len	# save copy of len
1789	mov	%rcx,$_key	# save copy of key
1790	mov	%r8,$_ivp	# save copy of ivp
1791	movl	\$0,$mark	# copy of aes_key->rounds = 0;
1792	mov	%r8,%rbp	# rearrange input arguments
1793	mov	%r9,%rbx
1794	mov	%rsi,$out
1795	mov	%rdi,$inp
1796	mov	%rcx,$key
1797
1798	mov	240($key),%eax		# key->rounds
1799	# do we copy key schedule to stack?
1800	mov	$key,%r10
1801	sub	$sbox,%r10
1802	and	\$0xfff,%r10
1803	cmp	\$2304,%r10
1804	jb	.Lcbc_do_ecopy
1805	cmp	\$4096-248,%r10
1806	jb	.Lcbc_skip_ecopy
1807.align	4
1808.Lcbc_do_ecopy:
1809		mov	$key,%rsi
1810		lea	$aes_key,%rdi
1811		lea	$aes_key,$key
1812		mov	\$240/8,%ecx
1813		.long	0x90A548F3	# rep movsq
1814		mov	%eax,(%rdi)	# copy aes_key->rounds
1815.Lcbc_skip_ecopy:
1816	mov	$key,$keyp	# save key pointer
1817
1818	mov	\$18,%ecx
1819.align	4
1820.Lcbc_prefetch_te:
1821		mov	0($sbox),%r10
1822		mov	32($sbox),%r11
1823		mov	64($sbox),%r12
1824		mov	96($sbox),%r13
1825		lea	128($sbox),$sbox
1826		sub	\$1,%ecx
1827	jnz	.Lcbc_prefetch_te
1828	lea	-2304($sbox),$sbox
1829
1830	cmp	\$0,%rbx
1831	je	.LFAST_DECRYPT
1832
1833#----------------------------- ENCRYPT -----------------------------#
1834	mov	0(%rbp),$s0		# load iv
1835	mov	4(%rbp),$s1
1836	mov	8(%rbp),$s2
1837	mov	12(%rbp),$s3
1838
1839.align	4
1840.Lcbc_fast_enc_loop:
1841		xor	0($inp),$s0
1842		xor	4($inp),$s1
1843		xor	8($inp),$s2
1844		xor	12($inp),$s3
1845		mov	$keyp,$key	# restore key
1846		mov	$inp,$_inp	# if ($verticalspin) save inp
1847
1848		call	_x86_64_AES_encrypt
1849
1850		mov	$_inp,$inp	# if ($verticalspin) restore inp
1851		mov	$_len,%r10
1852		mov	$s0,0($out)
1853		mov	$s1,4($out)
1854		mov	$s2,8($out)
1855		mov	$s3,12($out)
1856
1857		lea	16($inp),$inp
1858		lea	16($out),$out
1859		sub	\$16,%r10
1860		test	\$-16,%r10
1861		mov	%r10,$_len
1862	jnz	.Lcbc_fast_enc_loop
1863	mov	$_ivp,%rbp	# restore ivp
1864	mov	$s0,0(%rbp)	# save ivec
1865	mov	$s1,4(%rbp)
1866	mov	$s2,8(%rbp)
1867	mov	$s3,12(%rbp)
1868
1869	jmp	.Lcbc_fast_cleanup
1870
1871#----------------------------- DECRYPT -----------------------------#
1872.align	16
1873.LFAST_DECRYPT:
1874	cmp	$inp,$out
1875	je	.Lcbc_fast_dec_in_place
1876
1877	mov	%rbp,$ivec
1878.align	4
1879.Lcbc_fast_dec_loop:
1880		mov	0($inp),$s0	# read input
1881		mov	4($inp),$s1
1882		mov	8($inp),$s2
1883		mov	12($inp),$s3
1884		mov	$keyp,$key	# restore key
1885		mov	$inp,$_inp	# if ($verticalspin) save inp
1886
1887		call	_x86_64_AES_decrypt
1888
1889		mov	$ivec,%rbp	# load ivp
1890		mov	$_inp,$inp	# if ($verticalspin) restore inp
1891		mov	$_len,%r10	# load len
1892		xor	0(%rbp),$s0	# xor iv
1893		xor	4(%rbp),$s1
1894		xor	8(%rbp),$s2
1895		xor	12(%rbp),$s3
1896		mov	$inp,%rbp	# current input, next iv
1897
1898		sub	\$16,%r10
1899		mov	%r10,$_len	# update len
1900		mov	%rbp,$ivec	# update ivp
1901
1902		mov	$s0,0($out)	# write output
1903		mov	$s1,4($out)
1904		mov	$s2,8($out)
1905		mov	$s3,12($out)
1906
1907		lea	16($inp),$inp
1908		lea	16($out),$out
1909	jnz	.Lcbc_fast_dec_loop
1910	mov	$_ivp,%r12		# load user ivp
1911	mov	0(%rbp),%r10		# load iv
1912	mov	8(%rbp),%r11
1913	mov	%r10,0(%r12)		# copy back to user
1914	mov	%r11,8(%r12)
1915	jmp	.Lcbc_fast_cleanup
1916
1917.align	16
1918.Lcbc_fast_dec_in_place:
1919	mov	0(%rbp),%r10		# copy iv to stack
1920	mov	8(%rbp),%r11
1921	mov	%r10,0+$ivec
1922	mov	%r11,8+$ivec
1923.align	4
1924.Lcbc_fast_dec_in_place_loop:
1925		mov	0($inp),$s0	# load input
1926		mov	4($inp),$s1
1927		mov	8($inp),$s2
1928		mov	12($inp),$s3
1929		mov	$keyp,$key	# restore key
1930		mov	$inp,$_inp	# if ($verticalspin) save inp
1931
1932		call	_x86_64_AES_decrypt
1933
1934		mov	$_inp,$inp	# if ($verticalspin) restore inp
1935		mov	$_len,%r10
1936		xor	0+$ivec,$s0
1937		xor	4+$ivec,$s1
1938		xor	8+$ivec,$s2
1939		xor	12+$ivec,$s3
1940
1941		mov	0($inp),%r11	# load input
1942		mov	8($inp),%r12
1943		sub	\$16,%r10
1944		jz	.Lcbc_fast_dec_in_place_done
1945
1946		mov	%r11,0+$ivec	# copy input to iv
1947		mov	%r12,8+$ivec
1948
1949		mov	$s0,0($out)	# save output [zaps input]
1950		mov	$s1,4($out)
1951		mov	$s2,8($out)
1952		mov	$s3,12($out)
1953
1954		lea	16($inp),$inp
1955		lea	16($out),$out
1956		mov	%r10,$_len
1957	jmp	.Lcbc_fast_dec_in_place_loop
1958.Lcbc_fast_dec_in_place_done:
1959	mov	$_ivp,%rdi
1960	mov	%r11,0(%rdi)	# copy iv back to user
1961	mov	%r12,8(%rdi)
1962
1963	mov	$s0,0($out)	# save output [zaps input]
1964	mov	$s1,4($out)
1965	mov	$s2,8($out)
1966	mov	$s3,12($out)
1967
1968.align	4
1969.Lcbc_fast_cleanup:
1970	cmpl	\$0,$mark	# was the key schedule copied?
1971	lea	$aes_key,%rdi
1972	je	.Lcbc_exit
1973		mov	\$240/8,%ecx
1974		xor	%rax,%rax
1975		.long	0x90AB48F3	# rep stosq
1976
1977	jmp	.Lcbc_exit
1978
1979#--------------------------- SLOW ROUTINE ---------------------------#
1980.align	16
1981.Lcbc_slow_prologue:
1982	# allocate aligned stack frame...
1983	lea	-88(%rsp),%rbp
1984	and	\$-64,%rbp
1985	# ... just "above" key schedule
1986	lea	-88-63(%rcx),%r10
1987	sub	%rbp,%r10
1988	neg	%r10
1989	and	\$0x3c0,%r10
1990	sub	%r10,%rbp
1991
1992	xchg	%rsp,%rbp
1993	#add	\$8,%rsp	# reserve for return address!
1994	mov	%rbp,$_rsp	# save %rsp
1995.Lcbc_slow_body:
1996	#mov	%rdi,$_inp	# save copy of inp
1997	#mov	%rsi,$_out	# save copy of out
1998	#mov	%rdx,$_len	# save copy of len
1999	#mov	%rcx,$_key	# save copy of key
2000	mov	%r8,$_ivp	# save copy of ivp
2001	mov	%r8,%rbp	# rearrange input arguments
2002	mov	%r9,%rbx
2003	mov	%rsi,$out
2004	mov	%rdi,$inp
2005	mov	%rcx,$key
2006	mov	%rdx,%r10
2007
2008	mov	240($key),%eax
2009	mov	$key,$keyp	# save key pointer
2010	shl	\$4,%eax
2011	lea	($key,%rax),%rax
2012	mov	%rax,$keyend
2013
2014	# pick Te4 copy which can't "overlap" with stack frame or key schedule
2015	lea	2048($sbox),$sbox
2016	lea	768-8(%rsp),%rax
2017	sub	$sbox,%rax
2018	and	\$0x300,%rax
2019	lea	($sbox,%rax),$sbox
2020
2021	cmp	\$0,%rbx
2022	je	.LSLOW_DECRYPT
2023
2024#--------------------------- SLOW ENCRYPT ---------------------------#
2025	test	\$-16,%r10		# check upon length
2026	mov	0(%rbp),$s0		# load iv
2027	mov	4(%rbp),$s1
2028	mov	8(%rbp),$s2
2029	mov	12(%rbp),$s3
2030	jz	.Lcbc_slow_enc_tail	# short input...
2031
2032.align	4
2033.Lcbc_slow_enc_loop:
2034		xor	0($inp),$s0
2035		xor	4($inp),$s1
2036		xor	8($inp),$s2
2037		xor	12($inp),$s3
2038		mov	$keyp,$key	# restore key
2039		mov	$inp,$_inp	# save inp
2040		mov	$out,$_out	# save out
2041		mov	%r10,$_len	# save len
2042
2043		call	_x86_64_AES_encrypt_compact
2044
2045		mov	$_inp,$inp	# restore inp
2046		mov	$_out,$out	# restore out
2047		mov	$_len,%r10	# restore len
2048		mov	$s0,0($out)
2049		mov	$s1,4($out)
2050		mov	$s2,8($out)
2051		mov	$s3,12($out)
2052
2053		lea	16($inp),$inp
2054		lea	16($out),$out
2055		sub	\$16,%r10
2056		test	\$-16,%r10
2057	jnz	.Lcbc_slow_enc_loop
2058	test	\$15,%r10
2059	jnz	.Lcbc_slow_enc_tail
2060	mov	$_ivp,%rbp	# restore ivp
2061	mov	$s0,0(%rbp)	# save ivec
2062	mov	$s1,4(%rbp)
2063	mov	$s2,8(%rbp)
2064	mov	$s3,12(%rbp)
2065
2066	jmp	.Lcbc_exit
2067
2068.align	4
2069.Lcbc_slow_enc_tail:
2070	mov	%rax,%r11
2071	mov	%rcx,%r12
2072	mov	%r10,%rcx
2073	mov	$inp,%rsi
2074	mov	$out,%rdi
2075	.long	0x9066A4F3		# rep movsb
2076	mov	\$16,%rcx		# zero tail
2077	sub	%r10,%rcx
2078	xor	%rax,%rax
2079	.long	0x9066AAF3		# rep stosb
2080	mov	$out,$inp		# this is not a mistake!
2081	mov	\$16,%r10		# len=16
2082	mov	%r11,%rax
2083	mov	%r12,%rcx
2084	jmp	.Lcbc_slow_enc_loop	# one more spin...
2085#--------------------------- SLOW DECRYPT ---------------------------#
2086.align	16
2087.LSLOW_DECRYPT:
2088	shr	\$3,%rax
2089	add	%rax,$sbox		# recall "magic" constants!
2090
2091	mov	0(%rbp),%r11		# copy iv to stack
2092	mov	8(%rbp),%r12
2093	mov	%r11,0+$ivec
2094	mov	%r12,8+$ivec
2095
2096.align	4
2097.Lcbc_slow_dec_loop:
2098		mov	0($inp),$s0	# load input
2099		mov	4($inp),$s1
2100		mov	8($inp),$s2
2101		mov	12($inp),$s3
2102		mov	$keyp,$key	# restore key
2103		mov	$inp,$_inp	# save inp
2104		mov	$out,$_out	# save out
2105		mov	%r10,$_len	# save len
2106
2107		call	_x86_64_AES_decrypt_compact
2108
2109		mov	$_inp,$inp	# restore inp
2110		mov	$_out,$out	# restore out
2111		mov	$_len,%r10
2112		xor	0+$ivec,$s0
2113		xor	4+$ivec,$s1
2114		xor	8+$ivec,$s2
2115		xor	12+$ivec,$s3
2116
2117		mov	0($inp),%r11	# load input
2118		mov	8($inp),%r12
2119		sub	\$16,%r10
2120		jc	.Lcbc_slow_dec_partial
2121		jz	.Lcbc_slow_dec_done
2122
2123		mov	%r11,0+$ivec	# copy input to iv
2124		mov	%r12,8+$ivec
2125
2126		mov	$s0,0($out)	# save output [can zap input]
2127		mov	$s1,4($out)
2128		mov	$s2,8($out)
2129		mov	$s3,12($out)
2130
2131		lea	16($inp),$inp
2132		lea	16($out),$out
2133	jmp	.Lcbc_slow_dec_loop
2134.Lcbc_slow_dec_done:
2135	mov	$_ivp,%rdi
2136	mov	%r11,0(%rdi)		# copy iv back to user
2137	mov	%r12,8(%rdi)
2138
2139	mov	$s0,0($out)		# save output [can zap input]
2140	mov	$s1,4($out)
2141	mov	$s2,8($out)
2142	mov	$s3,12($out)
2143
2144	jmp	.Lcbc_exit
2145
2146.align	4
2147.Lcbc_slow_dec_partial:
2148	mov	$_ivp,%rdi
2149	mov	%r11,0(%rdi)		# copy iv back to user
2150	mov	%r12,8(%rdi)
2151
2152	mov	$s0,0+$ivec		# save output to stack
2153	mov	$s1,4+$ivec
2154	mov	$s2,8+$ivec
2155	mov	$s3,12+$ivec
2156
2157	mov	$out,%rdi
2158	lea	$ivec,%rsi
2159	lea	16(%r10),%rcx
2160	.long	0x9066A4F3	# rep movsb
2161	jmp	.Lcbc_exit
2162
2163.align	16
2164.Lcbc_exit:
2165	mov	$_rsp,%rsi
2166.cfi_def_cfa	%rsi,64
2167	mov	(%rsi),%r15
2168.cfi_restore	%r15
2169	mov	8(%rsi),%r14
2170.cfi_restore	%r14
2171	mov	16(%rsi),%r13
2172.cfi_restore	%r13
2173	mov	24(%rsi),%r12
2174.cfi_restore	%r12
2175	mov	32(%rsi),%rbp
2176.cfi_restore	%rbp
2177	mov	40(%rsi),%rbx
2178.cfi_restore	%rbx
2179	lea	48(%rsi),%rsp
2180.cfi_def_cfa	%rsp,16
2181.Lcbc_popfq:
2182	popfq
2183.cfi_pop	49		# %rflags
2184.Lcbc_epilogue:
2185	ret
2186.cfi_endproc
2187.size	aes_nohw_cbc_encrypt,.-aes_nohw_cbc_encrypt
2188___
2189}
2190
2191$code.=<<___;
2192.align	64
2193.LAES_Te:
2194___
2195	&_data_word(0xa56363c6, 0x847c7cf8, 0x997777ee, 0x8d7b7bf6);
2196	&_data_word(0x0df2f2ff, 0xbd6b6bd6, 0xb16f6fde, 0x54c5c591);
2197	&_data_word(0x50303060, 0x03010102, 0xa96767ce, 0x7d2b2b56);
2198	&_data_word(0x19fefee7, 0x62d7d7b5, 0xe6abab4d, 0x9a7676ec);
2199	&_data_word(0x45caca8f, 0x9d82821f, 0x40c9c989, 0x877d7dfa);
2200	&_data_word(0x15fafaef, 0xeb5959b2, 0xc947478e, 0x0bf0f0fb);
2201	&_data_word(0xecadad41, 0x67d4d4b3, 0xfda2a25f, 0xeaafaf45);
2202	&_data_word(0xbf9c9c23, 0xf7a4a453, 0x967272e4, 0x5bc0c09b);
2203	&_data_word(0xc2b7b775, 0x1cfdfde1, 0xae93933d, 0x6a26264c);
2204	&_data_word(0x5a36366c, 0x413f3f7e, 0x02f7f7f5, 0x4fcccc83);
2205	&_data_word(0x5c343468, 0xf4a5a551, 0x34e5e5d1, 0x08f1f1f9);
2206	&_data_word(0x937171e2, 0x73d8d8ab, 0x53313162, 0x3f15152a);
2207	&_data_word(0x0c040408, 0x52c7c795, 0x65232346, 0x5ec3c39d);
2208	&_data_word(0x28181830, 0xa1969637, 0x0f05050a, 0xb59a9a2f);
2209	&_data_word(0x0907070e, 0x36121224, 0x9b80801b, 0x3de2e2df);
2210	&_data_word(0x26ebebcd, 0x6927274e, 0xcdb2b27f, 0x9f7575ea);
2211	&_data_word(0x1b090912, 0x9e83831d, 0x742c2c58, 0x2e1a1a34);
2212	&_data_word(0x2d1b1b36, 0xb26e6edc, 0xee5a5ab4, 0xfba0a05b);
2213	&_data_word(0xf65252a4, 0x4d3b3b76, 0x61d6d6b7, 0xceb3b37d);
2214	&_data_word(0x7b292952, 0x3ee3e3dd, 0x712f2f5e, 0x97848413);
2215	&_data_word(0xf55353a6, 0x68d1d1b9, 0x00000000, 0x2cededc1);
2216	&_data_word(0x60202040, 0x1ffcfce3, 0xc8b1b179, 0xed5b5bb6);
2217	&_data_word(0xbe6a6ad4, 0x46cbcb8d, 0xd9bebe67, 0x4b393972);
2218	&_data_word(0xde4a4a94, 0xd44c4c98, 0xe85858b0, 0x4acfcf85);
2219	&_data_word(0x6bd0d0bb, 0x2aefefc5, 0xe5aaaa4f, 0x16fbfbed);
2220	&_data_word(0xc5434386, 0xd74d4d9a, 0x55333366, 0x94858511);
2221	&_data_word(0xcf45458a, 0x10f9f9e9, 0x06020204, 0x817f7ffe);
2222	&_data_word(0xf05050a0, 0x443c3c78, 0xba9f9f25, 0xe3a8a84b);
2223	&_data_word(0xf35151a2, 0xfea3a35d, 0xc0404080, 0x8a8f8f05);
2224	&_data_word(0xad92923f, 0xbc9d9d21, 0x48383870, 0x04f5f5f1);
2225	&_data_word(0xdfbcbc63, 0xc1b6b677, 0x75dadaaf, 0x63212142);
2226	&_data_word(0x30101020, 0x1affffe5, 0x0ef3f3fd, 0x6dd2d2bf);
2227	&_data_word(0x4ccdcd81, 0x140c0c18, 0x35131326, 0x2fececc3);
2228	&_data_word(0xe15f5fbe, 0xa2979735, 0xcc444488, 0x3917172e);
2229	&_data_word(0x57c4c493, 0xf2a7a755, 0x827e7efc, 0x473d3d7a);
2230	&_data_word(0xac6464c8, 0xe75d5dba, 0x2b191932, 0x957373e6);
2231	&_data_word(0xa06060c0, 0x98818119, 0xd14f4f9e, 0x7fdcdca3);
2232	&_data_word(0x66222244, 0x7e2a2a54, 0xab90903b, 0x8388880b);
2233	&_data_word(0xca46468c, 0x29eeeec7, 0xd3b8b86b, 0x3c141428);
2234	&_data_word(0x79dedea7, 0xe25e5ebc, 0x1d0b0b16, 0x76dbdbad);
2235	&_data_word(0x3be0e0db, 0x56323264, 0x4e3a3a74, 0x1e0a0a14);
2236	&_data_word(0xdb494992, 0x0a06060c, 0x6c242448, 0xe45c5cb8);
2237	&_data_word(0x5dc2c29f, 0x6ed3d3bd, 0xefacac43, 0xa66262c4);
2238	&_data_word(0xa8919139, 0xa4959531, 0x37e4e4d3, 0x8b7979f2);
2239	&_data_word(0x32e7e7d5, 0x43c8c88b, 0x5937376e, 0xb76d6dda);
2240	&_data_word(0x8c8d8d01, 0x64d5d5b1, 0xd24e4e9c, 0xe0a9a949);
2241	&_data_word(0xb46c6cd8, 0xfa5656ac, 0x07f4f4f3, 0x25eaeacf);
2242	&_data_word(0xaf6565ca, 0x8e7a7af4, 0xe9aeae47, 0x18080810);
2243	&_data_word(0xd5baba6f, 0x887878f0, 0x6f25254a, 0x722e2e5c);
2244	&_data_word(0x241c1c38, 0xf1a6a657, 0xc7b4b473, 0x51c6c697);
2245	&_data_word(0x23e8e8cb, 0x7cdddda1, 0x9c7474e8, 0x211f1f3e);
2246	&_data_word(0xdd4b4b96, 0xdcbdbd61, 0x868b8b0d, 0x858a8a0f);
2247	&_data_word(0x907070e0, 0x423e3e7c, 0xc4b5b571, 0xaa6666cc);
2248	&_data_word(0xd8484890, 0x05030306, 0x01f6f6f7, 0x120e0e1c);
2249	&_data_word(0xa36161c2, 0x5f35356a, 0xf95757ae, 0xd0b9b969);
2250	&_data_word(0x91868617, 0x58c1c199, 0x271d1d3a, 0xb99e9e27);
2251	&_data_word(0x38e1e1d9, 0x13f8f8eb, 0xb398982b, 0x33111122);
2252	&_data_word(0xbb6969d2, 0x70d9d9a9, 0x898e8e07, 0xa7949433);
2253	&_data_word(0xb69b9b2d, 0x221e1e3c, 0x92878715, 0x20e9e9c9);
2254	&_data_word(0x49cece87, 0xff5555aa, 0x78282850, 0x7adfdfa5);
2255	&_data_word(0x8f8c8c03, 0xf8a1a159, 0x80898909, 0x170d0d1a);
2256	&_data_word(0xdabfbf65, 0x31e6e6d7, 0xc6424284, 0xb86868d0);
2257	&_data_word(0xc3414182, 0xb0999929, 0x772d2d5a, 0x110f0f1e);
2258	&_data_word(0xcbb0b07b, 0xfc5454a8, 0xd6bbbb6d, 0x3a16162c);
2259
2260#Te4	# four copies of Te4 to choose from to avoid L1 aliasing
2261	&data_byte(0x63, 0x7c, 0x77, 0x7b, 0xf2, 0x6b, 0x6f, 0xc5);
2262	&data_byte(0x30, 0x01, 0x67, 0x2b, 0xfe, 0xd7, 0xab, 0x76);
2263	&data_byte(0xca, 0x82, 0xc9, 0x7d, 0xfa, 0x59, 0x47, 0xf0);
2264	&data_byte(0xad, 0xd4, 0xa2, 0xaf, 0x9c, 0xa4, 0x72, 0xc0);
2265	&data_byte(0xb7, 0xfd, 0x93, 0x26, 0x36, 0x3f, 0xf7, 0xcc);
2266	&data_byte(0x34, 0xa5, 0xe5, 0xf1, 0x71, 0xd8, 0x31, 0x15);
2267	&data_byte(0x04, 0xc7, 0x23, 0xc3, 0x18, 0x96, 0x05, 0x9a);
2268	&data_byte(0x07, 0x12, 0x80, 0xe2, 0xeb, 0x27, 0xb2, 0x75);
2269	&data_byte(0x09, 0x83, 0x2c, 0x1a, 0x1b, 0x6e, 0x5a, 0xa0);
2270	&data_byte(0x52, 0x3b, 0xd6, 0xb3, 0x29, 0xe3, 0x2f, 0x84);
2271	&data_byte(0x53, 0xd1, 0x00, 0xed, 0x20, 0xfc, 0xb1, 0x5b);
2272	&data_byte(0x6a, 0xcb, 0xbe, 0x39, 0x4a, 0x4c, 0x58, 0xcf);
2273	&data_byte(0xd0, 0xef, 0xaa, 0xfb, 0x43, 0x4d, 0x33, 0x85);
2274	&data_byte(0x45, 0xf9, 0x02, 0x7f, 0x50, 0x3c, 0x9f, 0xa8);
2275	&data_byte(0x51, 0xa3, 0x40, 0x8f, 0x92, 0x9d, 0x38, 0xf5);
2276	&data_byte(0xbc, 0xb6, 0xda, 0x21, 0x10, 0xff, 0xf3, 0xd2);
2277	&data_byte(0xcd, 0x0c, 0x13, 0xec, 0x5f, 0x97, 0x44, 0x17);
2278	&data_byte(0xc4, 0xa7, 0x7e, 0x3d, 0x64, 0x5d, 0x19, 0x73);
2279	&data_byte(0x60, 0x81, 0x4f, 0xdc, 0x22, 0x2a, 0x90, 0x88);
2280	&data_byte(0x46, 0xee, 0xb8, 0x14, 0xde, 0x5e, 0x0b, 0xdb);
2281	&data_byte(0xe0, 0x32, 0x3a, 0x0a, 0x49, 0x06, 0x24, 0x5c);
2282	&data_byte(0xc2, 0xd3, 0xac, 0x62, 0x91, 0x95, 0xe4, 0x79);
2283	&data_byte(0xe7, 0xc8, 0x37, 0x6d, 0x8d, 0xd5, 0x4e, 0xa9);
2284	&data_byte(0x6c, 0x56, 0xf4, 0xea, 0x65, 0x7a, 0xae, 0x08);
2285	&data_byte(0xba, 0x78, 0x25, 0x2e, 0x1c, 0xa6, 0xb4, 0xc6);
2286	&data_byte(0xe8, 0xdd, 0x74, 0x1f, 0x4b, 0xbd, 0x8b, 0x8a);
2287	&data_byte(0x70, 0x3e, 0xb5, 0x66, 0x48, 0x03, 0xf6, 0x0e);
2288	&data_byte(0x61, 0x35, 0x57, 0xb9, 0x86, 0xc1, 0x1d, 0x9e);
2289	&data_byte(0xe1, 0xf8, 0x98, 0x11, 0x69, 0xd9, 0x8e, 0x94);
2290	&data_byte(0x9b, 0x1e, 0x87, 0xe9, 0xce, 0x55, 0x28, 0xdf);
2291	&data_byte(0x8c, 0xa1, 0x89, 0x0d, 0xbf, 0xe6, 0x42, 0x68);
2292	&data_byte(0x41, 0x99, 0x2d, 0x0f, 0xb0, 0x54, 0xbb, 0x16);
2293
2294	&data_byte(0x63, 0x7c, 0x77, 0x7b, 0xf2, 0x6b, 0x6f, 0xc5);
2295	&data_byte(0x30, 0x01, 0x67, 0x2b, 0xfe, 0xd7, 0xab, 0x76);
2296	&data_byte(0xca, 0x82, 0xc9, 0x7d, 0xfa, 0x59, 0x47, 0xf0);
2297	&data_byte(0xad, 0xd4, 0xa2, 0xaf, 0x9c, 0xa4, 0x72, 0xc0);
2298	&data_byte(0xb7, 0xfd, 0x93, 0x26, 0x36, 0x3f, 0xf7, 0xcc);
2299	&data_byte(0x34, 0xa5, 0xe5, 0xf1, 0x71, 0xd8, 0x31, 0x15);
2300	&data_byte(0x04, 0xc7, 0x23, 0xc3, 0x18, 0x96, 0x05, 0x9a);
2301	&data_byte(0x07, 0x12, 0x80, 0xe2, 0xeb, 0x27, 0xb2, 0x75);
2302	&data_byte(0x09, 0x83, 0x2c, 0x1a, 0x1b, 0x6e, 0x5a, 0xa0);
2303	&data_byte(0x52, 0x3b, 0xd6, 0xb3, 0x29, 0xe3, 0x2f, 0x84);
2304	&data_byte(0x53, 0xd1, 0x00, 0xed, 0x20, 0xfc, 0xb1, 0x5b);
2305	&data_byte(0x6a, 0xcb, 0xbe, 0x39, 0x4a, 0x4c, 0x58, 0xcf);
2306	&data_byte(0xd0, 0xef, 0xaa, 0xfb, 0x43, 0x4d, 0x33, 0x85);
2307	&data_byte(0x45, 0xf9, 0x02, 0x7f, 0x50, 0x3c, 0x9f, 0xa8);
2308	&data_byte(0x51, 0xa3, 0x40, 0x8f, 0x92, 0x9d, 0x38, 0xf5);
2309	&data_byte(0xbc, 0xb6, 0xda, 0x21, 0x10, 0xff, 0xf3, 0xd2);
2310	&data_byte(0xcd, 0x0c, 0x13, 0xec, 0x5f, 0x97, 0x44, 0x17);
2311	&data_byte(0xc4, 0xa7, 0x7e, 0x3d, 0x64, 0x5d, 0x19, 0x73);
2312	&data_byte(0x60, 0x81, 0x4f, 0xdc, 0x22, 0x2a, 0x90, 0x88);
2313	&data_byte(0x46, 0xee, 0xb8, 0x14, 0xde, 0x5e, 0x0b, 0xdb);
2314	&data_byte(0xe0, 0x32, 0x3a, 0x0a, 0x49, 0x06, 0x24, 0x5c);
2315	&data_byte(0xc2, 0xd3, 0xac, 0x62, 0x91, 0x95, 0xe4, 0x79);
2316	&data_byte(0xe7, 0xc8, 0x37, 0x6d, 0x8d, 0xd5, 0x4e, 0xa9);
2317	&data_byte(0x6c, 0x56, 0xf4, 0xea, 0x65, 0x7a, 0xae, 0x08);
2318	&data_byte(0xba, 0x78, 0x25, 0x2e, 0x1c, 0xa6, 0xb4, 0xc6);
2319	&data_byte(0xe8, 0xdd, 0x74, 0x1f, 0x4b, 0xbd, 0x8b, 0x8a);
2320	&data_byte(0x70, 0x3e, 0xb5, 0x66, 0x48, 0x03, 0xf6, 0x0e);
2321	&data_byte(0x61, 0x35, 0x57, 0xb9, 0x86, 0xc1, 0x1d, 0x9e);
2322	&data_byte(0xe1, 0xf8, 0x98, 0x11, 0x69, 0xd9, 0x8e, 0x94);
2323	&data_byte(0x9b, 0x1e, 0x87, 0xe9, 0xce, 0x55, 0x28, 0xdf);
2324	&data_byte(0x8c, 0xa1, 0x89, 0x0d, 0xbf, 0xe6, 0x42, 0x68);
2325	&data_byte(0x41, 0x99, 0x2d, 0x0f, 0xb0, 0x54, 0xbb, 0x16);
2326
2327	&data_byte(0x63, 0x7c, 0x77, 0x7b, 0xf2, 0x6b, 0x6f, 0xc5);
2328	&data_byte(0x30, 0x01, 0x67, 0x2b, 0xfe, 0xd7, 0xab, 0x76);
2329	&data_byte(0xca, 0x82, 0xc9, 0x7d, 0xfa, 0x59, 0x47, 0xf0);
2330	&data_byte(0xad, 0xd4, 0xa2, 0xaf, 0x9c, 0xa4, 0x72, 0xc0);
2331	&data_byte(0xb7, 0xfd, 0x93, 0x26, 0x36, 0x3f, 0xf7, 0xcc);
2332	&data_byte(0x34, 0xa5, 0xe5, 0xf1, 0x71, 0xd8, 0x31, 0x15);
2333	&data_byte(0x04, 0xc7, 0x23, 0xc3, 0x18, 0x96, 0x05, 0x9a);
2334	&data_byte(0x07, 0x12, 0x80, 0xe2, 0xeb, 0x27, 0xb2, 0x75);
2335	&data_byte(0x09, 0x83, 0x2c, 0x1a, 0x1b, 0x6e, 0x5a, 0xa0);
2336	&data_byte(0x52, 0x3b, 0xd6, 0xb3, 0x29, 0xe3, 0x2f, 0x84);
2337	&data_byte(0x53, 0xd1, 0x00, 0xed, 0x20, 0xfc, 0xb1, 0x5b);
2338	&data_byte(0x6a, 0xcb, 0xbe, 0x39, 0x4a, 0x4c, 0x58, 0xcf);
2339	&data_byte(0xd0, 0xef, 0xaa, 0xfb, 0x43, 0x4d, 0x33, 0x85);
2340	&data_byte(0x45, 0xf9, 0x02, 0x7f, 0x50, 0x3c, 0x9f, 0xa8);
2341	&data_byte(0x51, 0xa3, 0x40, 0x8f, 0x92, 0x9d, 0x38, 0xf5);
2342	&data_byte(0xbc, 0xb6, 0xda, 0x21, 0x10, 0xff, 0xf3, 0xd2);
2343	&data_byte(0xcd, 0x0c, 0x13, 0xec, 0x5f, 0x97, 0x44, 0x17);
2344	&data_byte(0xc4, 0xa7, 0x7e, 0x3d, 0x64, 0x5d, 0x19, 0x73);
2345	&data_byte(0x60, 0x81, 0x4f, 0xdc, 0x22, 0x2a, 0x90, 0x88);
2346	&data_byte(0x46, 0xee, 0xb8, 0x14, 0xde, 0x5e, 0x0b, 0xdb);
2347	&data_byte(0xe0, 0x32, 0x3a, 0x0a, 0x49, 0x06, 0x24, 0x5c);
2348	&data_byte(0xc2, 0xd3, 0xac, 0x62, 0x91, 0x95, 0xe4, 0x79);
2349	&data_byte(0xe7, 0xc8, 0x37, 0x6d, 0x8d, 0xd5, 0x4e, 0xa9);
2350	&data_byte(0x6c, 0x56, 0xf4, 0xea, 0x65, 0x7a, 0xae, 0x08);
2351	&data_byte(0xba, 0x78, 0x25, 0x2e, 0x1c, 0xa6, 0xb4, 0xc6);
2352	&data_byte(0xe8, 0xdd, 0x74, 0x1f, 0x4b, 0xbd, 0x8b, 0x8a);
2353	&data_byte(0x70, 0x3e, 0xb5, 0x66, 0x48, 0x03, 0xf6, 0x0e);
2354	&data_byte(0x61, 0x35, 0x57, 0xb9, 0x86, 0xc1, 0x1d, 0x9e);
2355	&data_byte(0xe1, 0xf8, 0x98, 0x11, 0x69, 0xd9, 0x8e, 0x94);
2356	&data_byte(0x9b, 0x1e, 0x87, 0xe9, 0xce, 0x55, 0x28, 0xdf);
2357	&data_byte(0x8c, 0xa1, 0x89, 0x0d, 0xbf, 0xe6, 0x42, 0x68);
2358	&data_byte(0x41, 0x99, 0x2d, 0x0f, 0xb0, 0x54, 0xbb, 0x16);
2359
2360	&data_byte(0x63, 0x7c, 0x77, 0x7b, 0xf2, 0x6b, 0x6f, 0xc5);
2361	&data_byte(0x30, 0x01, 0x67, 0x2b, 0xfe, 0xd7, 0xab, 0x76);
2362	&data_byte(0xca, 0x82, 0xc9, 0x7d, 0xfa, 0x59, 0x47, 0xf0);
2363	&data_byte(0xad, 0xd4, 0xa2, 0xaf, 0x9c, 0xa4, 0x72, 0xc0);
2364	&data_byte(0xb7, 0xfd, 0x93, 0x26, 0x36, 0x3f, 0xf7, 0xcc);
2365	&data_byte(0x34, 0xa5, 0xe5, 0xf1, 0x71, 0xd8, 0x31, 0x15);
2366	&data_byte(0x04, 0xc7, 0x23, 0xc3, 0x18, 0x96, 0x05, 0x9a);
2367	&data_byte(0x07, 0x12, 0x80, 0xe2, 0xeb, 0x27, 0xb2, 0x75);
2368	&data_byte(0x09, 0x83, 0x2c, 0x1a, 0x1b, 0x6e, 0x5a, 0xa0);
2369	&data_byte(0x52, 0x3b, 0xd6, 0xb3, 0x29, 0xe3, 0x2f, 0x84);
2370	&data_byte(0x53, 0xd1, 0x00, 0xed, 0x20, 0xfc, 0xb1, 0x5b);
2371	&data_byte(0x6a, 0xcb, 0xbe, 0x39, 0x4a, 0x4c, 0x58, 0xcf);
2372	&data_byte(0xd0, 0xef, 0xaa, 0xfb, 0x43, 0x4d, 0x33, 0x85);
2373	&data_byte(0x45, 0xf9, 0x02, 0x7f, 0x50, 0x3c, 0x9f, 0xa8);
2374	&data_byte(0x51, 0xa3, 0x40, 0x8f, 0x92, 0x9d, 0x38, 0xf5);
2375	&data_byte(0xbc, 0xb6, 0xda, 0x21, 0x10, 0xff, 0xf3, 0xd2);
2376	&data_byte(0xcd, 0x0c, 0x13, 0xec, 0x5f, 0x97, 0x44, 0x17);
2377	&data_byte(0xc4, 0xa7, 0x7e, 0x3d, 0x64, 0x5d, 0x19, 0x73);
2378	&data_byte(0x60, 0x81, 0x4f, 0xdc, 0x22, 0x2a, 0x90, 0x88);
2379	&data_byte(0x46, 0xee, 0xb8, 0x14, 0xde, 0x5e, 0x0b, 0xdb);
2380	&data_byte(0xe0, 0x32, 0x3a, 0x0a, 0x49, 0x06, 0x24, 0x5c);
2381	&data_byte(0xc2, 0xd3, 0xac, 0x62, 0x91, 0x95, 0xe4, 0x79);
2382	&data_byte(0xe7, 0xc8, 0x37, 0x6d, 0x8d, 0xd5, 0x4e, 0xa9);
2383	&data_byte(0x6c, 0x56, 0xf4, 0xea, 0x65, 0x7a, 0xae, 0x08);
2384	&data_byte(0xba, 0x78, 0x25, 0x2e, 0x1c, 0xa6, 0xb4, 0xc6);
2385	&data_byte(0xe8, 0xdd, 0x74, 0x1f, 0x4b, 0xbd, 0x8b, 0x8a);
2386	&data_byte(0x70, 0x3e, 0xb5, 0x66, 0x48, 0x03, 0xf6, 0x0e);
2387	&data_byte(0x61, 0x35, 0x57, 0xb9, 0x86, 0xc1, 0x1d, 0x9e);
2388	&data_byte(0xe1, 0xf8, 0x98, 0x11, 0x69, 0xd9, 0x8e, 0x94);
2389	&data_byte(0x9b, 0x1e, 0x87, 0xe9, 0xce, 0x55, 0x28, 0xdf);
2390	&data_byte(0x8c, 0xa1, 0x89, 0x0d, 0xbf, 0xe6, 0x42, 0x68);
2391	&data_byte(0x41, 0x99, 0x2d, 0x0f, 0xb0, 0x54, 0xbb, 0x16);
2392#rcon:
2393$code.=<<___;
2394	.long	0x00000001, 0x00000002, 0x00000004, 0x00000008
2395	.long	0x00000010, 0x00000020, 0x00000040, 0x00000080
2396	.long	0x0000001b, 0x00000036, 0x80808080, 0x80808080
2397	.long	0xfefefefe, 0xfefefefe, 0x1b1b1b1b, 0x1b1b1b1b
2398___
2399$code.=<<___;
2400.align	64
2401.LAES_Td:
2402___
2403	&_data_word(0x50a7f451, 0x5365417e, 0xc3a4171a, 0x965e273a);
2404	&_data_word(0xcb6bab3b, 0xf1459d1f, 0xab58faac, 0x9303e34b);
2405	&_data_word(0x55fa3020, 0xf66d76ad, 0x9176cc88, 0x254c02f5);
2406	&_data_word(0xfcd7e54f, 0xd7cb2ac5, 0x80443526, 0x8fa362b5);
2407	&_data_word(0x495ab1de, 0x671bba25, 0x980eea45, 0xe1c0fe5d);
2408	&_data_word(0x02752fc3, 0x12f04c81, 0xa397468d, 0xc6f9d36b);
2409	&_data_word(0xe75f8f03, 0x959c9215, 0xeb7a6dbf, 0xda595295);
2410	&_data_word(0x2d83bed4, 0xd3217458, 0x2969e049, 0x44c8c98e);
2411	&_data_word(0x6a89c275, 0x78798ef4, 0x6b3e5899, 0xdd71b927);
2412	&_data_word(0xb64fe1be, 0x17ad88f0, 0x66ac20c9, 0xb43ace7d);
2413	&_data_word(0x184adf63, 0x82311ae5, 0x60335197, 0x457f5362);
2414	&_data_word(0xe07764b1, 0x84ae6bbb, 0x1ca081fe, 0x942b08f9);
2415	&_data_word(0x58684870, 0x19fd458f, 0x876cde94, 0xb7f87b52);
2416	&_data_word(0x23d373ab, 0xe2024b72, 0x578f1fe3, 0x2aab5566);
2417	&_data_word(0x0728ebb2, 0x03c2b52f, 0x9a7bc586, 0xa50837d3);
2418	&_data_word(0xf2872830, 0xb2a5bf23, 0xba6a0302, 0x5c8216ed);
2419	&_data_word(0x2b1ccf8a, 0x92b479a7, 0xf0f207f3, 0xa1e2694e);
2420	&_data_word(0xcdf4da65, 0xd5be0506, 0x1f6234d1, 0x8afea6c4);
2421	&_data_word(0x9d532e34, 0xa055f3a2, 0x32e18a05, 0x75ebf6a4);
2422	&_data_word(0x39ec830b, 0xaaef6040, 0x069f715e, 0x51106ebd);
2423	&_data_word(0xf98a213e, 0x3d06dd96, 0xae053edd, 0x46bde64d);
2424	&_data_word(0xb58d5491, 0x055dc471, 0x6fd40604, 0xff155060);
2425	&_data_word(0x24fb9819, 0x97e9bdd6, 0xcc434089, 0x779ed967);
2426	&_data_word(0xbd42e8b0, 0x888b8907, 0x385b19e7, 0xdbeec879);
2427	&_data_word(0x470a7ca1, 0xe90f427c, 0xc91e84f8, 0x00000000);
2428	&_data_word(0x83868009, 0x48ed2b32, 0xac70111e, 0x4e725a6c);
2429	&_data_word(0xfbff0efd, 0x5638850f, 0x1ed5ae3d, 0x27392d36);
2430	&_data_word(0x64d90f0a, 0x21a65c68, 0xd1545b9b, 0x3a2e3624);
2431	&_data_word(0xb1670a0c, 0x0fe75793, 0xd296eeb4, 0x9e919b1b);
2432	&_data_word(0x4fc5c080, 0xa220dc61, 0x694b775a, 0x161a121c);
2433	&_data_word(0x0aba93e2, 0xe52aa0c0, 0x43e0223c, 0x1d171b12);
2434	&_data_word(0x0b0d090e, 0xadc78bf2, 0xb9a8b62d, 0xc8a91e14);
2435	&_data_word(0x8519f157, 0x4c0775af, 0xbbdd99ee, 0xfd607fa3);
2436	&_data_word(0x9f2601f7, 0xbcf5725c, 0xc53b6644, 0x347efb5b);
2437	&_data_word(0x7629438b, 0xdcc623cb, 0x68fcedb6, 0x63f1e4b8);
2438	&_data_word(0xcadc31d7, 0x10856342, 0x40229713, 0x2011c684);
2439	&_data_word(0x7d244a85, 0xf83dbbd2, 0x1132f9ae, 0x6da129c7);
2440	&_data_word(0x4b2f9e1d, 0xf330b2dc, 0xec52860d, 0xd0e3c177);
2441	&_data_word(0x6c16b32b, 0x99b970a9, 0xfa489411, 0x2264e947);
2442	&_data_word(0xc48cfca8, 0x1a3ff0a0, 0xd82c7d56, 0xef903322);
2443	&_data_word(0xc74e4987, 0xc1d138d9, 0xfea2ca8c, 0x360bd498);
2444	&_data_word(0xcf81f5a6, 0x28de7aa5, 0x268eb7da, 0xa4bfad3f);
2445	&_data_word(0xe49d3a2c, 0x0d927850, 0x9bcc5f6a, 0x62467e54);
2446	&_data_word(0xc2138df6, 0xe8b8d890, 0x5ef7392e, 0xf5afc382);
2447	&_data_word(0xbe805d9f, 0x7c93d069, 0xa92dd56f, 0xb31225cf);
2448	&_data_word(0x3b99acc8, 0xa77d1810, 0x6e639ce8, 0x7bbb3bdb);
2449	&_data_word(0x097826cd, 0xf418596e, 0x01b79aec, 0xa89a4f83);
2450	&_data_word(0x656e95e6, 0x7ee6ffaa, 0x08cfbc21, 0xe6e815ef);
2451	&_data_word(0xd99be7ba, 0xce366f4a, 0xd4099fea, 0xd67cb029);
2452	&_data_word(0xafb2a431, 0x31233f2a, 0x3094a5c6, 0xc066a235);
2453	&_data_word(0x37bc4e74, 0xa6ca82fc, 0xb0d090e0, 0x15d8a733);
2454	&_data_word(0x4a9804f1, 0xf7daec41, 0x0e50cd7f, 0x2ff69117);
2455	&_data_word(0x8dd64d76, 0x4db0ef43, 0x544daacc, 0xdf0496e4);
2456	&_data_word(0xe3b5d19e, 0x1b886a4c, 0xb81f2cc1, 0x7f516546);
2457	&_data_word(0x04ea5e9d, 0x5d358c01, 0x737487fa, 0x2e410bfb);
2458	&_data_word(0x5a1d67b3, 0x52d2db92, 0x335610e9, 0x1347d66d);
2459	&_data_word(0x8c61d79a, 0x7a0ca137, 0x8e14f859, 0x893c13eb);
2460	&_data_word(0xee27a9ce, 0x35c961b7, 0xede51ce1, 0x3cb1477a);
2461	&_data_word(0x59dfd29c, 0x3f73f255, 0x79ce1418, 0xbf37c773);
2462	&_data_word(0xeacdf753, 0x5baafd5f, 0x146f3ddf, 0x86db4478);
2463	&_data_word(0x81f3afca, 0x3ec468b9, 0x2c342438, 0x5f40a3c2);
2464	&_data_word(0x72c31d16, 0x0c25e2bc, 0x8b493c28, 0x41950dff);
2465	&_data_word(0x7101a839, 0xdeb30c08, 0x9ce4b4d8, 0x90c15664);
2466	&_data_word(0x6184cb7b, 0x70b632d5, 0x745c6c48, 0x4257b8d0);
2467
2468#Td4:	# four copies of Td4 to choose from to avoid L1 aliasing
2469	&data_byte(0x52, 0x09, 0x6a, 0xd5, 0x30, 0x36, 0xa5, 0x38);
2470	&data_byte(0xbf, 0x40, 0xa3, 0x9e, 0x81, 0xf3, 0xd7, 0xfb);
2471	&data_byte(0x7c, 0xe3, 0x39, 0x82, 0x9b, 0x2f, 0xff, 0x87);
2472	&data_byte(0x34, 0x8e, 0x43, 0x44, 0xc4, 0xde, 0xe9, 0xcb);
2473	&data_byte(0x54, 0x7b, 0x94, 0x32, 0xa6, 0xc2, 0x23, 0x3d);
2474	&data_byte(0xee, 0x4c, 0x95, 0x0b, 0x42, 0xfa, 0xc3, 0x4e);
2475	&data_byte(0x08, 0x2e, 0xa1, 0x66, 0x28, 0xd9, 0x24, 0xb2);
2476	&data_byte(0x76, 0x5b, 0xa2, 0x49, 0x6d, 0x8b, 0xd1, 0x25);
2477	&data_byte(0x72, 0xf8, 0xf6, 0x64, 0x86, 0x68, 0x98, 0x16);
2478	&data_byte(0xd4, 0xa4, 0x5c, 0xcc, 0x5d, 0x65, 0xb6, 0x92);
2479	&data_byte(0x6c, 0x70, 0x48, 0x50, 0xfd, 0xed, 0xb9, 0xda);
2480	&data_byte(0x5e, 0x15, 0x46, 0x57, 0xa7, 0x8d, 0x9d, 0x84);
2481	&data_byte(0x90, 0xd8, 0xab, 0x00, 0x8c, 0xbc, 0xd3, 0x0a);
2482	&data_byte(0xf7, 0xe4, 0x58, 0x05, 0xb8, 0xb3, 0x45, 0x06);
2483	&data_byte(0xd0, 0x2c, 0x1e, 0x8f, 0xca, 0x3f, 0x0f, 0x02);
2484	&data_byte(0xc1, 0xaf, 0xbd, 0x03, 0x01, 0x13, 0x8a, 0x6b);
2485	&data_byte(0x3a, 0x91, 0x11, 0x41, 0x4f, 0x67, 0xdc, 0xea);
2486	&data_byte(0x97, 0xf2, 0xcf, 0xce, 0xf0, 0xb4, 0xe6, 0x73);
2487	&data_byte(0x96, 0xac, 0x74, 0x22, 0xe7, 0xad, 0x35, 0x85);
2488	&data_byte(0xe2, 0xf9, 0x37, 0xe8, 0x1c, 0x75, 0xdf, 0x6e);
2489	&data_byte(0x47, 0xf1, 0x1a, 0x71, 0x1d, 0x29, 0xc5, 0x89);
2490	&data_byte(0x6f, 0xb7, 0x62, 0x0e, 0xaa, 0x18, 0xbe, 0x1b);
2491	&data_byte(0xfc, 0x56, 0x3e, 0x4b, 0xc6, 0xd2, 0x79, 0x20);
2492	&data_byte(0x9a, 0xdb, 0xc0, 0xfe, 0x78, 0xcd, 0x5a, 0xf4);
2493	&data_byte(0x1f, 0xdd, 0xa8, 0x33, 0x88, 0x07, 0xc7, 0x31);
2494	&data_byte(0xb1, 0x12, 0x10, 0x59, 0x27, 0x80, 0xec, 0x5f);
2495	&data_byte(0x60, 0x51, 0x7f, 0xa9, 0x19, 0xb5, 0x4a, 0x0d);
2496	&data_byte(0x2d, 0xe5, 0x7a, 0x9f, 0x93, 0xc9, 0x9c, 0xef);
2497	&data_byte(0xa0, 0xe0, 0x3b, 0x4d, 0xae, 0x2a, 0xf5, 0xb0);
2498	&data_byte(0xc8, 0xeb, 0xbb, 0x3c, 0x83, 0x53, 0x99, 0x61);
2499	&data_byte(0x17, 0x2b, 0x04, 0x7e, 0xba, 0x77, 0xd6, 0x26);
2500	&data_byte(0xe1, 0x69, 0x14, 0x63, 0x55, 0x21, 0x0c, 0x7d);
2501$code.=<<___;
2502	.long	0x80808080, 0x80808080, 0xfefefefe, 0xfefefefe
2503	.long	0x1b1b1b1b, 0x1b1b1b1b, 0, 0
2504___
2505	&data_byte(0x52, 0x09, 0x6a, 0xd5, 0x30, 0x36, 0xa5, 0x38);
2506	&data_byte(0xbf, 0x40, 0xa3, 0x9e, 0x81, 0xf3, 0xd7, 0xfb);
2507	&data_byte(0x7c, 0xe3, 0x39, 0x82, 0x9b, 0x2f, 0xff, 0x87);
2508	&data_byte(0x34, 0x8e, 0x43, 0x44, 0xc4, 0xde, 0xe9, 0xcb);
2509	&data_byte(0x54, 0x7b, 0x94, 0x32, 0xa6, 0xc2, 0x23, 0x3d);
2510	&data_byte(0xee, 0x4c, 0x95, 0x0b, 0x42, 0xfa, 0xc3, 0x4e);
2511	&data_byte(0x08, 0x2e, 0xa1, 0x66, 0x28, 0xd9, 0x24, 0xb2);
2512	&data_byte(0x76, 0x5b, 0xa2, 0x49, 0x6d, 0x8b, 0xd1, 0x25);
2513	&data_byte(0x72, 0xf8, 0xf6, 0x64, 0x86, 0x68, 0x98, 0x16);
2514	&data_byte(0xd4, 0xa4, 0x5c, 0xcc, 0x5d, 0x65, 0xb6, 0x92);
2515	&data_byte(0x6c, 0x70, 0x48, 0x50, 0xfd, 0xed, 0xb9, 0xda);
2516	&data_byte(0x5e, 0x15, 0x46, 0x57, 0xa7, 0x8d, 0x9d, 0x84);
2517	&data_byte(0x90, 0xd8, 0xab, 0x00, 0x8c, 0xbc, 0xd3, 0x0a);
2518	&data_byte(0xf7, 0xe4, 0x58, 0x05, 0xb8, 0xb3, 0x45, 0x06);
2519	&data_byte(0xd0, 0x2c, 0x1e, 0x8f, 0xca, 0x3f, 0x0f, 0x02);
2520	&data_byte(0xc1, 0xaf, 0xbd, 0x03, 0x01, 0x13, 0x8a, 0x6b);
2521	&data_byte(0x3a, 0x91, 0x11, 0x41, 0x4f, 0x67, 0xdc, 0xea);
2522	&data_byte(0x97, 0xf2, 0xcf, 0xce, 0xf0, 0xb4, 0xe6, 0x73);
2523	&data_byte(0x96, 0xac, 0x74, 0x22, 0xe7, 0xad, 0x35, 0x85);
2524	&data_byte(0xe2, 0xf9, 0x37, 0xe8, 0x1c, 0x75, 0xdf, 0x6e);
2525	&data_byte(0x47, 0xf1, 0x1a, 0x71, 0x1d, 0x29, 0xc5, 0x89);
2526	&data_byte(0x6f, 0xb7, 0x62, 0x0e, 0xaa, 0x18, 0xbe, 0x1b);
2527	&data_byte(0xfc, 0x56, 0x3e, 0x4b, 0xc6, 0xd2, 0x79, 0x20);
2528	&data_byte(0x9a, 0xdb, 0xc0, 0xfe, 0x78, 0xcd, 0x5a, 0xf4);
2529	&data_byte(0x1f, 0xdd, 0xa8, 0x33, 0x88, 0x07, 0xc7, 0x31);
2530	&data_byte(0xb1, 0x12, 0x10, 0x59, 0x27, 0x80, 0xec, 0x5f);
2531	&data_byte(0x60, 0x51, 0x7f, 0xa9, 0x19, 0xb5, 0x4a, 0x0d);
2532	&data_byte(0x2d, 0xe5, 0x7a, 0x9f, 0x93, 0xc9, 0x9c, 0xef);
2533	&data_byte(0xa0, 0xe0, 0x3b, 0x4d, 0xae, 0x2a, 0xf5, 0xb0);
2534	&data_byte(0xc8, 0xeb, 0xbb, 0x3c, 0x83, 0x53, 0x99, 0x61);
2535	&data_byte(0x17, 0x2b, 0x04, 0x7e, 0xba, 0x77, 0xd6, 0x26);
2536	&data_byte(0xe1, 0x69, 0x14, 0x63, 0x55, 0x21, 0x0c, 0x7d);
2537$code.=<<___;
2538	.long	0x80808080, 0x80808080, 0xfefefefe, 0xfefefefe
2539	.long	0x1b1b1b1b, 0x1b1b1b1b, 0, 0
2540___
2541	&data_byte(0x52, 0x09, 0x6a, 0xd5, 0x30, 0x36, 0xa5, 0x38);
2542	&data_byte(0xbf, 0x40, 0xa3, 0x9e, 0x81, 0xf3, 0xd7, 0xfb);
2543	&data_byte(0x7c, 0xe3, 0x39, 0x82, 0x9b, 0x2f, 0xff, 0x87);
2544	&data_byte(0x34, 0x8e, 0x43, 0x44, 0xc4, 0xde, 0xe9, 0xcb);
2545	&data_byte(0x54, 0x7b, 0x94, 0x32, 0xa6, 0xc2, 0x23, 0x3d);
2546	&data_byte(0xee, 0x4c, 0x95, 0x0b, 0x42, 0xfa, 0xc3, 0x4e);
2547	&data_byte(0x08, 0x2e, 0xa1, 0x66, 0x28, 0xd9, 0x24, 0xb2);
2548	&data_byte(0x76, 0x5b, 0xa2, 0x49, 0x6d, 0x8b, 0xd1, 0x25);
2549	&data_byte(0x72, 0xf8, 0xf6, 0x64, 0x86, 0x68, 0x98, 0x16);
2550	&data_byte(0xd4, 0xa4, 0x5c, 0xcc, 0x5d, 0x65, 0xb6, 0x92);
2551	&data_byte(0x6c, 0x70, 0x48, 0x50, 0xfd, 0xed, 0xb9, 0xda);
2552	&data_byte(0x5e, 0x15, 0x46, 0x57, 0xa7, 0x8d, 0x9d, 0x84);
2553	&data_byte(0x90, 0xd8, 0xab, 0x00, 0x8c, 0xbc, 0xd3, 0x0a);
2554	&data_byte(0xf7, 0xe4, 0x58, 0x05, 0xb8, 0xb3, 0x45, 0x06);
2555	&data_byte(0xd0, 0x2c, 0x1e, 0x8f, 0xca, 0x3f, 0x0f, 0x02);
2556	&data_byte(0xc1, 0xaf, 0xbd, 0x03, 0x01, 0x13, 0x8a, 0x6b);
2557	&data_byte(0x3a, 0x91, 0x11, 0x41, 0x4f, 0x67, 0xdc, 0xea);
2558	&data_byte(0x97, 0xf2, 0xcf, 0xce, 0xf0, 0xb4, 0xe6, 0x73);
2559	&data_byte(0x96, 0xac, 0x74, 0x22, 0xe7, 0xad, 0x35, 0x85);
2560	&data_byte(0xe2, 0xf9, 0x37, 0xe8, 0x1c, 0x75, 0xdf, 0x6e);
2561	&data_byte(0x47, 0xf1, 0x1a, 0x71, 0x1d, 0x29, 0xc5, 0x89);
2562	&data_byte(0x6f, 0xb7, 0x62, 0x0e, 0xaa, 0x18, 0xbe, 0x1b);
2563	&data_byte(0xfc, 0x56, 0x3e, 0x4b, 0xc6, 0xd2, 0x79, 0x20);
2564	&data_byte(0x9a, 0xdb, 0xc0, 0xfe, 0x78, 0xcd, 0x5a, 0xf4);
2565	&data_byte(0x1f, 0xdd, 0xa8, 0x33, 0x88, 0x07, 0xc7, 0x31);
2566	&data_byte(0xb1, 0x12, 0x10, 0x59, 0x27, 0x80, 0xec, 0x5f);
2567	&data_byte(0x60, 0x51, 0x7f, 0xa9, 0x19, 0xb5, 0x4a, 0x0d);
2568	&data_byte(0x2d, 0xe5, 0x7a, 0x9f, 0x93, 0xc9, 0x9c, 0xef);
2569	&data_byte(0xa0, 0xe0, 0x3b, 0x4d, 0xae, 0x2a, 0xf5, 0xb0);
2570	&data_byte(0xc8, 0xeb, 0xbb, 0x3c, 0x83, 0x53, 0x99, 0x61);
2571	&data_byte(0x17, 0x2b, 0x04, 0x7e, 0xba, 0x77, 0xd6, 0x26);
2572	&data_byte(0xe1, 0x69, 0x14, 0x63, 0x55, 0x21, 0x0c, 0x7d);
2573$code.=<<___;
2574	.long	0x80808080, 0x80808080, 0xfefefefe, 0xfefefefe
2575	.long	0x1b1b1b1b, 0x1b1b1b1b, 0, 0
2576___
2577	&data_byte(0x52, 0x09, 0x6a, 0xd5, 0x30, 0x36, 0xa5, 0x38);
2578	&data_byte(0xbf, 0x40, 0xa3, 0x9e, 0x81, 0xf3, 0xd7, 0xfb);
2579	&data_byte(0x7c, 0xe3, 0x39, 0x82, 0x9b, 0x2f, 0xff, 0x87);
2580	&data_byte(0x34, 0x8e, 0x43, 0x44, 0xc4, 0xde, 0xe9, 0xcb);
2581	&data_byte(0x54, 0x7b, 0x94, 0x32, 0xa6, 0xc2, 0x23, 0x3d);
2582	&data_byte(0xee, 0x4c, 0x95, 0x0b, 0x42, 0xfa, 0xc3, 0x4e);
2583	&data_byte(0x08, 0x2e, 0xa1, 0x66, 0x28, 0xd9, 0x24, 0xb2);
2584	&data_byte(0x76, 0x5b, 0xa2, 0x49, 0x6d, 0x8b, 0xd1, 0x25);
2585	&data_byte(0x72, 0xf8, 0xf6, 0x64, 0x86, 0x68, 0x98, 0x16);
2586	&data_byte(0xd4, 0xa4, 0x5c, 0xcc, 0x5d, 0x65, 0xb6, 0x92);
2587	&data_byte(0x6c, 0x70, 0x48, 0x50, 0xfd, 0xed, 0xb9, 0xda);
2588	&data_byte(0x5e, 0x15, 0x46, 0x57, 0xa7, 0x8d, 0x9d, 0x84);
2589	&data_byte(0x90, 0xd8, 0xab, 0x00, 0x8c, 0xbc, 0xd3, 0x0a);
2590	&data_byte(0xf7, 0xe4, 0x58, 0x05, 0xb8, 0xb3, 0x45, 0x06);
2591	&data_byte(0xd0, 0x2c, 0x1e, 0x8f, 0xca, 0x3f, 0x0f, 0x02);
2592	&data_byte(0xc1, 0xaf, 0xbd, 0x03, 0x01, 0x13, 0x8a, 0x6b);
2593	&data_byte(0x3a, 0x91, 0x11, 0x41, 0x4f, 0x67, 0xdc, 0xea);
2594	&data_byte(0x97, 0xf2, 0xcf, 0xce, 0xf0, 0xb4, 0xe6, 0x73);
2595	&data_byte(0x96, 0xac, 0x74, 0x22, 0xe7, 0xad, 0x35, 0x85);
2596	&data_byte(0xe2, 0xf9, 0x37, 0xe8, 0x1c, 0x75, 0xdf, 0x6e);
2597	&data_byte(0x47, 0xf1, 0x1a, 0x71, 0x1d, 0x29, 0xc5, 0x89);
2598	&data_byte(0x6f, 0xb7, 0x62, 0x0e, 0xaa, 0x18, 0xbe, 0x1b);
2599	&data_byte(0xfc, 0x56, 0x3e, 0x4b, 0xc6, 0xd2, 0x79, 0x20);
2600	&data_byte(0x9a, 0xdb, 0xc0, 0xfe, 0x78, 0xcd, 0x5a, 0xf4);
2601	&data_byte(0x1f, 0xdd, 0xa8, 0x33, 0x88, 0x07, 0xc7, 0x31);
2602	&data_byte(0xb1, 0x12, 0x10, 0x59, 0x27, 0x80, 0xec, 0x5f);
2603	&data_byte(0x60, 0x51, 0x7f, 0xa9, 0x19, 0xb5, 0x4a, 0x0d);
2604	&data_byte(0x2d, 0xe5, 0x7a, 0x9f, 0x93, 0xc9, 0x9c, 0xef);
2605	&data_byte(0xa0, 0xe0, 0x3b, 0x4d, 0xae, 0x2a, 0xf5, 0xb0);
2606	&data_byte(0xc8, 0xeb, 0xbb, 0x3c, 0x83, 0x53, 0x99, 0x61);
2607	&data_byte(0x17, 0x2b, 0x04, 0x7e, 0xba, 0x77, 0xd6, 0x26);
2608	&data_byte(0xe1, 0x69, 0x14, 0x63, 0x55, 0x21, 0x0c, 0x7d);
2609$code.=<<___;
2610	.long	0x80808080, 0x80808080, 0xfefefefe, 0xfefefefe
2611	.long	0x1b1b1b1b, 0x1b1b1b1b, 0, 0
2612.asciz  "AES for x86_64, CRYPTOGAMS by <appro\@openssl.org>"
2613.align	64
2614___
2615
2616# EXCEPTION_DISPOSITION handler (EXCEPTION_RECORD *rec,ULONG64 frame,
2617#		CONTEXT *context,DISPATCHER_CONTEXT *disp)
2618if ($win64) {
2619$rec="%rcx";
2620$frame="%rdx";
2621$context="%r8";
2622$disp="%r9";
2623
2624$code.=<<___;
2625.extern	__imp_RtlVirtualUnwind
2626.type	block_se_handler,\@abi-omnipotent
2627.align	16
2628block_se_handler:
2629	push	%rsi
2630	push	%rdi
2631	push	%rbx
2632	push	%rbp
2633	push	%r12
2634	push	%r13
2635	push	%r14
2636	push	%r15
2637	pushfq
2638	sub	\$64,%rsp
2639
2640	mov	120($context),%rax	# pull context->Rax
2641	mov	248($context),%rbx	# pull context->Rip
2642
2643	mov	8($disp),%rsi		# disp->ImageBase
2644	mov	56($disp),%r11		# disp->HandlerData
2645
2646	mov	0(%r11),%r10d		# HandlerData[0]
2647	lea	(%rsi,%r10),%r10	# prologue label
2648	cmp	%r10,%rbx		# context->Rip<prologue label
2649	jb	.Lin_block_prologue
2650
2651	mov	152($context),%rax	# pull context->Rsp
2652
2653	mov	4(%r11),%r10d		# HandlerData[1]
2654	lea	(%rsi,%r10),%r10	# epilogue label
2655	cmp	%r10,%rbx		# context->Rip>=epilogue label
2656	jae	.Lin_block_prologue
2657
2658	mov	24(%rax),%rax		# pull saved real stack pointer
2659
2660	mov	-8(%rax),%rbx
2661	mov	-16(%rax),%rbp
2662	mov	-24(%rax),%r12
2663	mov	-32(%rax),%r13
2664	mov	-40(%rax),%r14
2665	mov	-48(%rax),%r15
2666	mov	%rbx,144($context)	# restore context->Rbx
2667	mov	%rbp,160($context)	# restore context->Rbp
2668	mov	%r12,216($context)	# restore context->R12
2669	mov	%r13,224($context)	# restore context->R13
2670	mov	%r14,232($context)	# restore context->R14
2671	mov	%r15,240($context)	# restore context->R15
2672
2673.Lin_block_prologue:
2674	mov	8(%rax),%rdi
2675	mov	16(%rax),%rsi
2676	mov	%rax,152($context)	# restore context->Rsp
2677	mov	%rsi,168($context)	# restore context->Rsi
2678	mov	%rdi,176($context)	# restore context->Rdi
2679
2680	jmp	.Lcommon_seh_exit
2681.size	block_se_handler,.-block_se_handler
2682
2683.type	key_se_handler,\@abi-omnipotent
2684.align	16
2685key_se_handler:
2686	push	%rsi
2687	push	%rdi
2688	push	%rbx
2689	push	%rbp
2690	push	%r12
2691	push	%r13
2692	push	%r14
2693	push	%r15
2694	pushfq
2695	sub	\$64,%rsp
2696
2697	mov	120($context),%rax	# pull context->Rax
2698	mov	248($context),%rbx	# pull context->Rip
2699
2700	mov	8($disp),%rsi		# disp->ImageBase
2701	mov	56($disp),%r11		# disp->HandlerData
2702
2703	mov	0(%r11),%r10d		# HandlerData[0]
2704	lea	(%rsi,%r10),%r10	# prologue label
2705	cmp	%r10,%rbx		# context->Rip<prologue label
2706	jb	.Lin_key_prologue
2707
2708	mov	152($context),%rax	# pull context->Rsp
2709
2710	mov	4(%r11),%r10d		# HandlerData[1]
2711	lea	(%rsi,%r10),%r10	# epilogue label
2712	cmp	%r10,%rbx		# context->Rip>=epilogue label
2713	jae	.Lin_key_prologue
2714
2715	lea	56(%rax),%rax
2716
2717	mov	-8(%rax),%rbx
2718	mov	-16(%rax),%rbp
2719	mov	-24(%rax),%r12
2720	mov	-32(%rax),%r13
2721	mov	-40(%rax),%r14
2722	mov	-48(%rax),%r15
2723	mov	%rbx,144($context)	# restore context->Rbx
2724	mov	%rbp,160($context)	# restore context->Rbp
2725	mov	%r12,216($context)	# restore context->R12
2726	mov	%r13,224($context)	# restore context->R13
2727	mov	%r14,232($context)	# restore context->R14
2728	mov	%r15,240($context)	# restore context->R15
2729
2730.Lin_key_prologue:
2731	mov	8(%rax),%rdi
2732	mov	16(%rax),%rsi
2733	mov	%rax,152($context)	# restore context->Rsp
2734	mov	%rsi,168($context)	# restore context->Rsi
2735	mov	%rdi,176($context)	# restore context->Rdi
2736
2737	jmp	.Lcommon_seh_exit
2738.size	key_se_handler,.-key_se_handler
2739
2740.type	cbc_se_handler,\@abi-omnipotent
2741.align	16
2742cbc_se_handler:
2743	push	%rsi
2744	push	%rdi
2745	push	%rbx
2746	push	%rbp
2747	push	%r12
2748	push	%r13
2749	push	%r14
2750	push	%r15
2751	pushfq
2752	sub	\$64,%rsp
2753
2754	mov	120($context),%rax	# pull context->Rax
2755	mov	248($context),%rbx	# pull context->Rip
2756
2757	lea	.Lcbc_prologue(%rip),%r10
2758	cmp	%r10,%rbx		# context->Rip<.Lcbc_prologue
2759	jb	.Lin_cbc_prologue
2760
2761	lea	.Lcbc_fast_body(%rip),%r10
2762	cmp	%r10,%rbx		# context->Rip<.Lcbc_fast_body
2763	jb	.Lin_cbc_frame_setup
2764
2765	lea	.Lcbc_slow_prologue(%rip),%r10
2766	cmp	%r10,%rbx		# context->Rip<.Lcbc_slow_prologue
2767	jb	.Lin_cbc_body
2768
2769	lea	.Lcbc_slow_body(%rip),%r10
2770	cmp	%r10,%rbx		# context->Rip<.Lcbc_slow_body
2771	jb	.Lin_cbc_frame_setup
2772
2773.Lin_cbc_body:
2774	mov	152($context),%rax	# pull context->Rsp
2775
2776	lea	.Lcbc_epilogue(%rip),%r10
2777	cmp	%r10,%rbx		# context->Rip>=.Lcbc_epilogue
2778	jae	.Lin_cbc_prologue
2779
2780	lea	8(%rax),%rax
2781
2782	lea	.Lcbc_popfq(%rip),%r10
2783	cmp	%r10,%rbx		# context->Rip>=.Lcbc_popfq
2784	jae	.Lin_cbc_prologue
2785
2786	mov	`16-8`(%rax),%rax	# biased $_rsp
2787	lea	56(%rax),%rax
2788
2789.Lin_cbc_frame_setup:
2790	mov	-16(%rax),%rbx
2791	mov	-24(%rax),%rbp
2792	mov	-32(%rax),%r12
2793	mov	-40(%rax),%r13
2794	mov	-48(%rax),%r14
2795	mov	-56(%rax),%r15
2796	mov	%rbx,144($context)	# restore context->Rbx
2797	mov	%rbp,160($context)	# restore context->Rbp
2798	mov	%r12,216($context)	# restore context->R12
2799	mov	%r13,224($context)	# restore context->R13
2800	mov	%r14,232($context)	# restore context->R14
2801	mov	%r15,240($context)	# restore context->R15
2802
2803.Lin_cbc_prologue:
2804	mov	8(%rax),%rdi
2805	mov	16(%rax),%rsi
2806	mov	%rax,152($context)	# restore context->Rsp
2807	mov	%rsi,168($context)	# restore context->Rsi
2808	mov	%rdi,176($context)	# restore context->Rdi
2809
2810.Lcommon_seh_exit:
2811
2812	mov	40($disp),%rdi		# disp->ContextRecord
2813	mov	$context,%rsi		# context
2814	mov	\$`1232/8`,%ecx		# sizeof(CONTEXT)
2815	.long	0xa548f3fc		# cld; rep movsq
2816
2817	mov	$disp,%rsi
2818	xor	%rcx,%rcx		# arg1, UNW_FLAG_NHANDLER
2819	mov	8(%rsi),%rdx		# arg2, disp->ImageBase
2820	mov	0(%rsi),%r8		# arg3, disp->ControlPc
2821	mov	16(%rsi),%r9		# arg4, disp->FunctionEntry
2822	mov	40(%rsi),%r10		# disp->ContextRecord
2823	lea	56(%rsi),%r11		# &disp->HandlerData
2824	lea	24(%rsi),%r12		# &disp->EstablisherFrame
2825	mov	%r10,32(%rsp)		# arg5
2826	mov	%r11,40(%rsp)		# arg6
2827	mov	%r12,48(%rsp)		# arg7
2828	mov	%rcx,56(%rsp)		# arg8, (NULL)
2829	call	*__imp_RtlVirtualUnwind(%rip)
2830
2831	mov	\$1,%eax		# ExceptionContinueSearch
2832	add	\$64,%rsp
2833	popfq
2834	pop	%r15
2835	pop	%r14
2836	pop	%r13
2837	pop	%r12
2838	pop	%rbp
2839	pop	%rbx
2840	pop	%rdi
2841	pop	%rsi
2842	ret
2843.size	cbc_se_handler,.-cbc_se_handler
2844
2845.section	.pdata
2846.align	4
2847	.rva	.LSEH_begin_aes_nohw_encrypt
2848	.rva	.LSEH_end_aes_nohw_encrypt
2849	.rva	.LSEH_info_aes_nohw_encrypt
2850
2851	.rva	.LSEH_begin_aes_nohw_decrypt
2852	.rva	.LSEH_end_aes_nohw_decrypt
2853	.rva	.LSEH_info_aes_nohw_decrypt
2854
2855	.rva	.LSEH_begin_aes_nohw_set_encrypt_key
2856	.rva	.LSEH_end_aes_nohw_set_encrypt_key
2857	.rva	.LSEH_info_aes_nohw_set_encrypt_key
2858
2859	.rva	.LSEH_begin_aes_nohw_set_decrypt_key
2860	.rva	.LSEH_end_aes_nohw_set_decrypt_key
2861	.rva	.LSEH_info_aes_nohw_set_decrypt_key
2862
2863	.rva	.LSEH_begin_aes_nohw_cbc_encrypt
2864	.rva	.LSEH_end_aes_nohw_cbc_encrypt
2865	.rva	.LSEH_info_aes_nohw_cbc_encrypt
2866
2867.section	.xdata
2868.align	8
2869.LSEH_info_aes_nohw_encrypt:
2870	.byte	9,0,0,0
2871	.rva	block_se_handler
2872	.rva	.Lenc_prologue,.Lenc_epilogue	# HandlerData[]
2873.LSEH_info_aes_nohw_decrypt:
2874	.byte	9,0,0,0
2875	.rva	block_se_handler
2876	.rva	.Ldec_prologue,.Ldec_epilogue	# HandlerData[]
2877.LSEH_info_aes_nohw_set_encrypt_key:
2878	.byte	9,0,0,0
2879	.rva	key_se_handler
2880	.rva	.Lenc_key_prologue,.Lenc_key_epilogue	# HandlerData[]
2881.LSEH_info_aes_nohw_set_decrypt_key:
2882	.byte	9,0,0,0
2883	.rva	key_se_handler
2884	.rva	.Ldec_key_prologue,.Ldec_key_epilogue	# HandlerData[]
2885.LSEH_info_aes_nohw_cbc_encrypt:
2886	.byte	9,0,0,0
2887	.rva	cbc_se_handler
2888___
2889}
2890
2891$code =~ s/\`([^\`]*)\`/eval($1)/gem;
2892
2893print $code;
2894
2895close STDOUT;
2896