aes-x86_64.pl revision 392aa7cc7d2b122614c5393c3e357da07fd07af3 
1#!/usr/bin/env perl
2#
3# ====================================================================
4# Written by Andy Polyakov <appro@fy.chalmers.se> for the OpenSSL
5# project. The module is, however, dual licensed under OpenSSL and
6# CRYPTOGAMS licenses depending on where you obtain it. For further
7# details see http://www.openssl.org/~appro/cryptogams/.
8# ====================================================================
9#
10# Version 2.1.
11#
12# aes-*-cbc benchmarks are improved by >70% [compared to gcc 3.3.2 on
13# Opteron 240 CPU] plus all the bells-n-whistles from 32-bit version
14# [you'll notice a lot of resemblance], such as compressed S-boxes
15# in little-endian byte order, prefetch of these tables in CBC mode,
16# as well as avoiding L1 cache aliasing between stack frame and key
17# schedule and already mentioned tables, compressed Td4...
18#
19# Performance in number of cycles per processed byte for 128-bit key:
20#
21#		ECB encrypt	ECB decrypt	CBC large chunk
22# AMD64		33		41		13.0
23# EM64T		38		59		18.6(*)
24# Core 2	30		43		14.5(*)
25#
26# (*) with hyper-threading off
27
28$flavour = shift;
29$output  = shift;
30if ($flavour =~ /\./) { $output = $flavour; undef $flavour; }
31
32$win64=0; $win64=1 if ($flavour =~ /[nm]asm|mingw64/ || $output =~ /\.asm$/);
33
34$0 =~ m/(.*[\/\\])[^\/\\]+$/; $dir=$1;
35( $xlate="${dir}x86_64-xlate.pl" and -f $xlate ) or
36( $xlate="${dir}../../perlasm/x86_64-xlate.pl" and -f $xlate) or
37die "can't locate x86_64-xlate.pl";
38
39open STDOUT,"| $^X $xlate $flavour $output";
40
41$verticalspin=1;	# unlike 32-bit version $verticalspin performs
42			# ~15% better on both AMD and Intel cores
43$speed_limit=512;	# see aes-586.pl for details
44
45$code=".text\n";
46
47$s0="%eax";
48$s1="%ebx";
49$s2="%ecx";
50$s3="%edx";
51$acc0="%esi";	$mask80="%rsi";
52$acc1="%edi";	$maskfe="%rdi";
53$acc2="%ebp";	$mask1b="%rbp";
54$inp="%r8";
55$out="%r9";
56$t0="%r10d";
57$t1="%r11d";
58$t2="%r12d";
59$rnds="%r13d";
60$sbox="%r14";
61$key="%r15";
62
63sub hi() { my $r=shift;	$r =~ s/%[er]([a-d])x/%\1h/;	$r; }
64sub lo() { my $r=shift;	$r =~ s/%[er]([a-d])x/%\1l/;
65			$r =~ s/%[er]([sd]i)/%\1l/;
66			$r =~ s/%(r[0-9]+)[d]?/%\1b/;	$r; }
67sub LO() { my $r=shift; $r =~ s/%r([a-z]+)/%e\1/;
68			$r =~ s/%r([0-9]+)/%r\1d/;	$r; }
69sub _data_word()
70{ my $i;
71    while(defined($i=shift)) { $code.=sprintf".long\t0x%08x,0x%08x\n",$i,$i; }
72}
73sub data_word()
74{ my $i;
75  my $last=pop(@_);
76    $code.=".long\t";
77    while(defined($i=shift)) { $code.=sprintf"0x%08x,",$i; }
78    $code.=sprintf"0x%08x\n",$last;
79}
80
81sub data_byte()
82{ my $i;
83  my $last=pop(@_);
84    $code.=".byte\t";
85    while(defined($i=shift)) { $code.=sprintf"0x%02x,",$i&0xff; }
86    $code.=sprintf"0x%02x\n",$last&0xff;
87}
88
89sub encvert()
90{ my $t3="%r8d";	# zaps $inp!
91
92$code.=<<___;
93	# favor 3-way issue Opteron pipeline...
94	movzb	`&lo("$s0")`,$acc0
95	movzb	`&lo("$s1")`,$acc1
96	movzb	`&lo("$s2")`,$acc2
97	mov	0($sbox,$acc0,8),$t0
98	mov	0($sbox,$acc1,8),$t1
99	mov	0($sbox,$acc2,8),$t2
100
101	movzb	`&hi("$s1")`,$acc0
102	movzb	`&hi("$s2")`,$acc1
103	movzb	`&lo("$s3")`,$acc2
104	xor	3($sbox,$acc0,8),$t0
105	xor	3($sbox,$acc1,8),$t1
106	mov	0($sbox,$acc2,8),$t3
107
108	movzb	`&hi("$s3")`,$acc0
109	shr	\$16,$s2
110	movzb	`&hi("$s0")`,$acc2
111	xor	3($sbox,$acc0,8),$t2
112	shr	\$16,$s3
113	xor	3($sbox,$acc2,8),$t3
114
115	shr	\$16,$s1
116	lea	16($key),$key
117	shr	\$16,$s0
118
119	movzb	`&lo("$s2")`,$acc0
120	movzb	`&lo("$s3")`,$acc1
121	movzb	`&lo("$s0")`,$acc2
122	xor	2($sbox,$acc0,8),$t0
123	xor	2($sbox,$acc1,8),$t1
124	xor	2($sbox,$acc2,8),$t2
125
126	movzb	`&hi("$s3")`,$acc0
127	movzb	`&hi("$s0")`,$acc1
128	movzb	`&lo("$s1")`,$acc2
129	xor	1($sbox,$acc0,8),$t0
130	xor	1($sbox,$acc1,8),$t1
131	xor	2($sbox,$acc2,8),$t3
132
133	mov	12($key),$s3
134	movzb	`&hi("$s1")`,$acc1
135	movzb	`&hi("$s2")`,$acc2
136	mov	0($key),$s0
137	xor	1($sbox,$acc1,8),$t2
138	xor	1($sbox,$acc2,8),$t3
139
140	mov	4($key),$s1
141	mov	8($key),$s2
142	xor	$t0,$s0
143	xor	$t1,$s1
144	xor	$t2,$s2
145	xor	$t3,$s3
146___
147}
148
149sub enclastvert()
150{ my $t3="%r8d";	# zaps $inp!
151
152$code.=<<___;
153	movzb	`&lo("$s0")`,$acc0
154	movzb	`&lo("$s1")`,$acc1
155	movzb	`&lo("$s2")`,$acc2
156	movzb	2($sbox,$acc0,8),$t0
157	movzb	2($sbox,$acc1,8),$t1
158	movzb	2($sbox,$acc2,8),$t2
159
160	movzb	`&lo("$s3")`,$acc0
161	movzb	`&hi("$s1")`,$acc1
162	movzb	`&hi("$s2")`,$acc2
163	movzb	2($sbox,$acc0,8),$t3
164	mov	0($sbox,$acc1,8),$acc1	#$t0
165	mov	0($sbox,$acc2,8),$acc2	#$t1
166
167	and	\$0x0000ff00,$acc1
168	and	\$0x0000ff00,$acc2
169
170	xor	$acc1,$t0
171	xor	$acc2,$t1
172	shr	\$16,$s2
173
174	movzb	`&hi("$s3")`,$acc0
175	movzb	`&hi("$s0")`,$acc1
176	shr	\$16,$s3
177	mov	0($sbox,$acc0,8),$acc0	#$t2
178	mov	0($sbox,$acc1,8),$acc1	#$t3
179
180	and	\$0x0000ff00,$acc0
181	and	\$0x0000ff00,$acc1
182	shr	\$16,$s1
183	xor	$acc0,$t2
184	xor	$acc1,$t3
185	shr	\$16,$s0
186
187	movzb	`&lo("$s2")`,$acc0
188	movzb	`&lo("$s3")`,$acc1
189	movzb	`&lo("$s0")`,$acc2
190	mov	0($sbox,$acc0,8),$acc0	#$t0
191	mov	0($sbox,$acc1,8),$acc1	#$t1
192	mov	0($sbox,$acc2,8),$acc2	#$t2
193
194	and	\$0x00ff0000,$acc0
195	and	\$0x00ff0000,$acc1
196	and	\$0x00ff0000,$acc2
197
198	xor	$acc0,$t0
199	xor	$acc1,$t1
200	xor	$acc2,$t2
201
202	movzb	`&lo("$s1")`,$acc0
203	movzb	`&hi("$s3")`,$acc1
204	movzb	`&hi("$s0")`,$acc2
205	mov	0($sbox,$acc0,8),$acc0	#$t3
206	mov	2($sbox,$acc1,8),$acc1	#$t0
207	mov	2($sbox,$acc2,8),$acc2	#$t1
208
209	and	\$0x00ff0000,$acc0
210	and	\$0xff000000,$acc1
211	and	\$0xff000000,$acc2
212
213	xor	$acc0,$t3
214	xor	$acc1,$t0
215	xor	$acc2,$t1
216
217	movzb	`&hi("$s1")`,$acc0
218	movzb	`&hi("$s2")`,$acc1
219	mov	16+12($key),$s3
220	mov	2($sbox,$acc0,8),$acc0	#$t2
221	mov	2($sbox,$acc1,8),$acc1	#$t3
222	mov	16+0($key),$s0
223
224	and	\$0xff000000,$acc0
225	and	\$0xff000000,$acc1
226
227	xor	$acc0,$t2
228	xor	$acc1,$t3
229
230	mov	16+4($key),$s1
231	mov	16+8($key),$s2
232	xor	$t0,$s0
233	xor	$t1,$s1
234	xor	$t2,$s2
235	xor	$t3,$s3
236___
237}
238
239sub encstep()
240{ my ($i,@s) = @_;
241  my $tmp0=$acc0;
242  my $tmp1=$acc1;
243  my $tmp2=$acc2;
244  my $out=($t0,$t1,$t2,$s[0])[$i];
245
246	if ($i==3) {
247		$tmp0=$s[1];
248		$tmp1=$s[2];
249		$tmp2=$s[3];
250	}
251	$code.="	movzb	".&lo($s[0]).",$out\n";
252	$code.="	mov	$s[2],$tmp1\n"		if ($i!=3);
253	$code.="	lea	16($key),$key\n"	if ($i==0);
254
255	$code.="	movzb	".&hi($s[1]).",$tmp0\n";
256	$code.="	mov	0($sbox,$out,8),$out\n";
257
258	$code.="	shr	\$16,$tmp1\n";
259	$code.="	mov	$s[3],$tmp2\n"		if ($i!=3);
260	$code.="	xor	3($sbox,$tmp0,8),$out\n";
261
262	$code.="	movzb	".&lo($tmp1).",$tmp1\n";
263	$code.="	shr	\$24,$tmp2\n";
264	$code.="	xor	4*$i($key),$out\n";
265
266	$code.="	xor	2($sbox,$tmp1,8),$out\n";
267	$code.="	xor	1($sbox,$tmp2,8),$out\n";
268
269	$code.="	mov	$t0,$s[1]\n"		if ($i==3);
270	$code.="	mov	$t1,$s[2]\n"		if ($i==3);
271	$code.="	mov	$t2,$s[3]\n"		if ($i==3);
272	$code.="\n";
273}
274
275sub enclast()
276{ my ($i,@s)=@_;
277  my $tmp0=$acc0;
278  my $tmp1=$acc1;
279  my $tmp2=$acc2;
280  my $out=($t0,$t1,$t2,$s[0])[$i];
281
282	if ($i==3) {
283		$tmp0=$s[1];
284		$tmp1=$s[2];
285		$tmp2=$s[3];
286	}
287	$code.="	movzb	".&lo($s[0]).",$out\n";
288	$code.="	mov	$s[2],$tmp1\n"		if ($i!=3);
289
290	$code.="	mov	2($sbox,$out,8),$out\n";
291	$code.="	shr	\$16,$tmp1\n";
292	$code.="	mov	$s[3],$tmp2\n"		if ($i!=3);
293
294	$code.="	and	\$0x000000ff,$out\n";
295	$code.="	movzb	".&hi($s[1]).",$tmp0\n";
296	$code.="	movzb	".&lo($tmp1).",$tmp1\n";
297	$code.="	shr	\$24,$tmp2\n";
298
299	$code.="	mov	0($sbox,$tmp0,8),$tmp0\n";
300	$code.="	mov	0($sbox,$tmp1,8),$tmp1\n";
301	$code.="	mov	2($sbox,$tmp2,8),$tmp2\n";
302
303	$code.="	and	\$0x0000ff00,$tmp0\n";
304	$code.="	and	\$0x00ff0000,$tmp1\n";
305	$code.="	and	\$0xff000000,$tmp2\n";
306
307	$code.="	xor	$tmp0,$out\n";
308	$code.="	mov	$t0,$s[1]\n"		if ($i==3);
309	$code.="	xor	$tmp1,$out\n";
310	$code.="	mov	$t1,$s[2]\n"		if ($i==3);
311	$code.="	xor	$tmp2,$out\n";
312	$code.="	mov	$t2,$s[3]\n"		if ($i==3);
313	$code.="\n";
314}
315
316$code.=<<___;
317.type	_x86_64_AES_encrypt,\@abi-omnipotent
318.align	16
319_x86_64_AES_encrypt:
320	xor	0($key),$s0			# xor with key
321	xor	4($key),$s1
322	xor	8($key),$s2
323	xor	12($key),$s3
324
325	mov	240($key),$rnds			# load key->rounds
326	sub	\$1,$rnds
327	jmp	.Lenc_loop
328.align	16
329.Lenc_loop:
330___
331	if ($verticalspin) { &encvert(); }
332	else {	&encstep(0,$s0,$s1,$s2,$s3);
333		&encstep(1,$s1,$s2,$s3,$s0);
334		&encstep(2,$s2,$s3,$s0,$s1);
335		&encstep(3,$s3,$s0,$s1,$s2);
336	}
337$code.=<<___;
338	sub	\$1,$rnds
339	jnz	.Lenc_loop
340___
341	if ($verticalspin) { &enclastvert(); }
342	else {	&enclast(0,$s0,$s1,$s2,$s3);
343		&enclast(1,$s1,$s2,$s3,$s0);
344		&enclast(2,$s2,$s3,$s0,$s1);
345		&enclast(3,$s3,$s0,$s1,$s2);
346		$code.=<<___;
347		xor	16+0($key),$s0		# xor with key
348		xor	16+4($key),$s1
349		xor	16+8($key),$s2
350		xor	16+12($key),$s3
351___
352	}
353$code.=<<___;
354	.byte	0xf3,0xc3			# rep ret
355.size	_x86_64_AES_encrypt,.-_x86_64_AES_encrypt
356___
357
358# it's possible to implement this by shifting tN by 8, filling least
359# significant byte with byte load and finally bswap-ing at the end,
360# but such partial register load kills Core 2...
361sub enccompactvert()
362{ my ($t3,$t4,$t5)=("%r8d","%r9d","%r13d");
363
364$code.=<<___;
365	movzb	`&lo("$s0")`,$t0
366	movzb	`&lo("$s1")`,$t1
367	movzb	`&lo("$s2")`,$t2
368	movzb	($sbox,$t0,1),$t0
369	movzb	($sbox,$t1,1),$t1
370	movzb	($sbox,$t2,1),$t2
371
372	movzb	`&lo("$s3")`,$t3
373	movzb	`&hi("$s1")`,$acc0
374	movzb	`&hi("$s2")`,$acc1
375	movzb	($sbox,$t3,1),$t3
376	movzb	($sbox,$acc0,1),$t4	#$t0
377	movzb	($sbox,$acc1,1),$t5	#$t1
378
379	movzb	`&hi("$s3")`,$acc2
380	movzb	`&hi("$s0")`,$acc0
381	shr	\$16,$s2
382	movzb	($sbox,$acc2,1),$acc2	#$t2
383	movzb	($sbox,$acc0,1),$acc0	#$t3
384	shr	\$16,$s3
385
386	movzb	`&lo("$s2")`,$acc1
387	shl	\$8,$t4
388	shl	\$8,$t5
389	movzb	($sbox,$acc1,1),$acc1	#$t0
390	xor	$t4,$t0
391	xor	$t5,$t1
392
393	movzb	`&lo("$s3")`,$t4
394	shr	\$16,$s0
395	shr	\$16,$s1
396	movzb	`&lo("$s0")`,$t5
397	shl	\$8,$acc2
398	shl	\$8,$acc0
399	movzb	($sbox,$t4,1),$t4	#$t1
400	movzb	($sbox,$t5,1),$t5	#$t2
401	xor	$acc2,$t2
402	xor	$acc0,$t3
403
404	movzb	`&lo("$s1")`,$acc2
405	movzb	`&hi("$s3")`,$acc0
406	shl	\$16,$acc1
407	movzb	($sbox,$acc2,1),$acc2	#$t3
408	movzb	($sbox,$acc0,1),$acc0	#$t0
409	xor	$acc1,$t0
410
411	movzb	`&hi("$s0")`,$acc1
412	shr	\$8,$s2
413	shr	\$8,$s1
414	movzb	($sbox,$acc1,1),$acc1	#$t1
415	movzb	($sbox,$s2,1),$s3	#$t3
416	movzb	($sbox,$s1,1),$s2	#$t2
417	shl	\$16,$t4
418	shl	\$16,$t5
419	shl	\$16,$acc2
420	xor	$t4,$t1
421	xor	$t5,$t2
422	xor	$acc2,$t3
423
424	shl	\$24,$acc0
425	shl	\$24,$acc1
426	shl	\$24,$s3
427	xor	$acc0,$t0
428	shl	\$24,$s2
429	xor	$acc1,$t1
430	mov	$t0,$s0
431	mov	$t1,$s1
432	xor	$t2,$s2
433	xor	$t3,$s3
434___
435}
436
437sub enctransform_ref()
438{ my $sn = shift;
439  my ($acc,$r2,$tmp)=("%r8d","%r9d","%r13d");
440
441$code.=<<___;
442	mov	$sn,$acc
443	and	\$0x80808080,$acc
444	mov	$acc,$tmp
445	shr	\$7,$tmp
446	lea	($sn,$sn),$r2
447	sub	$tmp,$acc
448	and	\$0xfefefefe,$r2
449	and	\$0x1b1b1b1b,$acc
450	mov	$sn,$tmp
451	xor	$acc,$r2
452
453	xor	$r2,$sn
454	rol	\$24,$sn
455	xor	$r2,$sn
456	ror	\$16,$tmp
457	xor	$tmp,$sn
458	ror	\$8,$tmp
459	xor	$tmp,$sn
460___
461}
462
463# unlike decrypt case it does not pay off to parallelize enctransform
464sub enctransform()
465{ my ($t3,$r20,$r21)=($acc2,"%r8d","%r9d");
466
467$code.=<<___;
468	mov	$s0,$acc0
469	mov	$s1,$acc1
470	and	\$0x80808080,$acc0
471	and	\$0x80808080,$acc1
472	mov	$acc0,$t0
473	mov	$acc1,$t1
474	shr	\$7,$t0
475	lea	($s0,$s0),$r20
476	shr	\$7,$t1
477	lea	($s1,$s1),$r21
478	sub	$t0,$acc0
479	sub	$t1,$acc1
480	and	\$0xfefefefe,$r20
481	and	\$0xfefefefe,$r21
482	and	\$0x1b1b1b1b,$acc0
483	and	\$0x1b1b1b1b,$acc1
484	mov	$s0,$t0
485	mov	$s1,$t1
486	xor	$acc0,$r20
487	xor	$acc1,$r21
488
489	xor	$r20,$s0
490	xor	$r21,$s1
491	 mov	$s2,$acc0
492	 mov	$s3,$acc1
493	rol	\$24,$s0
494	rol	\$24,$s1
495	 and	\$0x80808080,$acc0
496	 and	\$0x80808080,$acc1
497	xor	$r20,$s0
498	xor	$r21,$s1
499	 mov	$acc0,$t2
500	 mov	$acc1,$t3
501	ror	\$16,$t0
502	ror	\$16,$t1
503	 shr	\$7,$t2
504	 lea	($s2,$s2),$r20
505	xor	$t0,$s0
506	xor	$t1,$s1
507	 shr	\$7,$t3
508	 lea	($s3,$s3),$r21
509	ror	\$8,$t0
510	ror	\$8,$t1
511	 sub	$t2,$acc0
512	 sub	$t3,$acc1
513	xor	$t0,$s0
514	xor	$t1,$s1
515
516	and	\$0xfefefefe,$r20
517	and	\$0xfefefefe,$r21
518	and	\$0x1b1b1b1b,$acc0
519	and	\$0x1b1b1b1b,$acc1
520	mov	$s2,$t2
521	mov	$s3,$t3
522	xor	$acc0,$r20
523	xor	$acc1,$r21
524
525	xor	$r20,$s2
526	xor	$r21,$s3
527	rol	\$24,$s2
528	rol	\$24,$s3
529	xor	$r20,$s2
530	xor	$r21,$s3
531	mov	0($sbox),$acc0			# prefetch Te4
532	ror	\$16,$t2
533	ror	\$16,$t3
534	mov	64($sbox),$acc1
535	xor	$t2,$s2
536	xor	$t3,$s3
537	mov	128($sbox),$r20
538	ror	\$8,$t2
539	ror	\$8,$t3
540	mov	192($sbox),$r21
541	xor	$t2,$s2
542	xor	$t3,$s3
543___
544}
545
546$code.=<<___;
547.type	_x86_64_AES_encrypt_compact,\@abi-omnipotent
548.align	16
549_x86_64_AES_encrypt_compact:
550	lea	128($sbox),$inp			# size optimization
551	mov	0-128($inp),$acc1		# prefetch Te4
552	mov	32-128($inp),$acc2
553	mov	64-128($inp),$t0
554	mov	96-128($inp),$t1
555	mov	128-128($inp),$acc1
556	mov	160-128($inp),$acc2
557	mov	192-128($inp),$t0
558	mov	224-128($inp),$t1
559	jmp	.Lenc_loop_compact
560.align	16
561.Lenc_loop_compact:
562		xor	0($key),$s0		# xor with key
563		xor	4($key),$s1
564		xor	8($key),$s2
565		xor	12($key),$s3
566		lea	16($key),$key
567___
568		&enccompactvert();
569$code.=<<___;
570		cmp	16(%rsp),$key
571		je	.Lenc_compact_done
572___
573		&enctransform();
574$code.=<<___;
575	jmp	.Lenc_loop_compact
576.align	16
577.Lenc_compact_done:
578	xor	0($key),$s0
579	xor	4($key),$s1
580	xor	8($key),$s2
581	xor	12($key),$s3
582	.byte	0xf3,0xc3			# rep ret
583.size	_x86_64_AES_encrypt_compact,.-_x86_64_AES_encrypt_compact
584___
585
586# void AES_encrypt (const void *inp,void *out,const AES_KEY *key);
587$code.=<<___;
588.globl	AES_encrypt
589.type	AES_encrypt,\@function,3
590.align	16
591.globl	asm_AES_encrypt
592.hidden	asm_AES_encrypt
593asm_AES_encrypt:
594AES_encrypt:
595	push	%rbx
596	push	%rbp
597	push	%r12
598	push	%r13
599	push	%r14
600	push	%r15
601
602	# allocate frame "above" key schedule
603	mov	%rsp,%r10
604	lea	-63(%rdx),%rcx	# %rdx is key argument
605	and	\$-64,%rsp
606	sub	%rsp,%rcx
607	neg	%rcx
608	and	\$0x3c0,%rcx
609	sub	%rcx,%rsp
610	sub	\$32,%rsp
611
612	mov	%rsi,16(%rsp)	# save out
613	mov	%r10,24(%rsp)	# save real stack pointer
614.Lenc_prologue:
615
616	mov	%rdx,$key
617	mov	240($key),$rnds	# load rounds
618
619	mov	0(%rdi),$s0	# load input vector
620	mov	4(%rdi),$s1
621	mov	8(%rdi),$s2
622	mov	12(%rdi),$s3
623
624	shl	\$4,$rnds
625	lea	($key,$rnds),%rbp
626	mov	$key,(%rsp)	# key schedule
627	mov	%rbp,8(%rsp)	# end of key schedule
628
629	# pick Te4 copy which can't "overlap" with stack frame or key schedule
630	lea	.LAES_Te+2048(%rip),$sbox
631	lea	768(%rsp),%rbp
632	sub	$sbox,%rbp
633	and	\$0x300,%rbp
634	lea	($sbox,%rbp),$sbox
635
636	call	_x86_64_AES_encrypt_compact
637
638	mov	16(%rsp),$out	# restore out
639	mov	24(%rsp),%rsi	# restore saved stack pointer
640	mov	$s0,0($out)	# write output vector
641	mov	$s1,4($out)
642	mov	$s2,8($out)
643	mov	$s3,12($out)
644
645	mov	(%rsi),%r15
646	mov	8(%rsi),%r14
647	mov	16(%rsi),%r13
648	mov	24(%rsi),%r12
649	mov	32(%rsi),%rbp
650	mov	40(%rsi),%rbx
651	lea	48(%rsi),%rsp
652.Lenc_epilogue:
653	ret
654.size	AES_encrypt,.-AES_encrypt
655___
656
657#------------------------------------------------------------------#
658
659sub decvert()
660{ my $t3="%r8d";	# zaps $inp!
661
662$code.=<<___;
663	# favor 3-way issue Opteron pipeline...
664	movzb	`&lo("$s0")`,$acc0
665	movzb	`&lo("$s1")`,$acc1
666	movzb	`&lo("$s2")`,$acc2
667	mov	0($sbox,$acc0,8),$t0
668	mov	0($sbox,$acc1,8),$t1
669	mov	0($sbox,$acc2,8),$t2
670
671	movzb	`&hi("$s3")`,$acc0
672	movzb	`&hi("$s0")`,$acc1
673	movzb	`&lo("$s3")`,$acc2
674	xor	3($sbox,$acc0,8),$t0
675	xor	3($sbox,$acc1,8),$t1
676	mov	0($sbox,$acc2,8),$t3
677
678	movzb	`&hi("$s1")`,$acc0
679	shr	\$16,$s0
680	movzb	`&hi("$s2")`,$acc2
681	xor	3($sbox,$acc0,8),$t2
682	shr	\$16,$s3
683	xor	3($sbox,$acc2,8),$t3
684
685	shr	\$16,$s1
686	lea	16($key),$key
687	shr	\$16,$s2
688
689	movzb	`&lo("$s2")`,$acc0
690	movzb	`&lo("$s3")`,$acc1
691	movzb	`&lo("$s0")`,$acc2
692	xor	2($sbox,$acc0,8),$t0
693	xor	2($sbox,$acc1,8),$t1
694	xor	2($sbox,$acc2,8),$t2
695
696	movzb	`&hi("$s1")`,$acc0
697	movzb	`&hi("$s2")`,$acc1
698	movzb	`&lo("$s1")`,$acc2
699	xor	1($sbox,$acc0,8),$t0
700	xor	1($sbox,$acc1,8),$t1
701	xor	2($sbox,$acc2,8),$t3
702
703	movzb	`&hi("$s3")`,$acc0
704	mov	12($key),$s3
705	movzb	`&hi("$s0")`,$acc2
706	xor	1($sbox,$acc0,8),$t2
707	mov	0($key),$s0
708	xor	1($sbox,$acc2,8),$t3
709
710	xor	$t0,$s0
711	mov	4($key),$s1
712	mov	8($key),$s2
713	xor	$t2,$s2
714	xor	$t1,$s1
715	xor	$t3,$s3
716___
717}
718
719sub declastvert()
720{ my $t3="%r8d";	# zaps $inp!
721
722$code.=<<___;
723	lea	2048($sbox),$sbox	# size optimization
724	movzb	`&lo("$s0")`,$acc0
725	movzb	`&lo("$s1")`,$acc1
726	movzb	`&lo("$s2")`,$acc2
727	movzb	($sbox,$acc0,1),$t0
728	movzb	($sbox,$acc1,1),$t1
729	movzb	($sbox,$acc2,1),$t2
730
731	movzb	`&lo("$s3")`,$acc0
732	movzb	`&hi("$s3")`,$acc1
733	movzb	`&hi("$s0")`,$acc2
734	movzb	($sbox,$acc0,1),$t3
735	movzb	($sbox,$acc1,1),$acc1	#$t0
736	movzb	($sbox,$acc2,1),$acc2	#$t1
737
738	shl	\$8,$acc1
739	shl	\$8,$acc2
740
741	xor	$acc1,$t0
742	xor	$acc2,$t1
743	shr	\$16,$s3
744
745	movzb	`&hi("$s1")`,$acc0
746	movzb	`&hi("$s2")`,$acc1
747	shr	\$16,$s0
748	movzb	($sbox,$acc0,1),$acc0	#$t2
749	movzb	($sbox,$acc1,1),$acc1	#$t3
750
751	shl	\$8,$acc0
752	shl	\$8,$acc1
753	shr	\$16,$s1
754	xor	$acc0,$t2
755	xor	$acc1,$t3
756	shr	\$16,$s2
757
758	movzb	`&lo("$s2")`,$acc0
759	movzb	`&lo("$s3")`,$acc1
760	movzb	`&lo("$s0")`,$acc2
761	movzb	($sbox,$acc0,1),$acc0	#$t0
762	movzb	($sbox,$acc1,1),$acc1	#$t1
763	movzb	($sbox,$acc2,1),$acc2	#$t2
764
765	shl	\$16,$acc0
766	shl	\$16,$acc1
767	shl	\$16,$acc2
768
769	xor	$acc0,$t0
770	xor	$acc1,$t1
771	xor	$acc2,$t2
772
773	movzb	`&lo("$s1")`,$acc0
774	movzb	`&hi("$s1")`,$acc1
775	movzb	`&hi("$s2")`,$acc2
776	movzb	($sbox,$acc0,1),$acc0	#$t3
777	movzb	($sbox,$acc1,1),$acc1	#$t0
778	movzb	($sbox,$acc2,1),$acc2	#$t1
779
780	shl	\$16,$acc0
781	shl	\$24,$acc1
782	shl	\$24,$acc2
783
784	xor	$acc0,$t3
785	xor	$acc1,$t0
786	xor	$acc2,$t1
787
788	movzb	`&hi("$s3")`,$acc0
789	movzb	`&hi("$s0")`,$acc1
790	mov	16+12($key),$s3
791	movzb	($sbox,$acc0,1),$acc0	#$t2
792	movzb	($sbox,$acc1,1),$acc1	#$t3
793	mov	16+0($key),$s0
794
795	shl	\$24,$acc0
796	shl	\$24,$acc1
797
798	xor	$acc0,$t2
799	xor	$acc1,$t3
800
801	mov	16+4($key),$s1
802	mov	16+8($key),$s2
803	lea	-2048($sbox),$sbox
804	xor	$t0,$s0
805	xor	$t1,$s1
806	xor	$t2,$s2
807	xor	$t3,$s3
808___
809}
810
811sub decstep()
812{ my ($i,@s) = @_;
813  my $tmp0=$acc0;
814  my $tmp1=$acc1;
815  my $tmp2=$acc2;
816  my $out=($t0,$t1,$t2,$s[0])[$i];
817
818	$code.="	mov	$s[0],$out\n"		if ($i!=3);
819			$tmp1=$s[2]			if ($i==3);
820	$code.="	mov	$s[2],$tmp1\n"		if ($i!=3);
821	$code.="	and	\$0xFF,$out\n";
822
823	$code.="	mov	0($sbox,$out,8),$out\n";
824	$code.="	shr	\$16,$tmp1\n";
825			$tmp2=$s[3]			if ($i==3);
826	$code.="	mov	$s[3],$tmp2\n"		if ($i!=3);
827
828			$tmp0=$s[1]			if ($i==3);
829	$code.="	movzb	".&hi($s[1]).",$tmp0\n";
830	$code.="	and	\$0xFF,$tmp1\n";
831	$code.="	shr	\$24,$tmp2\n";
832
833	$code.="	xor	3($sbox,$tmp0,8),$out\n";
834	$code.="	xor	2($sbox,$tmp1,8),$out\n";
835	$code.="	xor	1($sbox,$tmp2,8),$out\n";
836
837	$code.="	mov	$t2,$s[1]\n"		if ($i==3);
838	$code.="	mov	$t1,$s[2]\n"		if ($i==3);
839	$code.="	mov	$t0,$s[3]\n"		if ($i==3);
840	$code.="\n";
841}
842
843sub declast()
844{ my ($i,@s)=@_;
845  my $tmp0=$acc0;
846  my $tmp1=$acc1;
847  my $tmp2=$acc2;
848  my $out=($t0,$t1,$t2,$s[0])[$i];
849
850	$code.="	mov	$s[0],$out\n"		if ($i!=3);
851			$tmp1=$s[2]			if ($i==3);
852	$code.="	mov	$s[2],$tmp1\n"		if ($i!=3);
853	$code.="	and	\$0xFF,$out\n";
854
855	$code.="	movzb	2048($sbox,$out,1),$out\n";
856	$code.="	shr	\$16,$tmp1\n";
857			$tmp2=$s[3]			if ($i==3);
858	$code.="	mov	$s[3],$tmp2\n"		if ($i!=3);
859
860			$tmp0=$s[1]			if ($i==3);
861	$code.="	movzb	".&hi($s[1]).",$tmp0\n";
862	$code.="	and	\$0xFF,$tmp1\n";
863	$code.="	shr	\$24,$tmp2\n";
864
865	$code.="	movzb	2048($sbox,$tmp0,1),$tmp0\n";
866	$code.="	movzb	2048($sbox,$tmp1,1),$tmp1\n";
867	$code.="	movzb	2048($sbox,$tmp2,1),$tmp2\n";
868
869	$code.="	shl	\$8,$tmp0\n";
870	$code.="	shl	\$16,$tmp1\n";
871	$code.="	shl	\$24,$tmp2\n";
872
873	$code.="	xor	$tmp0,$out\n";
874	$code.="	mov	$t2,$s[1]\n"		if ($i==3);
875	$code.="	xor	$tmp1,$out\n";
876	$code.="	mov	$t1,$s[2]\n"		if ($i==3);
877	$code.="	xor	$tmp2,$out\n";
878	$code.="	mov	$t0,$s[3]\n"		if ($i==3);
879	$code.="\n";
880}
881
882$code.=<<___;
883.type	_x86_64_AES_decrypt,\@abi-omnipotent
884.align	16
885_x86_64_AES_decrypt:
886	xor	0($key),$s0			# xor with key
887	xor	4($key),$s1
888	xor	8($key),$s2
889	xor	12($key),$s3
890
891	mov	240($key),$rnds			# load key->rounds
892	sub	\$1,$rnds
893	jmp	.Ldec_loop
894.align	16
895.Ldec_loop:
896___
897	if ($verticalspin) { &decvert(); }
898	else {	&decstep(0,$s0,$s3,$s2,$s1);
899		&decstep(1,$s1,$s0,$s3,$s2);
900		&decstep(2,$s2,$s1,$s0,$s3);
901		&decstep(3,$s3,$s2,$s1,$s0);
902		$code.=<<___;
903		lea	16($key),$key
904		xor	0($key),$s0			# xor with key
905		xor	4($key),$s1
906		xor	8($key),$s2
907		xor	12($key),$s3
908___
909	}
910$code.=<<___;
911	sub	\$1,$rnds
912	jnz	.Ldec_loop
913___
914	if ($verticalspin) { &declastvert(); }
915	else {	&declast(0,$s0,$s3,$s2,$s1);
916		&declast(1,$s1,$s0,$s3,$s2);
917		&declast(2,$s2,$s1,$s0,$s3);
918		&declast(3,$s3,$s2,$s1,$s0);
919		$code.=<<___;
920		xor	16+0($key),$s0			# xor with key
921		xor	16+4($key),$s1
922		xor	16+8($key),$s2
923		xor	16+12($key),$s3
924___
925	}
926$code.=<<___;
927	.byte	0xf3,0xc3			# rep ret
928.size	_x86_64_AES_decrypt,.-_x86_64_AES_decrypt
929___
930
931sub deccompactvert()
932{ my ($t3,$t4,$t5)=("%r8d","%r9d","%r13d");
933
934$code.=<<___;
935	movzb	`&lo("$s0")`,$t0
936	movzb	`&lo("$s1")`,$t1
937	movzb	`&lo("$s2")`,$t2
938	movzb	($sbox,$t0,1),$t0
939	movzb	($sbox,$t1,1),$t1
940	movzb	($sbox,$t2,1),$t2
941
942	movzb	`&lo("$s3")`,$t3
943	movzb	`&hi("$s3")`,$acc0
944	movzb	`&hi("$s0")`,$acc1
945	movzb	($sbox,$t3,1),$t3
946	movzb	($sbox,$acc0,1),$t4	#$t0
947	movzb	($sbox,$acc1,1),$t5	#$t1
948
949	movzb	`&hi("$s1")`,$acc2
950	movzb	`&hi("$s2")`,$acc0
951	shr	\$16,$s2
952	movzb	($sbox,$acc2,1),$acc2	#$t2
953	movzb	($sbox,$acc0,1),$acc0	#$t3
954	shr	\$16,$s3
955
956	movzb	`&lo("$s2")`,$acc1
957	shl	\$8,$t4
958	shl	\$8,$t5
959	movzb	($sbox,$acc1,1),$acc1	#$t0
960	xor	$t4,$t0
961	xor	$t5,$t1
962
963	movzb	`&lo("$s3")`,$t4
964	shr	\$16,$s0
965	shr	\$16,$s1
966	movzb	`&lo("$s0")`,$t5
967	shl	\$8,$acc2
968	shl	\$8,$acc0
969	movzb	($sbox,$t4,1),$t4	#$t1
970	movzb	($sbox,$t5,1),$t5	#$t2
971	xor	$acc2,$t2
972	xor	$acc0,$t3
973
974	movzb	`&lo("$s1")`,$acc2
975	movzb	`&hi("$s1")`,$acc0
976	shl	\$16,$acc1
977	movzb	($sbox,$acc2,1),$acc2	#$t3
978	movzb	($sbox,$acc0,1),$acc0	#$t0
979	xor	$acc1,$t0
980
981	movzb	`&hi("$s2")`,$acc1
982	shl	\$16,$t4
983	shl	\$16,$t5
984	movzb	($sbox,$acc1,1),$s1	#$t1
985	xor	$t4,$t1
986	xor	$t5,$t2
987
988	movzb	`&hi("$s3")`,$acc1
989	shr	\$8,$s0
990	shl	\$16,$acc2
991	movzb	($sbox,$acc1,1),$s2	#$t2
992	movzb	($sbox,$s0,1),$s3	#$t3
993	xor	$acc2,$t3
994
995	shl	\$24,$acc0
996	shl	\$24,$s1
997	shl	\$24,$s2
998	xor	$acc0,$t0
999	shl	\$24,$s3
1000	xor	$t1,$s1
1001	mov	$t0,$s0
1002	xor	$t2,$s2
1003	xor	$t3,$s3
1004___
1005}
1006
1007# parallelized version! input is pair of 64-bit values: %rax=s1.s0
1008# and %rcx=s3.s2, output is four 32-bit values in %eax=s0, %ebx=s1,
1009# %ecx=s2 and %edx=s3.
1010sub dectransform()
1011{ my ($tp10,$tp20,$tp40,$tp80,$acc0)=("%rax","%r8", "%r9", "%r10","%rbx");
1012  my ($tp18,$tp28,$tp48,$tp88,$acc8)=("%rcx","%r11","%r12","%r13","%rdx");
1013  my $prefetch = shift;
1014
1015$code.=<<___;
1016	mov	$tp10,$acc0
1017	mov	$tp18,$acc8
1018	and	$mask80,$acc0
1019	and	$mask80,$acc8
1020	mov	$acc0,$tp40
1021	mov	$acc8,$tp48
1022	shr	\$7,$tp40
1023	lea	($tp10,$tp10),$tp20
1024	shr	\$7,$tp48
1025	lea	($tp18,$tp18),$tp28
1026	sub	$tp40,$acc0
1027	sub	$tp48,$acc8
1028	and	$maskfe,$tp20
1029	and	$maskfe,$tp28
1030	and	$mask1b,$acc0
1031	and	$mask1b,$acc8
1032	xor	$tp20,$acc0
1033	xor	$tp28,$acc8
1034	mov	$acc0,$tp20
1035	mov	$acc8,$tp28
1036
1037	and	$mask80,$acc0
1038	and	$mask80,$acc8
1039	mov	$acc0,$tp80
1040	mov	$acc8,$tp88
1041	shr	\$7,$tp80
1042	lea	($tp20,$tp20),$tp40
1043	shr	\$7,$tp88
1044	lea	($tp28,$tp28),$tp48
1045	sub	$tp80,$acc0
1046	sub	$tp88,$acc8
1047	and	$maskfe,$tp40
1048	and	$maskfe,$tp48
1049	and	$mask1b,$acc0
1050	and	$mask1b,$acc8
1051	xor	$tp40,$acc0
1052	xor	$tp48,$acc8
1053	mov	$acc0,$tp40
1054	mov	$acc8,$tp48
1055
1056	and	$mask80,$acc0
1057	and	$mask80,$acc8
1058	mov	$acc0,$tp80
1059	mov	$acc8,$tp88
1060	shr	\$7,$tp80
1061	 xor	$tp10,$tp20		# tp2^=tp1
1062	shr	\$7,$tp88
1063	 xor	$tp18,$tp28		# tp2^=tp1
1064	sub	$tp80,$acc0
1065	sub	$tp88,$acc8
1066	lea	($tp40,$tp40),$tp80
1067	lea	($tp48,$tp48),$tp88
1068	 xor	$tp10,$tp40		# tp4^=tp1
1069	 xor	$tp18,$tp48		# tp4^=tp1
1070	and	$maskfe,$tp80
1071	and	$maskfe,$tp88
1072	and	$mask1b,$acc0
1073	and	$mask1b,$acc8
1074	xor	$acc0,$tp80
1075	xor	$acc8,$tp88
1076
1077	xor	$tp80,$tp10		# tp1^=tp8
1078	xor	$tp88,$tp18		# tp1^=tp8
1079	xor	$tp80,$tp20		# tp2^tp1^=tp8
1080	xor	$tp88,$tp28		# tp2^tp1^=tp8
1081	mov	$tp10,$acc0
1082	mov	$tp18,$acc8
1083	xor	$tp80,$tp40		# tp4^tp1^=tp8
1084	xor	$tp88,$tp48		# tp4^tp1^=tp8
1085	shr	\$32,$acc0
1086	shr	\$32,$acc8
1087	xor	$tp20,$tp80		# tp8^=tp8^tp2^tp1=tp2^tp1
1088	xor	$tp28,$tp88		# tp8^=tp8^tp2^tp1=tp2^tp1
1089	rol	\$8,`&LO("$tp10")`	# ROTATE(tp1^tp8,8)
1090	rol	\$8,`&LO("$tp18")`	# ROTATE(tp1^tp8,8)
1091	xor	$tp40,$tp80		# tp2^tp1^=tp8^tp4^tp1=tp8^tp4^tp2
1092	xor	$tp48,$tp88		# tp2^tp1^=tp8^tp4^tp1=tp8^tp4^tp2
1093
1094	rol	\$8,`&LO("$acc0")`	# ROTATE(tp1^tp8,8)
1095	rol	\$8,`&LO("$acc8")`	# ROTATE(tp1^tp8,8)
1096	xor	`&LO("$tp80")`,`&LO("$tp10")`
1097	xor	`&LO("$tp88")`,`&LO("$tp18")`
1098	shr	\$32,$tp80
1099	shr	\$32,$tp88
1100	xor	`&LO("$tp80")`,`&LO("$acc0")`
1101	xor	`&LO("$tp88")`,`&LO("$acc8")`
1102
1103	mov	$tp20,$tp80
1104	mov	$tp28,$tp88
1105	shr	\$32,$tp80
1106	shr	\$32,$tp88
1107	rol	\$24,`&LO("$tp20")`	# ROTATE(tp2^tp1^tp8,24)
1108	rol	\$24,`&LO("$tp28")`	# ROTATE(tp2^tp1^tp8,24)
1109	rol	\$24,`&LO("$tp80")`	# ROTATE(tp2^tp1^tp8,24)
1110	rol	\$24,`&LO("$tp88")`	# ROTATE(tp2^tp1^tp8,24)
1111	xor	`&LO("$tp20")`,`&LO("$tp10")`
1112	xor	`&LO("$tp28")`,`&LO("$tp18")`
1113	mov	$tp40,$tp20
1114	mov	$tp48,$tp28
1115	xor	`&LO("$tp80")`,`&LO("$acc0")`
1116	xor	`&LO("$tp88")`,`&LO("$acc8")`
1117
1118	`"mov	0($sbox),$mask80"	if ($prefetch)`
1119	shr	\$32,$tp20
1120	shr	\$32,$tp28
1121	`"mov	64($sbox),$maskfe"	if ($prefetch)`
1122	rol	\$16,`&LO("$tp40")`	# ROTATE(tp4^tp1^tp8,16)
1123	rol	\$16,`&LO("$tp48")`	# ROTATE(tp4^tp1^tp8,16)
1124	`"mov	128($sbox),$mask1b"	if ($prefetch)`
1125	rol	\$16,`&LO("$tp20")`	# ROTATE(tp4^tp1^tp8,16)
1126	rol	\$16,`&LO("$tp28")`	# ROTATE(tp4^tp1^tp8,16)
1127	`"mov	192($sbox),$tp80"	if ($prefetch)`
1128	xor	`&LO("$tp40")`,`&LO("$tp10")`
1129	xor	`&LO("$tp48")`,`&LO("$tp18")`
1130	`"mov	256($sbox),$tp88"	if ($prefetch)`
1131	xor	`&LO("$tp20")`,`&LO("$acc0")`
1132	xor	`&LO("$tp28")`,`&LO("$acc8")`
1133___
1134}
1135
1136$code.=<<___;
1137.type	_x86_64_AES_decrypt_compact,\@abi-omnipotent
1138.align	16
1139_x86_64_AES_decrypt_compact:
1140	lea	128($sbox),$inp			# size optimization
1141	mov	0-128($inp),$acc1		# prefetch Td4
1142	mov	32-128($inp),$acc2
1143	mov	64-128($inp),$t0
1144	mov	96-128($inp),$t1
1145	mov	128-128($inp),$acc1
1146	mov	160-128($inp),$acc2
1147	mov	192-128($inp),$t0
1148	mov	224-128($inp),$t1
1149	jmp	.Ldec_loop_compact
1150
1151.align	16
1152.Ldec_loop_compact:
1153		xor	0($key),$s0		# xor with key
1154		xor	4($key),$s1
1155		xor	8($key),$s2
1156		xor	12($key),$s3
1157		lea	16($key),$key
1158___
1159		&deccompactvert();
1160$code.=<<___;
1161		cmp	16(%rsp),$key
1162		je	.Ldec_compact_done
1163
1164		mov	256+0($sbox),$mask80
1165		shl	\$32,%rbx
1166		shl	\$32,%rdx
1167		mov	256+8($sbox),$maskfe
1168		or	%rbx,%rax
1169		or	%rdx,%rcx
1170		mov	256+16($sbox),$mask1b
1171___
1172		&dectransform(1);
1173$code.=<<___;
1174	jmp	.Ldec_loop_compact
1175.align	16
1176.Ldec_compact_done:
1177	xor	0($key),$s0
1178	xor	4($key),$s1
1179	xor	8($key),$s2
1180	xor	12($key),$s3
1181	.byte	0xf3,0xc3			# rep ret
1182.size	_x86_64_AES_decrypt_compact,.-_x86_64_AES_decrypt_compact
1183___
1184
1185# void AES_decrypt (const void *inp,void *out,const AES_KEY *key);
1186$code.=<<___;
1187.globl	AES_decrypt
1188.type	AES_decrypt,\@function,3
1189.align	16
1190.globl	asm_AES_decrypt
1191.hidden	asm_AES_decrypt
1192asm_AES_decrypt:
1193AES_decrypt:
1194	push	%rbx
1195	push	%rbp
1196	push	%r12
1197	push	%r13
1198	push	%r14
1199	push	%r15
1200
1201	# allocate frame "above" key schedule
1202	mov	%rsp,%r10
1203	lea	-63(%rdx),%rcx	# %rdx is key argument
1204	and	\$-64,%rsp
1205	sub	%rsp,%rcx
1206	neg	%rcx
1207	and	\$0x3c0,%rcx
1208	sub	%rcx,%rsp
1209	sub	\$32,%rsp
1210
1211	mov	%rsi,16(%rsp)	# save out
1212	mov	%r10,24(%rsp)	# save real stack pointer
1213.Ldec_prologue:
1214
1215	mov	%rdx,$key
1216	mov	240($key),$rnds	# load rounds
1217
1218	mov	0(%rdi),$s0	# load input vector
1219	mov	4(%rdi),$s1
1220	mov	8(%rdi),$s2
1221	mov	12(%rdi),$s3
1222
1223	shl	\$4,$rnds
1224	lea	($key,$rnds),%rbp
1225	mov	$key,(%rsp)	# key schedule
1226	mov	%rbp,8(%rsp)	# end of key schedule
1227
1228	# pick Td4 copy which can't "overlap" with stack frame or key schedule
1229	lea	.LAES_Td+2048(%rip),$sbox
1230	lea	768(%rsp),%rbp
1231	sub	$sbox,%rbp
1232	and	\$0x300,%rbp
1233	lea	($sbox,%rbp),$sbox
1234	shr	\$3,%rbp	# recall "magic" constants!
1235	add	%rbp,$sbox
1236
1237	call	_x86_64_AES_decrypt_compact
1238
1239	mov	16(%rsp),$out	# restore out
1240	mov	24(%rsp),%rsi	# restore saved stack pointer
1241	mov	$s0,0($out)	# write output vector
1242	mov	$s1,4($out)
1243	mov	$s2,8($out)
1244	mov	$s3,12($out)
1245
1246	mov	(%rsi),%r15
1247	mov	8(%rsi),%r14
1248	mov	16(%rsi),%r13
1249	mov	24(%rsi),%r12
1250	mov	32(%rsi),%rbp
1251	mov	40(%rsi),%rbx
1252	lea	48(%rsi),%rsp
1253.Ldec_epilogue:
1254	ret
1255.size	AES_decrypt,.-AES_decrypt
1256___
1257#------------------------------------------------------------------#
1258
1259sub enckey()
1260{
1261$code.=<<___;
1262	movz	%dl,%esi		# rk[i]>>0
1263	movzb	-128(%rbp,%rsi),%ebx
1264	movz	%dh,%esi		# rk[i]>>8
1265	shl	\$24,%ebx
1266	xor	%ebx,%eax
1267
1268	movzb	-128(%rbp,%rsi),%ebx
1269	shr	\$16,%edx
1270	movz	%dl,%esi		# rk[i]>>16
1271	xor	%ebx,%eax
1272
1273	movzb	-128(%rbp,%rsi),%ebx
1274	movz	%dh,%esi		# rk[i]>>24
1275	shl	\$8,%ebx
1276	xor	%ebx,%eax
1277
1278	movzb	-128(%rbp,%rsi),%ebx
1279	shl	\$16,%ebx
1280	xor	%ebx,%eax
1281
1282	xor	1024-128(%rbp,%rcx,4),%eax		# rcon
1283___
1284}
1285
1286# int private_AES_set_encrypt_key(const unsigned char *userKey, const int bits,
1287#                        AES_KEY *key)
1288$code.=<<___;
1289.globl	private_AES_set_encrypt_key
1290.type	private_AES_set_encrypt_key,\@function,3
1291.align	16
1292private_AES_set_encrypt_key:
1293	push	%rbx
1294	push	%rbp
1295	push	%r12			# redundant, but allows to share
1296	push	%r13			# exception handler...
1297	push	%r14
1298	push	%r15
1299	sub	\$8,%rsp
1300.Lenc_key_prologue:
1301
1302	call	_x86_64_AES_set_encrypt_key
1303
1304	mov	8(%rsp),%r15
1305	mov	16(%rsp),%r14
1306	mov	24(%rsp),%r13
1307	mov	32(%rsp),%r12
1308	mov	40(%rsp),%rbp
1309	mov	48(%rsp),%rbx
1310	add	\$56,%rsp
1311.Lenc_key_epilogue:
1312	ret
1313.size	private_AES_set_encrypt_key,.-private_AES_set_encrypt_key
1314
1315.type	_x86_64_AES_set_encrypt_key,\@abi-omnipotent
1316.align	16
1317_x86_64_AES_set_encrypt_key:
1318	mov	%esi,%ecx			# %ecx=bits
1319	mov	%rdi,%rsi			# %rsi=userKey
1320	mov	%rdx,%rdi			# %rdi=key
1321
1322	test	\$-1,%rsi
1323	jz	.Lbadpointer
1324	test	\$-1,%rdi
1325	jz	.Lbadpointer
1326
1327	lea	.LAES_Te(%rip),%rbp
1328	lea	2048+128(%rbp),%rbp
1329
1330	# prefetch Te4
1331	mov	0-128(%rbp),%eax
1332	mov	32-128(%rbp),%ebx
1333	mov	64-128(%rbp),%r8d
1334	mov	96-128(%rbp),%edx
1335	mov	128-128(%rbp),%eax
1336	mov	160-128(%rbp),%ebx
1337	mov	192-128(%rbp),%r8d
1338	mov	224-128(%rbp),%edx
1339
1340	cmp	\$128,%ecx
1341	je	.L10rounds
1342	cmp	\$192,%ecx
1343	je	.L12rounds
1344	cmp	\$256,%ecx
1345	je	.L14rounds
1346	mov	\$-2,%rax			# invalid number of bits
1347	jmp	.Lexit
1348
1349.L10rounds:
1350	mov	0(%rsi),%rax			# copy first 4 dwords
1351	mov	8(%rsi),%rdx
1352	mov	%rax,0(%rdi)
1353	mov	%rdx,8(%rdi)
1354
1355	shr	\$32,%rdx
1356	xor	%ecx,%ecx
1357	jmp	.L10shortcut
1358.align	4
1359.L10loop:
1360		mov	0(%rdi),%eax			# rk[0]
1361		mov	12(%rdi),%edx			# rk[3]
1362.L10shortcut:
1363___
1364		&enckey	();
1365$code.=<<___;
1366		mov	%eax,16(%rdi)			# rk[4]
1367		xor	4(%rdi),%eax
1368		mov	%eax,20(%rdi)			# rk[5]
1369		xor	8(%rdi),%eax
1370		mov	%eax,24(%rdi)			# rk[6]
1371		xor	12(%rdi),%eax
1372		mov	%eax,28(%rdi)			# rk[7]
1373		add	\$1,%ecx
1374		lea	16(%rdi),%rdi
1375		cmp	\$10,%ecx
1376	jl	.L10loop
1377
1378	movl	\$10,80(%rdi)			# setup number of rounds
1379	xor	%rax,%rax
1380	jmp	.Lexit
1381
1382.L12rounds:
1383	mov	0(%rsi),%rax			# copy first 6 dwords
1384	mov	8(%rsi),%rbx
1385	mov	16(%rsi),%rdx
1386	mov	%rax,0(%rdi)
1387	mov	%rbx,8(%rdi)
1388	mov	%rdx,16(%rdi)
1389
1390	shr	\$32,%rdx
1391	xor	%ecx,%ecx
1392	jmp	.L12shortcut
1393.align	4
1394.L12loop:
1395		mov	0(%rdi),%eax			# rk[0]
1396		mov	20(%rdi),%edx			# rk[5]
1397.L12shortcut:
1398___
1399		&enckey	();
1400$code.=<<___;
1401		mov	%eax,24(%rdi)			# rk[6]
1402		xor	4(%rdi),%eax
1403		mov	%eax,28(%rdi)			# rk[7]
1404		xor	8(%rdi),%eax
1405		mov	%eax,32(%rdi)			# rk[8]
1406		xor	12(%rdi),%eax
1407		mov	%eax,36(%rdi)			# rk[9]
1408
1409		cmp	\$7,%ecx
1410		je	.L12break
1411		add	\$1,%ecx
1412
1413		xor	16(%rdi),%eax
1414		mov	%eax,40(%rdi)			# rk[10]
1415		xor	20(%rdi),%eax
1416		mov	%eax,44(%rdi)			# rk[11]
1417
1418		lea	24(%rdi),%rdi
1419	jmp	.L12loop
1420.L12break:
1421	movl	\$12,72(%rdi)		# setup number of rounds
1422	xor	%rax,%rax
1423	jmp	.Lexit
1424
1425.L14rounds:
1426	mov	0(%rsi),%rax			# copy first 8 dwords
1427	mov	8(%rsi),%rbx
1428	mov	16(%rsi),%rcx
1429	mov	24(%rsi),%rdx
1430	mov	%rax,0(%rdi)
1431	mov	%rbx,8(%rdi)
1432	mov	%rcx,16(%rdi)
1433	mov	%rdx,24(%rdi)
1434
1435	shr	\$32,%rdx
1436	xor	%ecx,%ecx
1437	jmp	.L14shortcut
1438.align	4
1439.L14loop:
1440		mov	0(%rdi),%eax			# rk[0]
1441		mov	28(%rdi),%edx			# rk[4]
1442.L14shortcut:
1443___
1444		&enckey	();
1445$code.=<<___;
1446		mov	%eax,32(%rdi)			# rk[8]
1447		xor	4(%rdi),%eax
1448		mov	%eax,36(%rdi)			# rk[9]
1449		xor	8(%rdi),%eax
1450		mov	%eax,40(%rdi)			# rk[10]
1451		xor	12(%rdi),%eax
1452		mov	%eax,44(%rdi)			# rk[11]
1453
1454		cmp	\$6,%ecx
1455		je	.L14break
1456		add	\$1,%ecx
1457
1458		mov	%eax,%edx
1459		mov	16(%rdi),%eax			# rk[4]
1460		movz	%dl,%esi			# rk[11]>>0
1461		movzb	-128(%rbp,%rsi),%ebx
1462		movz	%dh,%esi			# rk[11]>>8
1463		xor	%ebx,%eax
1464
1465		movzb	-128(%rbp,%rsi),%ebx
1466		shr	\$16,%edx
1467		shl	\$8,%ebx
1468		movz	%dl,%esi			# rk[11]>>16
1469		xor	%ebx,%eax
1470
1471		movzb	-128(%rbp,%rsi),%ebx
1472		movz	%dh,%esi			# rk[11]>>24
1473		shl	\$16,%ebx
1474		xor	%ebx,%eax
1475
1476		movzb	-128(%rbp,%rsi),%ebx
1477		shl	\$24,%ebx
1478		xor	%ebx,%eax
1479
1480		mov	%eax,48(%rdi)			# rk[12]
1481		xor	20(%rdi),%eax
1482		mov	%eax,52(%rdi)			# rk[13]
1483		xor	24(%rdi),%eax
1484		mov	%eax,56(%rdi)			# rk[14]
1485		xor	28(%rdi),%eax
1486		mov	%eax,60(%rdi)			# rk[15]
1487
1488		lea	32(%rdi),%rdi
1489	jmp	.L14loop
1490.L14break:
1491	movl	\$14,48(%rdi)		# setup number of rounds
1492	xor	%rax,%rax
1493	jmp	.Lexit
1494
1495.Lbadpointer:
1496	mov	\$-1,%rax
1497.Lexit:
1498	.byte	0xf3,0xc3			# rep ret
1499.size	_x86_64_AES_set_encrypt_key,.-_x86_64_AES_set_encrypt_key
1500___
1501
1502sub deckey_ref()
1503{ my ($i,$ptr,$te,$td) = @_;
1504  my ($tp1,$tp2,$tp4,$tp8,$acc)=("%eax","%ebx","%edi","%edx","%r8d");
1505$code.=<<___;
1506	mov	$i($ptr),$tp1
1507	mov	$tp1,$acc
1508	and	\$0x80808080,$acc
1509	mov	$acc,$tp4
1510	shr	\$7,$tp4
1511	lea	0($tp1,$tp1),$tp2
1512	sub	$tp4,$acc
1513	and	\$0xfefefefe,$tp2
1514	and	\$0x1b1b1b1b,$acc
1515	xor	$tp2,$acc
1516	mov	$acc,$tp2
1517
1518	and	\$0x80808080,$acc
1519	mov	$acc,$tp8
1520	shr	\$7,$tp8
1521	lea	0($tp2,$tp2),$tp4
1522	sub	$tp8,$acc
1523	and	\$0xfefefefe,$tp4
1524	and	\$0x1b1b1b1b,$acc
1525	 xor	$tp1,$tp2		# tp2^tp1
1526	xor	$tp4,$acc
1527	mov	$acc,$tp4
1528
1529	and	\$0x80808080,$acc
1530	mov	$acc,$tp8
1531	shr	\$7,$tp8
1532	sub	$tp8,$acc
1533	lea	0($tp4,$tp4),$tp8
1534	 xor	$tp1,$tp4		# tp4^tp1
1535	and	\$0xfefefefe,$tp8
1536	and	\$0x1b1b1b1b,$acc
1537	xor	$acc,$tp8
1538
1539	xor	$tp8,$tp1		# tp1^tp8
1540	rol	\$8,$tp1		# ROTATE(tp1^tp8,8)
1541	xor	$tp8,$tp2		# tp2^tp1^tp8
1542	xor	$tp8,$tp4		# tp4^tp1^tp8
1543	xor	$tp2,$tp8
1544	xor	$tp4,$tp8		# tp8^(tp8^tp4^tp1)^(tp8^tp2^tp1)=tp8^tp4^tp2
1545
1546	xor	$tp8,$tp1
1547	rol	\$24,$tp2		# ROTATE(tp2^tp1^tp8,24)
1548	xor	$tp2,$tp1
1549	rol	\$16,$tp4		# ROTATE(tp4^tp1^tp8,16)
1550	xor	$tp4,$tp1
1551
1552	mov	$tp1,$i($ptr)
1553___
1554}
1555
1556# int private_AES_set_decrypt_key(const unsigned char *userKey, const int bits,
1557#                        AES_KEY *key)
1558$code.=<<___;
1559.globl	private_AES_set_decrypt_key
1560.type	private_AES_set_decrypt_key,\@function,3
1561.align	16
1562private_AES_set_decrypt_key:
1563	push	%rbx
1564	push	%rbp
1565	push	%r12
1566	push	%r13
1567	push	%r14
1568	push	%r15
1569	push	%rdx			# save key schedule
1570.Ldec_key_prologue:
1571
1572	call	_x86_64_AES_set_encrypt_key
1573	mov	(%rsp),%r8		# restore key schedule
1574	cmp	\$0,%eax
1575	jne	.Labort
1576
1577	mov	240(%r8),%r14d		# pull number of rounds
1578	xor	%rdi,%rdi
1579	lea	(%rdi,%r14d,4),%rcx
1580	mov	%r8,%rsi
1581	lea	(%r8,%rcx,4),%rdi	# pointer to last chunk
1582.align	4
1583.Linvert:
1584		mov	0(%rsi),%rax
1585		mov	8(%rsi),%rbx
1586		mov	0(%rdi),%rcx
1587		mov	8(%rdi),%rdx
1588		mov	%rax,0(%rdi)
1589		mov	%rbx,8(%rdi)
1590		mov	%rcx,0(%rsi)
1591		mov	%rdx,8(%rsi)
1592		lea	16(%rsi),%rsi
1593		lea	-16(%rdi),%rdi
1594		cmp	%rsi,%rdi
1595	jne	.Linvert
1596
1597	lea	.LAES_Te+2048+1024(%rip),%rax	# rcon
1598
1599	mov	40(%rax),$mask80
1600	mov	48(%rax),$maskfe
1601	mov	56(%rax),$mask1b
1602
1603	mov	%r8,$key
1604	sub	\$1,%r14d
1605.align	4
1606.Lpermute:
1607		lea	16($key),$key
1608		mov	0($key),%rax
1609		mov	8($key),%rcx
1610___
1611		&dectransform ();
1612$code.=<<___;
1613		mov	%eax,0($key)
1614		mov	%ebx,4($key)
1615		mov	%ecx,8($key)
1616		mov	%edx,12($key)
1617		sub	\$1,%r14d
1618	jnz	.Lpermute
1619
1620	xor	%rax,%rax
1621.Labort:
1622	mov	8(%rsp),%r15
1623	mov	16(%rsp),%r14
1624	mov	24(%rsp),%r13
1625	mov	32(%rsp),%r12
1626	mov	40(%rsp),%rbp
1627	mov	48(%rsp),%rbx
1628	add	\$56,%rsp
1629.Ldec_key_epilogue:
1630	ret
1631.size	private_AES_set_decrypt_key,.-private_AES_set_decrypt_key
1632___
1633
1634# void AES_cbc_encrypt (const void char *inp, unsigned char *out,
1635#			size_t length, const AES_KEY *key,
1636#			unsigned char *ivp,const int enc);
1637{
1638# stack frame layout
1639# -8(%rsp)		return address
1640my $keyp="0(%rsp)";		# one to pass as $key
1641my $keyend="8(%rsp)";		# &(keyp->rd_key[4*keyp->rounds])
1642my $_rsp="16(%rsp)";		# saved %rsp
1643my $_inp="24(%rsp)";		# copy of 1st parameter, inp
1644my $_out="32(%rsp)";		# copy of 2nd parameter, out
1645my $_len="40(%rsp)";		# copy of 3rd parameter, length
1646my $_key="48(%rsp)";		# copy of 4th parameter, key
1647my $_ivp="56(%rsp)";		# copy of 5th parameter, ivp
1648my $ivec="64(%rsp)";		# ivec[16]
1649my $aes_key="80(%rsp)";		# copy of aes_key
1650my $mark="80+240(%rsp)";	# copy of aes_key->rounds
1651
1652$code.=<<___;
1653.globl	AES_cbc_encrypt
1654.type	AES_cbc_encrypt,\@function,6
1655.align	16
1656.extern	OPENSSL_ia32cap_P
1657.globl	asm_AES_cbc_encrypt
1658.hidden	asm_AES_cbc_encrypt
1659asm_AES_cbc_encrypt:
1660AES_cbc_encrypt:
1661	cmp	\$0,%rdx	# check length
1662	je	.Lcbc_epilogue
1663	pushfq
1664	push	%rbx
1665	push	%rbp
1666	push	%r12
1667	push	%r13
1668	push	%r14
1669	push	%r15
1670.Lcbc_prologue:
1671
1672	cld
1673	mov	%r9d,%r9d	# clear upper half of enc
1674
1675	lea	.LAES_Te(%rip),$sbox
1676	cmp	\$0,%r9
1677	jne	.Lcbc_picked_te
1678	lea	.LAES_Td(%rip),$sbox
1679.Lcbc_picked_te:
1680
1681	mov	OPENSSL_ia32cap_P(%rip),%r10d
1682	cmp	\$$speed_limit,%rdx
1683	jb	.Lcbc_slow_prologue
1684	test	\$15,%rdx
1685	jnz	.Lcbc_slow_prologue
1686	bt	\$28,%r10d
1687	jc	.Lcbc_slow_prologue
1688
1689	# allocate aligned stack frame...
1690	lea	-88-248(%rsp),$key
1691	and	\$-64,$key
1692
1693	# ... and make sure it doesn't alias with AES_T[ed] modulo 4096
1694	mov	$sbox,%r10
1695	lea	2304($sbox),%r11
1696	mov	$key,%r12
1697	and	\$0xFFF,%r10	# s = $sbox&0xfff
1698	and	\$0xFFF,%r11	# e = ($sbox+2048)&0xfff
1699	and	\$0xFFF,%r12	# p = %rsp&0xfff
1700
1701	cmp	%r11,%r12	# if (p=>e) %rsp =- (p-e);
1702	jb	.Lcbc_te_break_out
1703	sub	%r11,%r12
1704	sub	%r12,$key
1705	jmp	.Lcbc_te_ok
1706.Lcbc_te_break_out:		# else %rsp -= (p-s)&0xfff + framesz
1707	sub	%r10,%r12
1708	and	\$0xFFF,%r12
1709	add	\$320,%r12
1710	sub	%r12,$key
1711.align	4
1712.Lcbc_te_ok:
1713
1714	xchg	%rsp,$key
1715	#add	\$8,%rsp	# reserve for return address!
1716	mov	$key,$_rsp	# save %rsp
1717.Lcbc_fast_body:
1718	mov	%rdi,$_inp	# save copy of inp
1719	mov	%rsi,$_out	# save copy of out
1720	mov	%rdx,$_len	# save copy of len
1721	mov	%rcx,$_key	# save copy of key
1722	mov	%r8,$_ivp	# save copy of ivp
1723	movl	\$0,$mark	# copy of aes_key->rounds = 0;
1724	mov	%r8,%rbp	# rearrange input arguments
1725	mov	%r9,%rbx
1726	mov	%rsi,$out
1727	mov	%rdi,$inp
1728	mov	%rcx,$key
1729
1730	mov	240($key),%eax		# key->rounds
1731	# do we copy key schedule to stack?
1732	mov	$key,%r10
1733	sub	$sbox,%r10
1734	and	\$0xfff,%r10
1735	cmp	\$2304,%r10
1736	jb	.Lcbc_do_ecopy
1737	cmp	\$4096-248,%r10
1738	jb	.Lcbc_skip_ecopy
1739.align	4
1740.Lcbc_do_ecopy:
1741		mov	$key,%rsi
1742		lea	$aes_key,%rdi
1743		lea	$aes_key,$key
1744		mov	\$240/8,%ecx
1745		.long	0x90A548F3	# rep movsq
1746		mov	%eax,(%rdi)	# copy aes_key->rounds
1747.Lcbc_skip_ecopy:
1748	mov	$key,$keyp	# save key pointer
1749
1750	mov	\$18,%ecx
1751.align	4
1752.Lcbc_prefetch_te:
1753		mov	0($sbox),%r10
1754		mov	32($sbox),%r11
1755		mov	64($sbox),%r12
1756		mov	96($sbox),%r13
1757		lea	128($sbox),$sbox
1758		sub	\$1,%ecx
1759	jnz	.Lcbc_prefetch_te
1760	lea	-2304($sbox),$sbox
1761
1762	cmp	\$0,%rbx
1763	je	.LFAST_DECRYPT
1764
1765#----------------------------- ENCRYPT -----------------------------#
1766	mov	0(%rbp),$s0		# load iv
1767	mov	4(%rbp),$s1
1768	mov	8(%rbp),$s2
1769	mov	12(%rbp),$s3
1770
1771.align	4
1772.Lcbc_fast_enc_loop:
1773		xor	0($inp),$s0
1774		xor	4($inp),$s1
1775		xor	8($inp),$s2
1776		xor	12($inp),$s3
1777		mov	$keyp,$key	# restore key
1778		mov	$inp,$_inp	# if ($verticalspin) save inp
1779
1780		call	_x86_64_AES_encrypt
1781
1782		mov	$_inp,$inp	# if ($verticalspin) restore inp
1783		mov	$_len,%r10
1784		mov	$s0,0($out)
1785		mov	$s1,4($out)
1786		mov	$s2,8($out)
1787		mov	$s3,12($out)
1788
1789		lea	16($inp),$inp
1790		lea	16($out),$out
1791		sub	\$16,%r10
1792		test	\$-16,%r10
1793		mov	%r10,$_len
1794	jnz	.Lcbc_fast_enc_loop
1795	mov	$_ivp,%rbp	# restore ivp
1796	mov	$s0,0(%rbp)	# save ivec
1797	mov	$s1,4(%rbp)
1798	mov	$s2,8(%rbp)
1799	mov	$s3,12(%rbp)
1800
1801	jmp	.Lcbc_fast_cleanup
1802
1803#----------------------------- DECRYPT -----------------------------#
1804.align	16
1805.LFAST_DECRYPT:
1806	cmp	$inp,$out
1807	je	.Lcbc_fast_dec_in_place
1808
1809	mov	%rbp,$ivec
1810.align	4
1811.Lcbc_fast_dec_loop:
1812		mov	0($inp),$s0	# read input
1813		mov	4($inp),$s1
1814		mov	8($inp),$s2
1815		mov	12($inp),$s3
1816		mov	$keyp,$key	# restore key
1817		mov	$inp,$_inp	# if ($verticalspin) save inp
1818
1819		call	_x86_64_AES_decrypt
1820
1821		mov	$ivec,%rbp	# load ivp
1822		mov	$_inp,$inp	# if ($verticalspin) restore inp
1823		mov	$_len,%r10	# load len
1824		xor	0(%rbp),$s0	# xor iv
1825		xor	4(%rbp),$s1
1826		xor	8(%rbp),$s2
1827		xor	12(%rbp),$s3
1828		mov	$inp,%rbp	# current input, next iv
1829
1830		sub	\$16,%r10
1831		mov	%r10,$_len	# update len
1832		mov	%rbp,$ivec	# update ivp
1833
1834		mov	$s0,0($out)	# write output
1835		mov	$s1,4($out)
1836		mov	$s2,8($out)
1837		mov	$s3,12($out)
1838
1839		lea	16($inp),$inp
1840		lea	16($out),$out
1841	jnz	.Lcbc_fast_dec_loop
1842	mov	$_ivp,%r12		# load user ivp
1843	mov	0(%rbp),%r10		# load iv
1844	mov	8(%rbp),%r11
1845	mov	%r10,0(%r12)		# copy back to user
1846	mov	%r11,8(%r12)
1847	jmp	.Lcbc_fast_cleanup
1848
1849.align	16
1850.Lcbc_fast_dec_in_place:
1851	mov	0(%rbp),%r10		# copy iv to stack
1852	mov	8(%rbp),%r11
1853	mov	%r10,0+$ivec
1854	mov	%r11,8+$ivec
1855.align	4
1856.Lcbc_fast_dec_in_place_loop:
1857		mov	0($inp),$s0	# load input
1858		mov	4($inp),$s1
1859		mov	8($inp),$s2
1860		mov	12($inp),$s3
1861		mov	$keyp,$key	# restore key
1862		mov	$inp,$_inp	# if ($verticalspin) save inp
1863
1864		call	_x86_64_AES_decrypt
1865
1866		mov	$_inp,$inp	# if ($verticalspin) restore inp
1867		mov	$_len,%r10
1868		xor	0+$ivec,$s0
1869		xor	4+$ivec,$s1
1870		xor	8+$ivec,$s2
1871		xor	12+$ivec,$s3
1872
1873		mov	0($inp),%r11	# load input
1874		mov	8($inp),%r12
1875		sub	\$16,%r10
1876		jz	.Lcbc_fast_dec_in_place_done
1877
1878		mov	%r11,0+$ivec	# copy input to iv
1879		mov	%r12,8+$ivec
1880
1881		mov	$s0,0($out)	# save output [zaps input]
1882		mov	$s1,4($out)
1883		mov	$s2,8($out)
1884		mov	$s3,12($out)
1885
1886		lea	16($inp),$inp
1887		lea	16($out),$out
1888		mov	%r10,$_len
1889	jmp	.Lcbc_fast_dec_in_place_loop
1890.Lcbc_fast_dec_in_place_done:
1891	mov	$_ivp,%rdi
1892	mov	%r11,0(%rdi)	# copy iv back to user
1893	mov	%r12,8(%rdi)
1894
1895	mov	$s0,0($out)	# save output [zaps input]
1896	mov	$s1,4($out)
1897	mov	$s2,8($out)
1898	mov	$s3,12($out)
1899
1900.align	4
1901.Lcbc_fast_cleanup:
1902	cmpl	\$0,$mark	# was the key schedule copied?
1903	lea	$aes_key,%rdi
1904	je	.Lcbc_exit
1905		mov	\$240/8,%ecx
1906		xor	%rax,%rax
1907		.long	0x90AB48F3	# rep stosq
1908
1909	jmp	.Lcbc_exit
1910
1911#--------------------------- SLOW ROUTINE ---------------------------#
1912.align	16
1913.Lcbc_slow_prologue:
1914	# allocate aligned stack frame...
1915	lea	-88(%rsp),%rbp
1916	and	\$-64,%rbp
1917	# ... just "above" key schedule
1918	lea	-88-63(%rcx),%r10
1919	sub	%rbp,%r10
1920	neg	%r10
1921	and	\$0x3c0,%r10
1922	sub	%r10,%rbp
1923
1924	xchg	%rsp,%rbp
1925	#add	\$8,%rsp	# reserve for return address!
1926	mov	%rbp,$_rsp	# save %rsp
1927.Lcbc_slow_body:
1928	#mov	%rdi,$_inp	# save copy of inp
1929	#mov	%rsi,$_out	# save copy of out
1930	#mov	%rdx,$_len	# save copy of len
1931	#mov	%rcx,$_key	# save copy of key
1932	mov	%r8,$_ivp	# save copy of ivp
1933	mov	%r8,%rbp	# rearrange input arguments
1934	mov	%r9,%rbx
1935	mov	%rsi,$out
1936	mov	%rdi,$inp
1937	mov	%rcx,$key
1938	mov	%rdx,%r10
1939
1940	mov	240($key),%eax
1941	mov	$key,$keyp	# save key pointer
1942	shl	\$4,%eax
1943	lea	($key,%rax),%rax
1944	mov	%rax,$keyend
1945
1946	# pick Te4 copy which can't "overlap" with stack frame or key scdedule
1947	lea	2048($sbox),$sbox
1948	lea	768-8(%rsp),%rax
1949	sub	$sbox,%rax
1950	and	\$0x300,%rax
1951	lea	($sbox,%rax),$sbox
1952
1953	cmp	\$0,%rbx
1954	je	.LSLOW_DECRYPT
1955
1956#--------------------------- SLOW ENCRYPT ---------------------------#
1957	test	\$-16,%r10		# check upon length
1958	mov	0(%rbp),$s0		# load iv
1959	mov	4(%rbp),$s1
1960	mov	8(%rbp),$s2
1961	mov	12(%rbp),$s3
1962	jz	.Lcbc_slow_enc_tail	# short input...
1963
1964.align	4
1965.Lcbc_slow_enc_loop:
1966		xor	0($inp),$s0
1967		xor	4($inp),$s1
1968		xor	8($inp),$s2
1969		xor	12($inp),$s3
1970		mov	$keyp,$key	# restore key
1971		mov	$inp,$_inp	# save inp
1972		mov	$out,$_out	# save out
1973		mov	%r10,$_len	# save len
1974
1975		call	_x86_64_AES_encrypt_compact
1976
1977		mov	$_inp,$inp	# restore inp
1978		mov	$_out,$out	# restore out
1979		mov	$_len,%r10	# restore len
1980		mov	$s0,0($out)
1981		mov	$s1,4($out)
1982		mov	$s2,8($out)
1983		mov	$s3,12($out)
1984
1985		lea	16($inp),$inp
1986		lea	16($out),$out
1987		sub	\$16,%r10
1988		test	\$-16,%r10
1989	jnz	.Lcbc_slow_enc_loop
1990	test	\$15,%r10
1991	jnz	.Lcbc_slow_enc_tail
1992	mov	$_ivp,%rbp	# restore ivp
1993	mov	$s0,0(%rbp)	# save ivec
1994	mov	$s1,4(%rbp)
1995	mov	$s2,8(%rbp)
1996	mov	$s3,12(%rbp)
1997
1998	jmp	.Lcbc_exit
1999
2000.align	4
2001.Lcbc_slow_enc_tail:
2002	mov	%rax,%r11
2003	mov	%rcx,%r12
2004	mov	%r10,%rcx
2005	mov	$inp,%rsi
2006	mov	$out,%rdi
2007	.long	0x9066A4F3		# rep movsb
2008	mov	\$16,%rcx		# zero tail
2009	sub	%r10,%rcx
2010	xor	%rax,%rax
2011	.long	0x9066AAF3		# rep stosb
2012	mov	$out,$inp		# this is not a mistake!
2013	mov	\$16,%r10		# len=16
2014	mov	%r11,%rax
2015	mov	%r12,%rcx
2016	jmp	.Lcbc_slow_enc_loop	# one more spin...
2017#--------------------------- SLOW DECRYPT ---------------------------#
2018.align	16
2019.LSLOW_DECRYPT:
2020	shr	\$3,%rax
2021	add	%rax,$sbox		# recall "magic" constants!
2022
2023	mov	0(%rbp),%r11		# copy iv to stack
2024	mov	8(%rbp),%r12
2025	mov	%r11,0+$ivec
2026	mov	%r12,8+$ivec
2027
2028.align	4
2029.Lcbc_slow_dec_loop:
2030		mov	0($inp),$s0	# load input
2031		mov	4($inp),$s1
2032		mov	8($inp),$s2
2033		mov	12($inp),$s3
2034		mov	$keyp,$key	# restore key
2035		mov	$inp,$_inp	# save inp
2036		mov	$out,$_out	# save out
2037		mov	%r10,$_len	# save len
2038
2039		call	_x86_64_AES_decrypt_compact
2040
2041		mov	$_inp,$inp	# restore inp
2042		mov	$_out,$out	# restore out
2043		mov	$_len,%r10
2044		xor	0+$ivec,$s0
2045		xor	4+$ivec,$s1
2046		xor	8+$ivec,$s2
2047		xor	12+$ivec,$s3
2048
2049		mov	0($inp),%r11	# load input
2050		mov	8($inp),%r12
2051		sub	\$16,%r10
2052		jc	.Lcbc_slow_dec_partial
2053		jz	.Lcbc_slow_dec_done
2054
2055		mov	%r11,0+$ivec	# copy input to iv
2056		mov	%r12,8+$ivec
2057
2058		mov	$s0,0($out)	# save output [can zap input]
2059		mov	$s1,4($out)
2060		mov	$s2,8($out)
2061		mov	$s3,12($out)
2062
2063		lea	16($inp),$inp
2064		lea	16($out),$out
2065	jmp	.Lcbc_slow_dec_loop
2066.Lcbc_slow_dec_done:
2067	mov	$_ivp,%rdi
2068	mov	%r11,0(%rdi)		# copy iv back to user
2069	mov	%r12,8(%rdi)
2070
2071	mov	$s0,0($out)		# save output [can zap input]
2072	mov	$s1,4($out)
2073	mov	$s2,8($out)
2074	mov	$s3,12($out)
2075
2076	jmp	.Lcbc_exit
2077
2078.align	4
2079.Lcbc_slow_dec_partial:
2080	mov	$_ivp,%rdi
2081	mov	%r11,0(%rdi)		# copy iv back to user
2082	mov	%r12,8(%rdi)
2083
2084	mov	$s0,0+$ivec		# save output to stack
2085	mov	$s1,4+$ivec
2086	mov	$s2,8+$ivec
2087	mov	$s3,12+$ivec
2088
2089	mov	$out,%rdi
2090	lea	$ivec,%rsi
2091	lea	16(%r10),%rcx
2092	.long	0x9066A4F3	# rep movsb
2093	jmp	.Lcbc_exit
2094
2095.align	16
2096.Lcbc_exit:
2097	mov	$_rsp,%rsi
2098	mov	(%rsi),%r15
2099	mov	8(%rsi),%r14
2100	mov	16(%rsi),%r13
2101	mov	24(%rsi),%r12
2102	mov	32(%rsi),%rbp
2103	mov	40(%rsi),%rbx
2104	lea	48(%rsi),%rsp
2105.Lcbc_popfq:
2106	popfq
2107.Lcbc_epilogue:
2108	ret
2109.size	AES_cbc_encrypt,.-AES_cbc_encrypt
2110___
2111}
2112
2113$code.=<<___;
2114.align	64
2115.LAES_Te:
2116___
2117	&_data_word(0xa56363c6, 0x847c7cf8, 0x997777ee, 0x8d7b7bf6);
2118	&_data_word(0x0df2f2ff, 0xbd6b6bd6, 0xb16f6fde, 0x54c5c591);
2119	&_data_word(0x50303060, 0x03010102, 0xa96767ce, 0x7d2b2b56);
2120	&_data_word(0x19fefee7, 0x62d7d7b5, 0xe6abab4d, 0x9a7676ec);
2121	&_data_word(0x45caca8f, 0x9d82821f, 0x40c9c989, 0x877d7dfa);
2122	&_data_word(0x15fafaef, 0xeb5959b2, 0xc947478e, 0x0bf0f0fb);
2123	&_data_word(0xecadad41, 0x67d4d4b3, 0xfda2a25f, 0xeaafaf45);
2124	&_data_word(0xbf9c9c23, 0xf7a4a453, 0x967272e4, 0x5bc0c09b);
2125	&_data_word(0xc2b7b775, 0x1cfdfde1, 0xae93933d, 0x6a26264c);
2126	&_data_word(0x5a36366c, 0x413f3f7e, 0x02f7f7f5, 0x4fcccc83);
2127	&_data_word(0x5c343468, 0xf4a5a551, 0x34e5e5d1, 0x08f1f1f9);
2128	&_data_word(0x937171e2, 0x73d8d8ab, 0x53313162, 0x3f15152a);
2129	&_data_word(0x0c040408, 0x52c7c795, 0x65232346, 0x5ec3c39d);
2130	&_data_word(0x28181830, 0xa1969637, 0x0f05050a, 0xb59a9a2f);
2131	&_data_word(0x0907070e, 0x36121224, 0x9b80801b, 0x3de2e2df);
2132	&_data_word(0x26ebebcd, 0x6927274e, 0xcdb2b27f, 0x9f7575ea);
2133	&_data_word(0x1b090912, 0x9e83831d, 0x742c2c58, 0x2e1a1a34);
2134	&_data_word(0x2d1b1b36, 0xb26e6edc, 0xee5a5ab4, 0xfba0a05b);
2135	&_data_word(0xf65252a4, 0x4d3b3b76, 0x61d6d6b7, 0xceb3b37d);
2136	&_data_word(0x7b292952, 0x3ee3e3dd, 0x712f2f5e, 0x97848413);
2137	&_data_word(0xf55353a6, 0x68d1d1b9, 0x00000000, 0x2cededc1);
2138	&_data_word(0x60202040, 0x1ffcfce3, 0xc8b1b179, 0xed5b5bb6);
2139	&_data_word(0xbe6a6ad4, 0x46cbcb8d, 0xd9bebe67, 0x4b393972);
2140	&_data_word(0xde4a4a94, 0xd44c4c98, 0xe85858b0, 0x4acfcf85);
2141	&_data_word(0x6bd0d0bb, 0x2aefefc5, 0xe5aaaa4f, 0x16fbfbed);
2142	&_data_word(0xc5434386, 0xd74d4d9a, 0x55333366, 0x94858511);
2143	&_data_word(0xcf45458a, 0x10f9f9e9, 0x06020204, 0x817f7ffe);
2144	&_data_word(0xf05050a0, 0x443c3c78, 0xba9f9f25, 0xe3a8a84b);
2145	&_data_word(0xf35151a2, 0xfea3a35d, 0xc0404080, 0x8a8f8f05);
2146	&_data_word(0xad92923f, 0xbc9d9d21, 0x48383870, 0x04f5f5f1);
2147	&_data_word(0xdfbcbc63, 0xc1b6b677, 0x75dadaaf, 0x63212142);
2148	&_data_word(0x30101020, 0x1affffe5, 0x0ef3f3fd, 0x6dd2d2bf);
2149	&_data_word(0x4ccdcd81, 0x140c0c18, 0x35131326, 0x2fececc3);
2150	&_data_word(0xe15f5fbe, 0xa2979735, 0xcc444488, 0x3917172e);
2151	&_data_word(0x57c4c493, 0xf2a7a755, 0x827e7efc, 0x473d3d7a);
2152	&_data_word(0xac6464c8, 0xe75d5dba, 0x2b191932, 0x957373e6);
2153	&_data_word(0xa06060c0, 0x98818119, 0xd14f4f9e, 0x7fdcdca3);
2154	&_data_word(0x66222244, 0x7e2a2a54, 0xab90903b, 0x8388880b);
2155	&_data_word(0xca46468c, 0x29eeeec7, 0xd3b8b86b, 0x3c141428);
2156	&_data_word(0x79dedea7, 0xe25e5ebc, 0x1d0b0b16, 0x76dbdbad);
2157	&_data_word(0x3be0e0db, 0x56323264, 0x4e3a3a74, 0x1e0a0a14);
2158	&_data_word(0xdb494992, 0x0a06060c, 0x6c242448, 0xe45c5cb8);
2159	&_data_word(0x5dc2c29f, 0x6ed3d3bd, 0xefacac43, 0xa66262c4);
2160	&_data_word(0xa8919139, 0xa4959531, 0x37e4e4d3, 0x8b7979f2);
2161	&_data_word(0x32e7e7d5, 0x43c8c88b, 0x5937376e, 0xb76d6dda);
2162	&_data_word(0x8c8d8d01, 0x64d5d5b1, 0xd24e4e9c, 0xe0a9a949);
2163	&_data_word(0xb46c6cd8, 0xfa5656ac, 0x07f4f4f3, 0x25eaeacf);
2164	&_data_word(0xaf6565ca, 0x8e7a7af4, 0xe9aeae47, 0x18080810);
2165	&_data_word(0xd5baba6f, 0x887878f0, 0x6f25254a, 0x722e2e5c);
2166	&_data_word(0x241c1c38, 0xf1a6a657, 0xc7b4b473, 0x51c6c697);
2167	&_data_word(0x23e8e8cb, 0x7cdddda1, 0x9c7474e8, 0x211f1f3e);
2168	&_data_word(0xdd4b4b96, 0xdcbdbd61, 0x868b8b0d, 0x858a8a0f);
2169	&_data_word(0x907070e0, 0x423e3e7c, 0xc4b5b571, 0xaa6666cc);
2170	&_data_word(0xd8484890, 0x05030306, 0x01f6f6f7, 0x120e0e1c);
2171	&_data_word(0xa36161c2, 0x5f35356a, 0xf95757ae, 0xd0b9b969);
2172	&_data_word(0x91868617, 0x58c1c199, 0x271d1d3a, 0xb99e9e27);
2173	&_data_word(0x38e1e1d9, 0x13f8f8eb, 0xb398982b, 0x33111122);
2174	&_data_word(0xbb6969d2, 0x70d9d9a9, 0x898e8e07, 0xa7949433);
2175	&_data_word(0xb69b9b2d, 0x221e1e3c, 0x92878715, 0x20e9e9c9);
2176	&_data_word(0x49cece87, 0xff5555aa, 0x78282850, 0x7adfdfa5);
2177	&_data_word(0x8f8c8c03, 0xf8a1a159, 0x80898909, 0x170d0d1a);
2178	&_data_word(0xdabfbf65, 0x31e6e6d7, 0xc6424284, 0xb86868d0);
2179	&_data_word(0xc3414182, 0xb0999929, 0x772d2d5a, 0x110f0f1e);
2180	&_data_word(0xcbb0b07b, 0xfc5454a8, 0xd6bbbb6d, 0x3a16162c);
2181
2182#Te4	# four copies of Te4 to choose from to avoid L1 aliasing
2183	&data_byte(0x63, 0x7c, 0x77, 0x7b, 0xf2, 0x6b, 0x6f, 0xc5);
2184	&data_byte(0x30, 0x01, 0x67, 0x2b, 0xfe, 0xd7, 0xab, 0x76);
2185	&data_byte(0xca, 0x82, 0xc9, 0x7d, 0xfa, 0x59, 0x47, 0xf0);
2186	&data_byte(0xad, 0xd4, 0xa2, 0xaf, 0x9c, 0xa4, 0x72, 0xc0);
2187	&data_byte(0xb7, 0xfd, 0x93, 0x26, 0x36, 0x3f, 0xf7, 0xcc);
2188	&data_byte(0x34, 0xa5, 0xe5, 0xf1, 0x71, 0xd8, 0x31, 0x15);
2189	&data_byte(0x04, 0xc7, 0x23, 0xc3, 0x18, 0x96, 0x05, 0x9a);
2190	&data_byte(0x07, 0x12, 0x80, 0xe2, 0xeb, 0x27, 0xb2, 0x75);
2191	&data_byte(0x09, 0x83, 0x2c, 0x1a, 0x1b, 0x6e, 0x5a, 0xa0);
2192	&data_byte(0x52, 0x3b, 0xd6, 0xb3, 0x29, 0xe3, 0x2f, 0x84);
2193	&data_byte(0x53, 0xd1, 0x00, 0xed, 0x20, 0xfc, 0xb1, 0x5b);
2194	&data_byte(0x6a, 0xcb, 0xbe, 0x39, 0x4a, 0x4c, 0x58, 0xcf);
2195	&data_byte(0xd0, 0xef, 0xaa, 0xfb, 0x43, 0x4d, 0x33, 0x85);
2196	&data_byte(0x45, 0xf9, 0x02, 0x7f, 0x50, 0x3c, 0x9f, 0xa8);
2197	&data_byte(0x51, 0xa3, 0x40, 0x8f, 0x92, 0x9d, 0x38, 0xf5);
2198	&data_byte(0xbc, 0xb6, 0xda, 0x21, 0x10, 0xff, 0xf3, 0xd2);
2199	&data_byte(0xcd, 0x0c, 0x13, 0xec, 0x5f, 0x97, 0x44, 0x17);
2200	&data_byte(0xc4, 0xa7, 0x7e, 0x3d, 0x64, 0x5d, 0x19, 0x73);
2201	&data_byte(0x60, 0x81, 0x4f, 0xdc, 0x22, 0x2a, 0x90, 0x88);
2202	&data_byte(0x46, 0xee, 0xb8, 0x14, 0xde, 0x5e, 0x0b, 0xdb);
2203	&data_byte(0xe0, 0x32, 0x3a, 0x0a, 0x49, 0x06, 0x24, 0x5c);
2204	&data_byte(0xc2, 0xd3, 0xac, 0x62, 0x91, 0x95, 0xe4, 0x79);
2205	&data_byte(0xe7, 0xc8, 0x37, 0x6d, 0x8d, 0xd5, 0x4e, 0xa9);
2206	&data_byte(0x6c, 0x56, 0xf4, 0xea, 0x65, 0x7a, 0xae, 0x08);
2207	&data_byte(0xba, 0x78, 0x25, 0x2e, 0x1c, 0xa6, 0xb4, 0xc6);
2208	&data_byte(0xe8, 0xdd, 0x74, 0x1f, 0x4b, 0xbd, 0x8b, 0x8a);
2209	&data_byte(0x70, 0x3e, 0xb5, 0x66, 0x48, 0x03, 0xf6, 0x0e);
2210	&data_byte(0x61, 0x35, 0x57, 0xb9, 0x86, 0xc1, 0x1d, 0x9e);
2211	&data_byte(0xe1, 0xf8, 0x98, 0x11, 0x69, 0xd9, 0x8e, 0x94);
2212	&data_byte(0x9b, 0x1e, 0x87, 0xe9, 0xce, 0x55, 0x28, 0xdf);
2213	&data_byte(0x8c, 0xa1, 0x89, 0x0d, 0xbf, 0xe6, 0x42, 0x68);
2214	&data_byte(0x41, 0x99, 0x2d, 0x0f, 0xb0, 0x54, 0xbb, 0x16);
2215
2216	&data_byte(0x63, 0x7c, 0x77, 0x7b, 0xf2, 0x6b, 0x6f, 0xc5);
2217	&data_byte(0x30, 0x01, 0x67, 0x2b, 0xfe, 0xd7, 0xab, 0x76);
2218	&data_byte(0xca, 0x82, 0xc9, 0x7d, 0xfa, 0x59, 0x47, 0xf0);
2219	&data_byte(0xad, 0xd4, 0xa2, 0xaf, 0x9c, 0xa4, 0x72, 0xc0);
2220	&data_byte(0xb7, 0xfd, 0x93, 0x26, 0x36, 0x3f, 0xf7, 0xcc);
2221	&data_byte(0x34, 0xa5, 0xe5, 0xf1, 0x71, 0xd8, 0x31, 0x15);
2222	&data_byte(0x04, 0xc7, 0x23, 0xc3, 0x18, 0x96, 0x05, 0x9a);
2223	&data_byte(0x07, 0x12, 0x80, 0xe2, 0xeb, 0x27, 0xb2, 0x75);
2224	&data_byte(0x09, 0x83, 0x2c, 0x1a, 0x1b, 0x6e, 0x5a, 0xa0);
2225	&data_byte(0x52, 0x3b, 0xd6, 0xb3, 0x29, 0xe3, 0x2f, 0x84);
2226	&data_byte(0x53, 0xd1, 0x00, 0xed, 0x20, 0xfc, 0xb1, 0x5b);
2227	&data_byte(0x6a, 0xcb, 0xbe, 0x39, 0x4a, 0x4c, 0x58, 0xcf);
2228	&data_byte(0xd0, 0xef, 0xaa, 0xfb, 0x43, 0x4d, 0x33, 0x85);
2229	&data_byte(0x45, 0xf9, 0x02, 0x7f, 0x50, 0x3c, 0x9f, 0xa8);
2230	&data_byte(0x51, 0xa3, 0x40, 0x8f, 0x92, 0x9d, 0x38, 0xf5);
2231	&data_byte(0xbc, 0xb6, 0xda, 0x21, 0x10, 0xff, 0xf3, 0xd2);
2232	&data_byte(0xcd, 0x0c, 0x13, 0xec, 0x5f, 0x97, 0x44, 0x17);
2233	&data_byte(0xc4, 0xa7, 0x7e, 0x3d, 0x64, 0x5d, 0x19, 0x73);
2234	&data_byte(0x60, 0x81, 0x4f, 0xdc, 0x22, 0x2a, 0x90, 0x88);
2235	&data_byte(0x46, 0xee, 0xb8, 0x14, 0xde, 0x5e, 0x0b, 0xdb);
2236	&data_byte(0xe0, 0x32, 0x3a, 0x0a, 0x49, 0x06, 0x24, 0x5c);
2237	&data_byte(0xc2, 0xd3, 0xac, 0x62, 0x91, 0x95, 0xe4, 0x79);
2238	&data_byte(0xe7, 0xc8, 0x37, 0x6d, 0x8d, 0xd5, 0x4e, 0xa9);
2239	&data_byte(0x6c, 0x56, 0xf4, 0xea, 0x65, 0x7a, 0xae, 0x08);
2240	&data_byte(0xba, 0x78, 0x25, 0x2e, 0x1c, 0xa6, 0xb4, 0xc6);
2241	&data_byte(0xe8, 0xdd, 0x74, 0x1f, 0x4b, 0xbd, 0x8b, 0x8a);
2242	&data_byte(0x70, 0x3e, 0xb5, 0x66, 0x48, 0x03, 0xf6, 0x0e);
2243	&data_byte(0x61, 0x35, 0x57, 0xb9, 0x86, 0xc1, 0x1d, 0x9e);
2244	&data_byte(0xe1, 0xf8, 0x98, 0x11, 0x69, 0xd9, 0x8e, 0x94);
2245	&data_byte(0x9b, 0x1e, 0x87, 0xe9, 0xce, 0x55, 0x28, 0xdf);
2246	&data_byte(0x8c, 0xa1, 0x89, 0x0d, 0xbf, 0xe6, 0x42, 0x68);
2247	&data_byte(0x41, 0x99, 0x2d, 0x0f, 0xb0, 0x54, 0xbb, 0x16);
2248
2249	&data_byte(0x63, 0x7c, 0x77, 0x7b, 0xf2, 0x6b, 0x6f, 0xc5);
2250	&data_byte(0x30, 0x01, 0x67, 0x2b, 0xfe, 0xd7, 0xab, 0x76);
2251	&data_byte(0xca, 0x82, 0xc9, 0x7d, 0xfa, 0x59, 0x47, 0xf0);
2252	&data_byte(0xad, 0xd4, 0xa2, 0xaf, 0x9c, 0xa4, 0x72, 0xc0);
2253	&data_byte(0xb7, 0xfd, 0x93, 0x26, 0x36, 0x3f, 0xf7, 0xcc);
2254	&data_byte(0x34, 0xa5, 0xe5, 0xf1, 0x71, 0xd8, 0x31, 0x15);
2255	&data_byte(0x04, 0xc7, 0x23, 0xc3, 0x18, 0x96, 0x05, 0x9a);
2256	&data_byte(0x07, 0x12, 0x80, 0xe2, 0xeb, 0x27, 0xb2, 0x75);
2257	&data_byte(0x09, 0x83, 0x2c, 0x1a, 0x1b, 0x6e, 0x5a, 0xa0);
2258	&data_byte(0x52, 0x3b, 0xd6, 0xb3, 0x29, 0xe3, 0x2f, 0x84);
2259	&data_byte(0x53, 0xd1, 0x00, 0xed, 0x20, 0xfc, 0xb1, 0x5b);
2260	&data_byte(0x6a, 0xcb, 0xbe, 0x39, 0x4a, 0x4c, 0x58, 0xcf);
2261	&data_byte(0xd0, 0xef, 0xaa, 0xfb, 0x43, 0x4d, 0x33, 0x85);
2262	&data_byte(0x45, 0xf9, 0x02, 0x7f, 0x50, 0x3c, 0x9f, 0xa8);
2263	&data_byte(0x51, 0xa3, 0x40, 0x8f, 0x92, 0x9d, 0x38, 0xf5);
2264	&data_byte(0xbc, 0xb6, 0xda, 0x21, 0x10, 0xff, 0xf3, 0xd2);
2265	&data_byte(0xcd, 0x0c, 0x13, 0xec, 0x5f, 0x97, 0x44, 0x17);
2266	&data_byte(0xc4, 0xa7, 0x7e, 0x3d, 0x64, 0x5d, 0x19, 0x73);
2267	&data_byte(0x60, 0x81, 0x4f, 0xdc, 0x22, 0x2a, 0x90, 0x88);
2268	&data_byte(0x46, 0xee, 0xb8, 0x14, 0xde, 0x5e, 0x0b, 0xdb);
2269	&data_byte(0xe0, 0x32, 0x3a, 0x0a, 0x49, 0x06, 0x24, 0x5c);
2270	&data_byte(0xc2, 0xd3, 0xac, 0x62, 0x91, 0x95, 0xe4, 0x79);
2271	&data_byte(0xe7, 0xc8, 0x37, 0x6d, 0x8d, 0xd5, 0x4e, 0xa9);
2272	&data_byte(0x6c, 0x56, 0xf4, 0xea, 0x65, 0x7a, 0xae, 0x08);
2273	&data_byte(0xba, 0x78, 0x25, 0x2e, 0x1c, 0xa6, 0xb4, 0xc6);
2274	&data_byte(0xe8, 0xdd, 0x74, 0x1f, 0x4b, 0xbd, 0x8b, 0x8a);
2275	&data_byte(0x70, 0x3e, 0xb5, 0x66, 0x48, 0x03, 0xf6, 0x0e);
2276	&data_byte(0x61, 0x35, 0x57, 0xb9, 0x86, 0xc1, 0x1d, 0x9e);
2277	&data_byte(0xe1, 0xf8, 0x98, 0x11, 0x69, 0xd9, 0x8e, 0x94);
2278	&data_byte(0x9b, 0x1e, 0x87, 0xe9, 0xce, 0x55, 0x28, 0xdf);
2279	&data_byte(0x8c, 0xa1, 0x89, 0x0d, 0xbf, 0xe6, 0x42, 0x68);
2280	&data_byte(0x41, 0x99, 0x2d, 0x0f, 0xb0, 0x54, 0xbb, 0x16);
2281
2282	&data_byte(0x63, 0x7c, 0x77, 0x7b, 0xf2, 0x6b, 0x6f, 0xc5);
2283	&data_byte(0x30, 0x01, 0x67, 0x2b, 0xfe, 0xd7, 0xab, 0x76);
2284	&data_byte(0xca, 0x82, 0xc9, 0x7d, 0xfa, 0x59, 0x47, 0xf0);
2285	&data_byte(0xad, 0xd4, 0xa2, 0xaf, 0x9c, 0xa4, 0x72, 0xc0);
2286	&data_byte(0xb7, 0xfd, 0x93, 0x26, 0x36, 0x3f, 0xf7, 0xcc);
2287	&data_byte(0x34, 0xa5, 0xe5, 0xf1, 0x71, 0xd8, 0x31, 0x15);
2288	&data_byte(0x04, 0xc7, 0x23, 0xc3, 0x18, 0x96, 0x05, 0x9a);
2289	&data_byte(0x07, 0x12, 0x80, 0xe2, 0xeb, 0x27, 0xb2, 0x75);
2290	&data_byte(0x09, 0x83, 0x2c, 0x1a, 0x1b, 0x6e, 0x5a, 0xa0);
2291	&data_byte(0x52, 0x3b, 0xd6, 0xb3, 0x29, 0xe3, 0x2f, 0x84);
2292	&data_byte(0x53, 0xd1, 0x00, 0xed, 0x20, 0xfc, 0xb1, 0x5b);
2293	&data_byte(0x6a, 0xcb, 0xbe, 0x39, 0x4a, 0x4c, 0x58, 0xcf);
2294	&data_byte(0xd0, 0xef, 0xaa, 0xfb, 0x43, 0x4d, 0x33, 0x85);
2295	&data_byte(0x45, 0xf9, 0x02, 0x7f, 0x50, 0x3c, 0x9f, 0xa8);
2296	&data_byte(0x51, 0xa3, 0x40, 0x8f, 0x92, 0x9d, 0x38, 0xf5);
2297	&data_byte(0xbc, 0xb6, 0xda, 0x21, 0x10, 0xff, 0xf3, 0xd2);
2298	&data_byte(0xcd, 0x0c, 0x13, 0xec, 0x5f, 0x97, 0x44, 0x17);
2299	&data_byte(0xc4, 0xa7, 0x7e, 0x3d, 0x64, 0x5d, 0x19, 0x73);
2300	&data_byte(0x60, 0x81, 0x4f, 0xdc, 0x22, 0x2a, 0x90, 0x88);
2301	&data_byte(0x46, 0xee, 0xb8, 0x14, 0xde, 0x5e, 0x0b, 0xdb);
2302	&data_byte(0xe0, 0x32, 0x3a, 0x0a, 0x49, 0x06, 0x24, 0x5c);
2303	&data_byte(0xc2, 0xd3, 0xac, 0x62, 0x91, 0x95, 0xe4, 0x79);
2304	&data_byte(0xe7, 0xc8, 0x37, 0x6d, 0x8d, 0xd5, 0x4e, 0xa9);
2305	&data_byte(0x6c, 0x56, 0xf4, 0xea, 0x65, 0x7a, 0xae, 0x08);
2306	&data_byte(0xba, 0x78, 0x25, 0x2e, 0x1c, 0xa6, 0xb4, 0xc6);
2307	&data_byte(0xe8, 0xdd, 0x74, 0x1f, 0x4b, 0xbd, 0x8b, 0x8a);
2308	&data_byte(0x70, 0x3e, 0xb5, 0x66, 0x48, 0x03, 0xf6, 0x0e);
2309	&data_byte(0x61, 0x35, 0x57, 0xb9, 0x86, 0xc1, 0x1d, 0x9e);
2310	&data_byte(0xe1, 0xf8, 0x98, 0x11, 0x69, 0xd9, 0x8e, 0x94);
2311	&data_byte(0x9b, 0x1e, 0x87, 0xe9, 0xce, 0x55, 0x28, 0xdf);
2312	&data_byte(0x8c, 0xa1, 0x89, 0x0d, 0xbf, 0xe6, 0x42, 0x68);
2313	&data_byte(0x41, 0x99, 0x2d, 0x0f, 0xb0, 0x54, 0xbb, 0x16);
2314#rcon:
2315$code.=<<___;
2316	.long	0x00000001, 0x00000002, 0x00000004, 0x00000008
2317	.long	0x00000010, 0x00000020, 0x00000040, 0x00000080
2318	.long	0x0000001b, 0x00000036, 0x80808080, 0x80808080
2319	.long	0xfefefefe, 0xfefefefe, 0x1b1b1b1b, 0x1b1b1b1b
2320___
2321$code.=<<___;
2322.align	64
2323.LAES_Td:
2324___
2325	&_data_word(0x50a7f451, 0x5365417e, 0xc3a4171a, 0x965e273a);
2326	&_data_word(0xcb6bab3b, 0xf1459d1f, 0xab58faac, 0x9303e34b);
2327	&_data_word(0x55fa3020, 0xf66d76ad, 0x9176cc88, 0x254c02f5);
2328	&_data_word(0xfcd7e54f, 0xd7cb2ac5, 0x80443526, 0x8fa362b5);
2329	&_data_word(0x495ab1de, 0x671bba25, 0x980eea45, 0xe1c0fe5d);
2330	&_data_word(0x02752fc3, 0x12f04c81, 0xa397468d, 0xc6f9d36b);
2331	&_data_word(0xe75f8f03, 0x959c9215, 0xeb7a6dbf, 0xda595295);
2332	&_data_word(0x2d83bed4, 0xd3217458, 0x2969e049, 0x44c8c98e);
2333	&_data_word(0x6a89c275, 0x78798ef4, 0x6b3e5899, 0xdd71b927);
2334	&_data_word(0xb64fe1be, 0x17ad88f0, 0x66ac20c9, 0xb43ace7d);
2335	&_data_word(0x184adf63, 0x82311ae5, 0x60335197, 0x457f5362);
2336	&_data_word(0xe07764b1, 0x84ae6bbb, 0x1ca081fe, 0x942b08f9);
2337	&_data_word(0x58684870, 0x19fd458f, 0x876cde94, 0xb7f87b52);
2338	&_data_word(0x23d373ab, 0xe2024b72, 0x578f1fe3, 0x2aab5566);
2339	&_data_word(0x0728ebb2, 0x03c2b52f, 0x9a7bc586, 0xa50837d3);
2340	&_data_word(0xf2872830, 0xb2a5bf23, 0xba6a0302, 0x5c8216ed);
2341	&_data_word(0x2b1ccf8a, 0x92b479a7, 0xf0f207f3, 0xa1e2694e);
2342	&_data_word(0xcdf4da65, 0xd5be0506, 0x1f6234d1, 0x8afea6c4);
2343	&_data_word(0x9d532e34, 0xa055f3a2, 0x32e18a05, 0x75ebf6a4);
2344	&_data_word(0x39ec830b, 0xaaef6040, 0x069f715e, 0x51106ebd);
2345	&_data_word(0xf98a213e, 0x3d06dd96, 0xae053edd, 0x46bde64d);
2346	&_data_word(0xb58d5491, 0x055dc471, 0x6fd40604, 0xff155060);
2347	&_data_word(0x24fb9819, 0x97e9bdd6, 0xcc434089, 0x779ed967);
2348	&_data_word(0xbd42e8b0, 0x888b8907, 0x385b19e7, 0xdbeec879);
2349	&_data_word(0x470a7ca1, 0xe90f427c, 0xc91e84f8, 0x00000000);
2350	&_data_word(0x83868009, 0x48ed2b32, 0xac70111e, 0x4e725a6c);
2351	&_data_word(0xfbff0efd, 0x5638850f, 0x1ed5ae3d, 0x27392d36);
2352	&_data_word(0x64d90f0a, 0x21a65c68, 0xd1545b9b, 0x3a2e3624);
2353	&_data_word(0xb1670a0c, 0x0fe75793, 0xd296eeb4, 0x9e919b1b);
2354	&_data_word(0x4fc5c080, 0xa220dc61, 0x694b775a, 0x161a121c);
2355	&_data_word(0x0aba93e2, 0xe52aa0c0, 0x43e0223c, 0x1d171b12);
2356	&_data_word(0x0b0d090e, 0xadc78bf2, 0xb9a8b62d, 0xc8a91e14);
2357	&_data_word(0x8519f157, 0x4c0775af, 0xbbdd99ee, 0xfd607fa3);
2358	&_data_word(0x9f2601f7, 0xbcf5725c, 0xc53b6644, 0x347efb5b);
2359	&_data_word(0x7629438b, 0xdcc623cb, 0x68fcedb6, 0x63f1e4b8);
2360	&_data_word(0xcadc31d7, 0x10856342, 0x40229713, 0x2011c684);
2361	&_data_word(0x7d244a85, 0xf83dbbd2, 0x1132f9ae, 0x6da129c7);
2362	&_data_word(0x4b2f9e1d, 0xf330b2dc, 0xec52860d, 0xd0e3c177);
2363	&_data_word(0x6c16b32b, 0x99b970a9, 0xfa489411, 0x2264e947);
2364	&_data_word(0xc48cfca8, 0x1a3ff0a0, 0xd82c7d56, 0xef903322);
2365	&_data_word(0xc74e4987, 0xc1d138d9, 0xfea2ca8c, 0x360bd498);
2366	&_data_word(0xcf81f5a6, 0x28de7aa5, 0x268eb7da, 0xa4bfad3f);
2367	&_data_word(0xe49d3a2c, 0x0d927850, 0x9bcc5f6a, 0x62467e54);
2368	&_data_word(0xc2138df6, 0xe8b8d890, 0x5ef7392e, 0xf5afc382);
2369	&_data_word(0xbe805d9f, 0x7c93d069, 0xa92dd56f, 0xb31225cf);
2370	&_data_word(0x3b99acc8, 0xa77d1810, 0x6e639ce8, 0x7bbb3bdb);
2371	&_data_word(0x097826cd, 0xf418596e, 0x01b79aec, 0xa89a4f83);
2372	&_data_word(0x656e95e6, 0x7ee6ffaa, 0x08cfbc21, 0xe6e815ef);
2373	&_data_word(0xd99be7ba, 0xce366f4a, 0xd4099fea, 0xd67cb029);
2374	&_data_word(0xafb2a431, 0x31233f2a, 0x3094a5c6, 0xc066a235);
2375	&_data_word(0x37bc4e74, 0xa6ca82fc, 0xb0d090e0, 0x15d8a733);
2376	&_data_word(0x4a9804f1, 0xf7daec41, 0x0e50cd7f, 0x2ff69117);
2377	&_data_word(0x8dd64d76, 0x4db0ef43, 0x544daacc, 0xdf0496e4);
2378	&_data_word(0xe3b5d19e, 0x1b886a4c, 0xb81f2cc1, 0x7f516546);
2379	&_data_word(0x04ea5e9d, 0x5d358c01, 0x737487fa, 0x2e410bfb);
2380	&_data_word(0x5a1d67b3, 0x52d2db92, 0x335610e9, 0x1347d66d);
2381	&_data_word(0x8c61d79a, 0x7a0ca137, 0x8e14f859, 0x893c13eb);
2382	&_data_word(0xee27a9ce, 0x35c961b7, 0xede51ce1, 0x3cb1477a);
2383	&_data_word(0x59dfd29c, 0x3f73f255, 0x79ce1418, 0xbf37c773);
2384	&_data_word(0xeacdf753, 0x5baafd5f, 0x146f3ddf, 0x86db4478);
2385	&_data_word(0x81f3afca, 0x3ec468b9, 0x2c342438, 0x5f40a3c2);
2386	&_data_word(0x72c31d16, 0x0c25e2bc, 0x8b493c28, 0x41950dff);
2387	&_data_word(0x7101a839, 0xdeb30c08, 0x9ce4b4d8, 0x90c15664);
2388	&_data_word(0x6184cb7b, 0x70b632d5, 0x745c6c48, 0x4257b8d0);
2389
2390#Td4:	# four copies of Td4 to choose from to avoid L1 aliasing
2391	&data_byte(0x52, 0x09, 0x6a, 0xd5, 0x30, 0x36, 0xa5, 0x38);
2392	&data_byte(0xbf, 0x40, 0xa3, 0x9e, 0x81, 0xf3, 0xd7, 0xfb);
2393	&data_byte(0x7c, 0xe3, 0x39, 0x82, 0x9b, 0x2f, 0xff, 0x87);
2394	&data_byte(0x34, 0x8e, 0x43, 0x44, 0xc4, 0xde, 0xe9, 0xcb);
2395	&data_byte(0x54, 0x7b, 0x94, 0x32, 0xa6, 0xc2, 0x23, 0x3d);
2396	&data_byte(0xee, 0x4c, 0x95, 0x0b, 0x42, 0xfa, 0xc3, 0x4e);
2397	&data_byte(0x08, 0x2e, 0xa1, 0x66, 0x28, 0xd9, 0x24, 0xb2);
2398	&data_byte(0x76, 0x5b, 0xa2, 0x49, 0x6d, 0x8b, 0xd1, 0x25);
2399	&data_byte(0x72, 0xf8, 0xf6, 0x64, 0x86, 0x68, 0x98, 0x16);
2400	&data_byte(0xd4, 0xa4, 0x5c, 0xcc, 0x5d, 0x65, 0xb6, 0x92);
2401	&data_byte(0x6c, 0x70, 0x48, 0x50, 0xfd, 0xed, 0xb9, 0xda);
2402	&data_byte(0x5e, 0x15, 0x46, 0x57, 0xa7, 0x8d, 0x9d, 0x84);
2403	&data_byte(0x90, 0xd8, 0xab, 0x00, 0x8c, 0xbc, 0xd3, 0x0a);
2404	&data_byte(0xf7, 0xe4, 0x58, 0x05, 0xb8, 0xb3, 0x45, 0x06);
2405	&data_byte(0xd0, 0x2c, 0x1e, 0x8f, 0xca, 0x3f, 0x0f, 0x02);
2406	&data_byte(0xc1, 0xaf, 0xbd, 0x03, 0x01, 0x13, 0x8a, 0x6b);
2407	&data_byte(0x3a, 0x91, 0x11, 0x41, 0x4f, 0x67, 0xdc, 0xea);
2408	&data_byte(0x97, 0xf2, 0xcf, 0xce, 0xf0, 0xb4, 0xe6, 0x73);
2409	&data_byte(0x96, 0xac, 0x74, 0x22, 0xe7, 0xad, 0x35, 0x85);
2410	&data_byte(0xe2, 0xf9, 0x37, 0xe8, 0x1c, 0x75, 0xdf, 0x6e);
2411	&data_byte(0x47, 0xf1, 0x1a, 0x71, 0x1d, 0x29, 0xc5, 0x89);
2412	&data_byte(0x6f, 0xb7, 0x62, 0x0e, 0xaa, 0x18, 0xbe, 0x1b);
2413	&data_byte(0xfc, 0x56, 0x3e, 0x4b, 0xc6, 0xd2, 0x79, 0x20);
2414	&data_byte(0x9a, 0xdb, 0xc0, 0xfe, 0x78, 0xcd, 0x5a, 0xf4);
2415	&data_byte(0x1f, 0xdd, 0xa8, 0x33, 0x88, 0x07, 0xc7, 0x31);
2416	&data_byte(0xb1, 0x12, 0x10, 0x59, 0x27, 0x80, 0xec, 0x5f);
2417	&data_byte(0x60, 0x51, 0x7f, 0xa9, 0x19, 0xb5, 0x4a, 0x0d);
2418	&data_byte(0x2d, 0xe5, 0x7a, 0x9f, 0x93, 0xc9, 0x9c, 0xef);
2419	&data_byte(0xa0, 0xe0, 0x3b, 0x4d, 0xae, 0x2a, 0xf5, 0xb0);
2420	&data_byte(0xc8, 0xeb, 0xbb, 0x3c, 0x83, 0x53, 0x99, 0x61);
2421	&data_byte(0x17, 0x2b, 0x04, 0x7e, 0xba, 0x77, 0xd6, 0x26);
2422	&data_byte(0xe1, 0x69, 0x14, 0x63, 0x55, 0x21, 0x0c, 0x7d);
2423$code.=<<___;
2424	.long	0x80808080, 0x80808080, 0xfefefefe, 0xfefefefe
2425	.long	0x1b1b1b1b, 0x1b1b1b1b, 0, 0
2426___
2427	&data_byte(0x52, 0x09, 0x6a, 0xd5, 0x30, 0x36, 0xa5, 0x38);
2428	&data_byte(0xbf, 0x40, 0xa3, 0x9e, 0x81, 0xf3, 0xd7, 0xfb);
2429	&data_byte(0x7c, 0xe3, 0x39, 0x82, 0x9b, 0x2f, 0xff, 0x87);
2430	&data_byte(0x34, 0x8e, 0x43, 0x44, 0xc4, 0xde, 0xe9, 0xcb);
2431	&data_byte(0x54, 0x7b, 0x94, 0x32, 0xa6, 0xc2, 0x23, 0x3d);
2432	&data_byte(0xee, 0x4c, 0x95, 0x0b, 0x42, 0xfa, 0xc3, 0x4e);
2433	&data_byte(0x08, 0x2e, 0xa1, 0x66, 0x28, 0xd9, 0x24, 0xb2);
2434	&data_byte(0x76, 0x5b, 0xa2, 0x49, 0x6d, 0x8b, 0xd1, 0x25);
2435	&data_byte(0x72, 0xf8, 0xf6, 0x64, 0x86, 0x68, 0x98, 0x16);
2436	&data_byte(0xd4, 0xa4, 0x5c, 0xcc, 0x5d, 0x65, 0xb6, 0x92);
2437	&data_byte(0x6c, 0x70, 0x48, 0x50, 0xfd, 0xed, 0xb9, 0xda);
2438	&data_byte(0x5e, 0x15, 0x46, 0x57, 0xa7, 0x8d, 0x9d, 0x84);
2439	&data_byte(0x90, 0xd8, 0xab, 0x00, 0x8c, 0xbc, 0xd3, 0x0a);
2440	&data_byte(0xf7, 0xe4, 0x58, 0x05, 0xb8, 0xb3, 0x45, 0x06);
2441	&data_byte(0xd0, 0x2c, 0x1e, 0x8f, 0xca, 0x3f, 0x0f, 0x02);
2442	&data_byte(0xc1, 0xaf, 0xbd, 0x03, 0x01, 0x13, 0x8a, 0x6b);
2443	&data_byte(0x3a, 0x91, 0x11, 0x41, 0x4f, 0x67, 0xdc, 0xea);
2444	&data_byte(0x97, 0xf2, 0xcf, 0xce, 0xf0, 0xb4, 0xe6, 0x73);
2445	&data_byte(0x96, 0xac, 0x74, 0x22, 0xe7, 0xad, 0x35, 0x85);
2446	&data_byte(0xe2, 0xf9, 0x37, 0xe8, 0x1c, 0x75, 0xdf, 0x6e);
2447	&data_byte(0x47, 0xf1, 0x1a, 0x71, 0x1d, 0x29, 0xc5, 0x89);
2448	&data_byte(0x6f, 0xb7, 0x62, 0x0e, 0xaa, 0x18, 0xbe, 0x1b);
2449	&data_byte(0xfc, 0x56, 0x3e, 0x4b, 0xc6, 0xd2, 0x79, 0x20);
2450	&data_byte(0x9a, 0xdb, 0xc0, 0xfe, 0x78, 0xcd, 0x5a, 0xf4);
2451	&data_byte(0x1f, 0xdd, 0xa8, 0x33, 0x88, 0x07, 0xc7, 0x31);
2452	&data_byte(0xb1, 0x12, 0x10, 0x59, 0x27, 0x80, 0xec, 0x5f);
2453	&data_byte(0x60, 0x51, 0x7f, 0xa9, 0x19, 0xb5, 0x4a, 0x0d);
2454	&data_byte(0x2d, 0xe5, 0x7a, 0x9f, 0x93, 0xc9, 0x9c, 0xef);
2455	&data_byte(0xa0, 0xe0, 0x3b, 0x4d, 0xae, 0x2a, 0xf5, 0xb0);
2456	&data_byte(0xc8, 0xeb, 0xbb, 0x3c, 0x83, 0x53, 0x99, 0x61);
2457	&data_byte(0x17, 0x2b, 0x04, 0x7e, 0xba, 0x77, 0xd6, 0x26);
2458	&data_byte(0xe1, 0x69, 0x14, 0x63, 0x55, 0x21, 0x0c, 0x7d);
2459$code.=<<___;
2460	.long	0x80808080, 0x80808080, 0xfefefefe, 0xfefefefe
2461	.long	0x1b1b1b1b, 0x1b1b1b1b, 0, 0
2462___
2463	&data_byte(0x52, 0x09, 0x6a, 0xd5, 0x30, 0x36, 0xa5, 0x38);
2464	&data_byte(0xbf, 0x40, 0xa3, 0x9e, 0x81, 0xf3, 0xd7, 0xfb);
2465	&data_byte(0x7c, 0xe3, 0x39, 0x82, 0x9b, 0x2f, 0xff, 0x87);
2466	&data_byte(0x34, 0x8e, 0x43, 0x44, 0xc4, 0xde, 0xe9, 0xcb);
2467	&data_byte(0x54, 0x7b, 0x94, 0x32, 0xa6, 0xc2, 0x23, 0x3d);
2468	&data_byte(0xee, 0x4c, 0x95, 0x0b, 0x42, 0xfa, 0xc3, 0x4e);
2469	&data_byte(0x08, 0x2e, 0xa1, 0x66, 0x28, 0xd9, 0x24, 0xb2);
2470	&data_byte(0x76, 0x5b, 0xa2, 0x49, 0x6d, 0x8b, 0xd1, 0x25);
2471	&data_byte(0x72, 0xf8, 0xf6, 0x64, 0x86, 0x68, 0x98, 0x16);
2472	&data_byte(0xd4, 0xa4, 0x5c, 0xcc, 0x5d, 0x65, 0xb6, 0x92);
2473	&data_byte(0x6c, 0x70, 0x48, 0x50, 0xfd, 0xed, 0xb9, 0xda);
2474	&data_byte(0x5e, 0x15, 0x46, 0x57, 0xa7, 0x8d, 0x9d, 0x84);
2475	&data_byte(0x90, 0xd8, 0xab, 0x00, 0x8c, 0xbc, 0xd3, 0x0a);
2476	&data_byte(0xf7, 0xe4, 0x58, 0x05, 0xb8, 0xb3, 0x45, 0x06);
2477	&data_byte(0xd0, 0x2c, 0x1e, 0x8f, 0xca, 0x3f, 0x0f, 0x02);
2478	&data_byte(0xc1, 0xaf, 0xbd, 0x03, 0x01, 0x13, 0x8a, 0x6b);
2479	&data_byte(0x3a, 0x91, 0x11, 0x41, 0x4f, 0x67, 0xdc, 0xea);
2480	&data_byte(0x97, 0xf2, 0xcf, 0xce, 0xf0, 0xb4, 0xe6, 0x73);
2481	&data_byte(0x96, 0xac, 0x74, 0x22, 0xe7, 0xad, 0x35, 0x85);
2482	&data_byte(0xe2, 0xf9, 0x37, 0xe8, 0x1c, 0x75, 0xdf, 0x6e);
2483	&data_byte(0x47, 0xf1, 0x1a, 0x71, 0x1d, 0x29, 0xc5, 0x89);
2484	&data_byte(0x6f, 0xb7, 0x62, 0x0e, 0xaa, 0x18, 0xbe, 0x1b);
2485	&data_byte(0xfc, 0x56, 0x3e, 0x4b, 0xc6, 0xd2, 0x79, 0x20);
2486	&data_byte(0x9a, 0xdb, 0xc0, 0xfe, 0x78, 0xcd, 0x5a, 0xf4);
2487	&data_byte(0x1f, 0xdd, 0xa8, 0x33, 0x88, 0x07, 0xc7, 0x31);
2488	&data_byte(0xb1, 0x12, 0x10, 0x59, 0x27, 0x80, 0xec, 0x5f);
2489	&data_byte(0x60, 0x51, 0x7f, 0xa9, 0x19, 0xb5, 0x4a, 0x0d);
2490	&data_byte(0x2d, 0xe5, 0x7a, 0x9f, 0x93, 0xc9, 0x9c, 0xef);
2491	&data_byte(0xa0, 0xe0, 0x3b, 0x4d, 0xae, 0x2a, 0xf5, 0xb0);
2492	&data_byte(0xc8, 0xeb, 0xbb, 0x3c, 0x83, 0x53, 0x99, 0x61);
2493	&data_byte(0x17, 0x2b, 0x04, 0x7e, 0xba, 0x77, 0xd6, 0x26);
2494	&data_byte(0xe1, 0x69, 0x14, 0x63, 0x55, 0x21, 0x0c, 0x7d);
2495$code.=<<___;
2496	.long	0x80808080, 0x80808080, 0xfefefefe, 0xfefefefe
2497	.long	0x1b1b1b1b, 0x1b1b1b1b, 0, 0
2498___
2499	&data_byte(0x52, 0x09, 0x6a, 0xd5, 0x30, 0x36, 0xa5, 0x38);
2500	&data_byte(0xbf, 0x40, 0xa3, 0x9e, 0x81, 0xf3, 0xd7, 0xfb);
2501	&data_byte(0x7c, 0xe3, 0x39, 0x82, 0x9b, 0x2f, 0xff, 0x87);
2502	&data_byte(0x34, 0x8e, 0x43, 0x44, 0xc4, 0xde, 0xe9, 0xcb);
2503	&data_byte(0x54, 0x7b, 0x94, 0x32, 0xa6, 0xc2, 0x23, 0x3d);
2504	&data_byte(0xee, 0x4c, 0x95, 0x0b, 0x42, 0xfa, 0xc3, 0x4e);
2505	&data_byte(0x08, 0x2e, 0xa1, 0x66, 0x28, 0xd9, 0x24, 0xb2);
2506	&data_byte(0x76, 0x5b, 0xa2, 0x49, 0x6d, 0x8b, 0xd1, 0x25);
2507	&data_byte(0x72, 0xf8, 0xf6, 0x64, 0x86, 0x68, 0x98, 0x16);
2508	&data_byte(0xd4, 0xa4, 0x5c, 0xcc, 0x5d, 0x65, 0xb6, 0x92);
2509	&data_byte(0x6c, 0x70, 0x48, 0x50, 0xfd, 0xed, 0xb9, 0xda);
2510	&data_byte(0x5e, 0x15, 0x46, 0x57, 0xa7, 0x8d, 0x9d, 0x84);
2511	&data_byte(0x90, 0xd8, 0xab, 0x00, 0x8c, 0xbc, 0xd3, 0x0a);
2512	&data_byte(0xf7, 0xe4, 0x58, 0x05, 0xb8, 0xb3, 0x45, 0x06);
2513	&data_byte(0xd0, 0x2c, 0x1e, 0x8f, 0xca, 0x3f, 0x0f, 0x02);
2514	&data_byte(0xc1, 0xaf, 0xbd, 0x03, 0x01, 0x13, 0x8a, 0x6b);
2515	&data_byte(0x3a, 0x91, 0x11, 0x41, 0x4f, 0x67, 0xdc, 0xea);
2516	&data_byte(0x97, 0xf2, 0xcf, 0xce, 0xf0, 0xb4, 0xe6, 0x73);
2517	&data_byte(0x96, 0xac, 0x74, 0x22, 0xe7, 0xad, 0x35, 0x85);
2518	&data_byte(0xe2, 0xf9, 0x37, 0xe8, 0x1c, 0x75, 0xdf, 0x6e);
2519	&data_byte(0x47, 0xf1, 0x1a, 0x71, 0x1d, 0x29, 0xc5, 0x89);
2520	&data_byte(0x6f, 0xb7, 0x62, 0x0e, 0xaa, 0x18, 0xbe, 0x1b);
2521	&data_byte(0xfc, 0x56, 0x3e, 0x4b, 0xc6, 0xd2, 0x79, 0x20);
2522	&data_byte(0x9a, 0xdb, 0xc0, 0xfe, 0x78, 0xcd, 0x5a, 0xf4);
2523	&data_byte(0x1f, 0xdd, 0xa8, 0x33, 0x88, 0x07, 0xc7, 0x31);
2524	&data_byte(0xb1, 0x12, 0x10, 0x59, 0x27, 0x80, 0xec, 0x5f);
2525	&data_byte(0x60, 0x51, 0x7f, 0xa9, 0x19, 0xb5, 0x4a, 0x0d);
2526	&data_byte(0x2d, 0xe5, 0x7a, 0x9f, 0x93, 0xc9, 0x9c, 0xef);
2527	&data_byte(0xa0, 0xe0, 0x3b, 0x4d, 0xae, 0x2a, 0xf5, 0xb0);
2528	&data_byte(0xc8, 0xeb, 0xbb, 0x3c, 0x83, 0x53, 0x99, 0x61);
2529	&data_byte(0x17, 0x2b, 0x04, 0x7e, 0xba, 0x77, 0xd6, 0x26);
2530	&data_byte(0xe1, 0x69, 0x14, 0x63, 0x55, 0x21, 0x0c, 0x7d);
2531$code.=<<___;
2532	.long	0x80808080, 0x80808080, 0xfefefefe, 0xfefefefe
2533	.long	0x1b1b1b1b, 0x1b1b1b1b, 0, 0
2534.asciz  "AES for x86_64, CRYPTOGAMS by <appro\@openssl.org>"
2535.align	64
2536___
2537
2538# EXCEPTION_DISPOSITION handler (EXCEPTION_RECORD *rec,ULONG64 frame,
2539#		CONTEXT *context,DISPATCHER_CONTEXT *disp)
2540if ($win64) {
2541$rec="%rcx";
2542$frame="%rdx";
2543$context="%r8";
2544$disp="%r9";
2545
2546$code.=<<___;
2547.extern	__imp_RtlVirtualUnwind
2548.type	block_se_handler,\@abi-omnipotent
2549.align	16
2550block_se_handler:
2551	push	%rsi
2552	push	%rdi
2553	push	%rbx
2554	push	%rbp
2555	push	%r12
2556	push	%r13
2557	push	%r14
2558	push	%r15
2559	pushfq
2560	sub	\$64,%rsp
2561
2562	mov	120($context),%rax	# pull context->Rax
2563	mov	248($context),%rbx	# pull context->Rip
2564
2565	mov	8($disp),%rsi		# disp->ImageBase
2566	mov	56($disp),%r11		# disp->HandlerData
2567
2568	mov	0(%r11),%r10d		# HandlerData[0]
2569	lea	(%rsi,%r10),%r10	# prologue label
2570	cmp	%r10,%rbx		# context->Rip<prologue label
2571	jb	.Lin_block_prologue
2572
2573	mov	152($context),%rax	# pull context->Rsp
2574
2575	mov	4(%r11),%r10d		# HandlerData[1]
2576	lea	(%rsi,%r10),%r10	# epilogue label
2577	cmp	%r10,%rbx		# context->Rip>=epilogue label
2578	jae	.Lin_block_prologue
2579
2580	mov	24(%rax),%rax		# pull saved real stack pointer
2581	lea	48(%rax),%rax		# adjust...
2582
2583	mov	-8(%rax),%rbx
2584	mov	-16(%rax),%rbp
2585	mov	-24(%rax),%r12
2586	mov	-32(%rax),%r13
2587	mov	-40(%rax),%r14
2588	mov	-48(%rax),%r15
2589	mov	%rbx,144($context)	# restore context->Rbx
2590	mov	%rbp,160($context)	# restore context->Rbp
2591	mov	%r12,216($context)	# restore context->R12
2592	mov	%r13,224($context)	# restore context->R13
2593	mov	%r14,232($context)	# restore context->R14
2594	mov	%r15,240($context)	# restore context->R15
2595
2596.Lin_block_prologue:
2597	mov	8(%rax),%rdi
2598	mov	16(%rax),%rsi
2599	mov	%rax,152($context)	# restore context->Rsp
2600	mov	%rsi,168($context)	# restore context->Rsi
2601	mov	%rdi,176($context)	# restore context->Rdi
2602
2603	jmp	.Lcommon_seh_exit
2604.size	block_se_handler,.-block_se_handler
2605
2606.type	key_se_handler,\@abi-omnipotent
2607.align	16
2608key_se_handler:
2609	push	%rsi
2610	push	%rdi
2611	push	%rbx
2612	push	%rbp
2613	push	%r12
2614	push	%r13
2615	push	%r14
2616	push	%r15
2617	pushfq
2618	sub	\$64,%rsp
2619
2620	mov	120($context),%rax	# pull context->Rax
2621	mov	248($context),%rbx	# pull context->Rip
2622
2623	mov	8($disp),%rsi		# disp->ImageBase
2624	mov	56($disp),%r11		# disp->HandlerData
2625
2626	mov	0(%r11),%r10d		# HandlerData[0]
2627	lea	(%rsi,%r10),%r10	# prologue label
2628	cmp	%r10,%rbx		# context->Rip<prologue label
2629	jb	.Lin_key_prologue
2630
2631	mov	152($context),%rax	# pull context->Rsp
2632
2633	mov	4(%r11),%r10d		# HandlerData[1]
2634	lea	(%rsi,%r10),%r10	# epilogue label
2635	cmp	%r10,%rbx		# context->Rip>=epilogue label
2636	jae	.Lin_key_prologue
2637
2638	lea	56(%rax),%rax
2639
2640	mov	-8(%rax),%rbx
2641	mov	-16(%rax),%rbp
2642	mov	-24(%rax),%r12
2643	mov	-32(%rax),%r13
2644	mov	-40(%rax),%r14
2645	mov	-48(%rax),%r15
2646	mov	%rbx,144($context)	# restore context->Rbx
2647	mov	%rbp,160($context)	# restore context->Rbp
2648	mov	%r12,216($context)	# restore context->R12
2649	mov	%r13,224($context)	# restore context->R13
2650	mov	%r14,232($context)	# restore context->R14
2651	mov	%r15,240($context)	# restore context->R15
2652
2653.Lin_key_prologue:
2654	mov	8(%rax),%rdi
2655	mov	16(%rax),%rsi
2656	mov	%rax,152($context)	# restore context->Rsp
2657	mov	%rsi,168($context)	# restore context->Rsi
2658	mov	%rdi,176($context)	# restore context->Rdi
2659
2660	jmp	.Lcommon_seh_exit
2661.size	key_se_handler,.-key_se_handler
2662
2663.type	cbc_se_handler,\@abi-omnipotent
2664.align	16
2665cbc_se_handler:
2666	push	%rsi
2667	push	%rdi
2668	push	%rbx
2669	push	%rbp
2670	push	%r12
2671	push	%r13
2672	push	%r14
2673	push	%r15
2674	pushfq
2675	sub	\$64,%rsp
2676
2677	mov	120($context),%rax	# pull context->Rax
2678	mov	248($context),%rbx	# pull context->Rip
2679
2680	lea	.Lcbc_prologue(%rip),%r10
2681	cmp	%r10,%rbx		# context->Rip<.Lcbc_prologue
2682	jb	.Lin_cbc_prologue
2683
2684	lea	.Lcbc_fast_body(%rip),%r10
2685	cmp	%r10,%rbx		# context->Rip<.Lcbc_fast_body
2686	jb	.Lin_cbc_frame_setup
2687
2688	lea	.Lcbc_slow_prologue(%rip),%r10
2689	cmp	%r10,%rbx		# context->Rip<.Lcbc_slow_prologue
2690	jb	.Lin_cbc_body
2691
2692	lea	.Lcbc_slow_body(%rip),%r10
2693	cmp	%r10,%rbx		# context->Rip<.Lcbc_slow_body
2694	jb	.Lin_cbc_frame_setup
2695
2696.Lin_cbc_body:
2697	mov	152($context),%rax	# pull context->Rsp
2698
2699	lea	.Lcbc_epilogue(%rip),%r10
2700	cmp	%r10,%rbx		# context->Rip>=.Lcbc_epilogue
2701	jae	.Lin_cbc_prologue
2702
2703	lea	8(%rax),%rax
2704
2705	lea	.Lcbc_popfq(%rip),%r10
2706	cmp	%r10,%rbx		# context->Rip>=.Lcbc_popfq
2707	jae	.Lin_cbc_prologue
2708
2709	mov	`16-8`(%rax),%rax	# biased $_rsp
2710	lea	56(%rax),%rax
2711
2712.Lin_cbc_frame_setup:
2713	mov	-16(%rax),%rbx
2714	mov	-24(%rax),%rbp
2715	mov	-32(%rax),%r12
2716	mov	-40(%rax),%r13
2717	mov	-48(%rax),%r14
2718	mov	-56(%rax),%r15
2719	mov	%rbx,144($context)	# restore context->Rbx
2720	mov	%rbp,160($context)	# restore context->Rbp
2721	mov	%r12,216($context)	# restore context->R12
2722	mov	%r13,224($context)	# restore context->R13
2723	mov	%r14,232($context)	# restore context->R14
2724	mov	%r15,240($context)	# restore context->R15
2725
2726.Lin_cbc_prologue:
2727	mov	8(%rax),%rdi
2728	mov	16(%rax),%rsi
2729	mov	%rax,152($context)	# restore context->Rsp
2730	mov	%rsi,168($context)	# restore context->Rsi
2731	mov	%rdi,176($context)	# restore context->Rdi
2732
2733.Lcommon_seh_exit:
2734
2735	mov	40($disp),%rdi		# disp->ContextRecord
2736	mov	$context,%rsi		# context
2737	mov	\$`1232/8`,%ecx		# sizeof(CONTEXT)
2738	.long	0xa548f3fc		# cld; rep movsq
2739
2740	mov	$disp,%rsi
2741	xor	%rcx,%rcx		# arg1, UNW_FLAG_NHANDLER
2742	mov	8(%rsi),%rdx		# arg2, disp->ImageBase
2743	mov	0(%rsi),%r8		# arg3, disp->ControlPc
2744	mov	16(%rsi),%r9		# arg4, disp->FunctionEntry
2745	mov	40(%rsi),%r10		# disp->ContextRecord
2746	lea	56(%rsi),%r11		# &disp->HandlerData
2747	lea	24(%rsi),%r12		# &disp->EstablisherFrame
2748	mov	%r10,32(%rsp)		# arg5
2749	mov	%r11,40(%rsp)		# arg6
2750	mov	%r12,48(%rsp)		# arg7
2751	mov	%rcx,56(%rsp)		# arg8, (NULL)
2752	call	*__imp_RtlVirtualUnwind(%rip)
2753
2754	mov	\$1,%eax		# ExceptionContinueSearch
2755	add	\$64,%rsp
2756	popfq
2757	pop	%r15
2758	pop	%r14
2759	pop	%r13
2760	pop	%r12
2761	pop	%rbp
2762	pop	%rbx
2763	pop	%rdi
2764	pop	%rsi
2765	ret
2766.size	cbc_se_handler,.-cbc_se_handler
2767
2768.section	.pdata
2769.align	4
2770	.rva	.LSEH_begin_AES_encrypt
2771	.rva	.LSEH_end_AES_encrypt
2772	.rva	.LSEH_info_AES_encrypt
2773
2774	.rva	.LSEH_begin_AES_decrypt
2775	.rva	.LSEH_end_AES_decrypt
2776	.rva	.LSEH_info_AES_decrypt
2777
2778	.rva	.LSEH_begin_private_AES_set_encrypt_key
2779	.rva	.LSEH_end_private_AES_set_encrypt_key
2780	.rva	.LSEH_info_private_AES_set_encrypt_key
2781
2782	.rva	.LSEH_begin_private_AES_set_decrypt_key
2783	.rva	.LSEH_end_private_AES_set_decrypt_key
2784	.rva	.LSEH_info_private_AES_set_decrypt_key
2785
2786	.rva	.LSEH_begin_AES_cbc_encrypt
2787	.rva	.LSEH_end_AES_cbc_encrypt
2788	.rva	.LSEH_info_AES_cbc_encrypt
2789
2790.section	.xdata
2791.align	8
2792.LSEH_info_AES_encrypt:
2793	.byte	9,0,0,0
2794	.rva	block_se_handler
2795	.rva	.Lenc_prologue,.Lenc_epilogue	# HandlerData[]
2796.LSEH_info_AES_decrypt:
2797	.byte	9,0,0,0
2798	.rva	block_se_handler
2799	.rva	.Ldec_prologue,.Ldec_epilogue	# HandlerData[]
2800.LSEH_info_private_AES_set_encrypt_key:
2801	.byte	9,0,0,0
2802	.rva	key_se_handler
2803	.rva	.Lenc_key_prologue,.Lenc_key_epilogue	# HandlerData[]
2804.LSEH_info_private_AES_set_decrypt_key:
2805	.byte	9,0,0,0
2806	.rva	key_se_handler
2807	.rva	.Ldec_key_prologue,.Ldec_key_epilogue	# HandlerData[]
2808.LSEH_info_AES_cbc_encrypt:
2809	.byte	9,0,0,0
2810	.rva	cbc_se_handler
2811___
2812}
2813
2814$code =~ s/\`([^\`]*)\`/eval($1)/gem;
2815
2816print $code;
2817
2818close STDOUT;
2819