xref: /openssl/crypto/sha/asm/sha1-parisc.pl (revision 33388b44) 
1#! /usr/bin/env perl
2# Copyright 2009-2020 The OpenSSL Project Authors. All Rights Reserved.
3#
4# Licensed under the Apache License 2.0 (the "License").  You may not use
5# this file except in compliance with the License.  You can obtain a copy
6# in the file LICENSE in the source distribution or at
7# https://www.openssl.org/source/license.html
8
9
10# ====================================================================
11# Written by Andy Polyakov <appro@openssl.org> for the OpenSSL
12# project. The module is, however, dual licensed under OpenSSL and
13# CRYPTOGAMS licenses depending on where you obtain it. For further
14# details see http://www.openssl.org/~appro/cryptogams/.
15# ====================================================================
16
17# SHA1 block procedure for PA-RISC.
18
19# June 2009.
20#
21# On PA-7100LC performance is >30% better than gcc 3.2 generated code
22# for aligned input and >50% better for unaligned. Compared to vendor
23# compiler on PA-8600 it's almost 60% faster in 64-bit build and just
24# few percent faster in 32-bit one (this for aligned input, data for
25# unaligned input is not available).
26#
27# Special thanks to polarhome.com for providing HP-UX account.
28
29# $output is the last argument if it looks like a file (it has an extension)
30# $flavour is the first argument if it doesn't look like a file
31$output = $#ARGV >= 0 && $ARGV[$#ARGV] =~ m|\.\w+$| ? pop : undef;
32$flavour = $#ARGV >= 0 && $ARGV[0] !~ m|\.| ? shift : undef;
33
34$output and open STDOUT,">$output";
35
36if ($flavour =~ /64/) {
37	$LEVEL		="2.0W";
38	$SIZE_T		=8;
39	$FRAME_MARKER	=80;
40	$SAVED_RP	=16;
41	$PUSH		="std";
42	$PUSHMA		="std,ma";
43	$POP		="ldd";
44	$POPMB		="ldd,mb";
45} else {
46	$LEVEL		="1.0";
47	$SIZE_T		=4;
48	$FRAME_MARKER	=48;
49	$SAVED_RP	=20;
50	$PUSH		="stw";
51	$PUSHMA		="stwm";
52	$POP		="ldw";
53	$POPMB		="ldwm";
54}
55
56$FRAME=14*$SIZE_T+$FRAME_MARKER;# 14 saved regs + frame marker
57				#                 [+ argument transfer]
58$ctx="%r26";		# arg0
59$inp="%r25";		# arg1
60$num="%r24";		# arg2
61
62$t0="%r28";
63$t1="%r29";
64$K="%r31";
65
66@X=("%r1", "%r2", "%r3", "%r4", "%r5", "%r6", "%r7", "%r8",
67    "%r9", "%r10","%r11","%r12","%r13","%r14","%r15","%r16",$t0);
68
69@V=($A,$B,$C,$D,$E)=("%r19","%r20","%r21","%r22","%r23");
70
71sub BODY_00_19 {
72my ($i,$a,$b,$c,$d,$e)=@_;
73my $j=$i+1;
74$code.=<<___ if ($i<15);
75	addl	$K,$e,$e	; $i
76	shd	$a,$a,27,$t1
77	addl	@X[$i],$e,$e
78	and	$c,$b,$t0
79	addl	$t1,$e,$e
80	andcm	$d,$b,$t1
81	shd	$b,$b,2,$b
82	or	$t1,$t0,$t0
83	addl	$t0,$e,$e
84___
85$code.=<<___ if ($i>=15);	# with forward Xupdate
86	addl	$K,$e,$e	; $i
87	shd	$a,$a,27,$t1
88	xor	@X[($j+2)%16],@X[$j%16],@X[$j%16]
89	addl	@X[$i%16],$e,$e
90	and	$c,$b,$t0
91	xor	@X[($j+8)%16],@X[$j%16],@X[$j%16]
92	addl	$t1,$e,$e
93	andcm	$d,$b,$t1
94	shd	$b,$b,2,$b
95	or	$t1,$t0,$t0
96	xor	@X[($j+13)%16],@X[$j%16],@X[$j%16]
97	add	$t0,$e,$e
98	shd	@X[$j%16],@X[$j%16],31,@X[$j%16]
99___
100}
101
102sub BODY_20_39 {
103my ($i,$a,$b,$c,$d,$e)=@_;
104my $j=$i+1;
105$code.=<<___ if ($i<79);
106	xor	@X[($j+2)%16],@X[$j%16],@X[$j%16]	; $i
107	addl	$K,$e,$e
108	shd	$a,$a,27,$t1
109	xor	@X[($j+8)%16],@X[$j%16],@X[$j%16]
110	addl	@X[$i%16],$e,$e
111	xor	$b,$c,$t0
112	xor	@X[($j+13)%16],@X[$j%16],@X[$j%16]
113	addl	$t1,$e,$e
114	shd	$b,$b,2,$b
115	xor	$d,$t0,$t0
116	shd	@X[$j%16],@X[$j%16],31,@X[$j%16]
117	addl	$t0,$e,$e
118___
119$code.=<<___ if ($i==79);	# with context load
120	ldw	0($ctx),@X[0]	; $i
121	addl	$K,$e,$e
122	shd	$a,$a,27,$t1
123	ldw	4($ctx),@X[1]
124	addl	@X[$i%16],$e,$e
125	xor	$b,$c,$t0
126	ldw	8($ctx),@X[2]
127	addl	$t1,$e,$e
128	shd	$b,$b,2,$b
129	xor	$d,$t0,$t0
130	ldw	12($ctx),@X[3]
131	addl	$t0,$e,$e
132	ldw	16($ctx),@X[4]
133___
134}
135
136sub BODY_40_59 {
137my ($i,$a,$b,$c,$d,$e)=@_;
138my $j=$i+1;
139$code.=<<___;
140	shd	$a,$a,27,$t1	; $i
141	addl	$K,$e,$e
142	xor	@X[($j+2)%16],@X[$j%16],@X[$j%16]
143	xor	$d,$c,$t0
144	addl	@X[$i%16],$e,$e
145	xor	@X[($j+8)%16],@X[$j%16],@X[$j%16]
146	and	$b,$t0,$t0
147	addl	$t1,$e,$e
148	shd	$b,$b,2,$b
149	xor	@X[($j+13)%16],@X[$j%16],@X[$j%16]
150	addl	$t0,$e,$e
151	and	$d,$c,$t1
152	shd	@X[$j%16],@X[$j%16],31,@X[$j%16]
153	addl	$t1,$e,$e
154___
155}
156
157$code=<<___;
158	.LEVEL	$LEVEL
159	.SPACE	\$TEXT\$
160	.SUBSPA	\$CODE\$,QUAD=0,ALIGN=8,ACCESS=0x2C,CODE_ONLY
161
162	.EXPORT	sha1_block_data_order,ENTRY,ARGW0=GR,ARGW1=GR,ARGW2=GR
163sha1_block_data_order
164	.PROC
165	.CALLINFO	FRAME=`$FRAME-14*$SIZE_T`,NO_CALLS,SAVE_RP,ENTRY_GR=16
166	.ENTRY
167	$PUSH	%r2,-$SAVED_RP(%sp)	; standard prologue
168	$PUSHMA	%r3,$FRAME(%sp)
169	$PUSH	%r4,`-$FRAME+1*$SIZE_T`(%sp)
170	$PUSH	%r5,`-$FRAME+2*$SIZE_T`(%sp)
171	$PUSH	%r6,`-$FRAME+3*$SIZE_T`(%sp)
172	$PUSH	%r7,`-$FRAME+4*$SIZE_T`(%sp)
173	$PUSH	%r8,`-$FRAME+5*$SIZE_T`(%sp)
174	$PUSH	%r9,`-$FRAME+6*$SIZE_T`(%sp)
175	$PUSH	%r10,`-$FRAME+7*$SIZE_T`(%sp)
176	$PUSH	%r11,`-$FRAME+8*$SIZE_T`(%sp)
177	$PUSH	%r12,`-$FRAME+9*$SIZE_T`(%sp)
178	$PUSH	%r13,`-$FRAME+10*$SIZE_T`(%sp)
179	$PUSH	%r14,`-$FRAME+11*$SIZE_T`(%sp)
180	$PUSH	%r15,`-$FRAME+12*$SIZE_T`(%sp)
181	$PUSH	%r16,`-$FRAME+13*$SIZE_T`(%sp)
182
183	ldw	0($ctx),$A
184	ldw	4($ctx),$B
185	ldw	8($ctx),$C
186	ldw	12($ctx),$D
187	ldw	16($ctx),$E
188
189	extru	$inp,31,2,$t0		; t0=inp&3;
190	sh3addl	$t0,%r0,$t0		; t0*=8;
191	subi	32,$t0,$t0		; t0=32-t0;
192	mtctl	$t0,%cr11		; %sar=t0;
193
194L\$oop
195	ldi	3,$t0
196	andcm	$inp,$t0,$t0		; 64-bit neutral
197___
198	for ($i=0;$i<15;$i++) {		# load input block
199	$code.="\tldw	`4*$i`($t0),@X[$i]\n";		}
200$code.=<<___;
201	cmpb,*=	$inp,$t0,L\$aligned
202	ldw	60($t0),@X[15]
203	ldw	64($t0),@X[16]
204___
205	for ($i=0;$i<16;$i++) {		# align input
206	$code.="\tvshd	@X[$i],@X[$i+1],@X[$i]\n";	}
207$code.=<<___;
208L\$aligned
209	ldil	L'0x5a827000,$K		; K_00_19
210	ldo	0x999($K),$K
211___
212for ($i=0;$i<20;$i++)   { &BODY_00_19($i,@V); unshift(@V,pop(@V)); }
213$code.=<<___;
214	ldil	L'0x6ed9e000,$K		; K_20_39
215	ldo	0xba1($K),$K
216___
217
218for (;$i<40;$i++)       { &BODY_20_39($i,@V); unshift(@V,pop(@V)); }
219$code.=<<___;
220	ldil	L'0x8f1bb000,$K		; K_40_59
221	ldo	0xcdc($K),$K
222___
223
224for (;$i<60;$i++)       { &BODY_40_59($i,@V); unshift(@V,pop(@V)); }
225$code.=<<___;
226	ldil	L'0xca62c000,$K		; K_60_79
227	ldo	0x1d6($K),$K
228___
229for (;$i<80;$i++)       { &BODY_20_39($i,@V); unshift(@V,pop(@V)); }
230
231$code.=<<___;
232	addl	@X[0],$A,$A
233	addl	@X[1],$B,$B
234	addl	@X[2],$C,$C
235	addl	@X[3],$D,$D
236	addl	@X[4],$E,$E
237	stw	$A,0($ctx)
238	stw	$B,4($ctx)
239	stw	$C,8($ctx)
240	stw	$D,12($ctx)
241	stw	$E,16($ctx)
242	addib,*<> -1,$num,L\$oop
243	ldo	64($inp),$inp
244
245	$POP	`-$FRAME-$SAVED_RP`(%sp),%r2	; standard epilogue
246	$POP	`-$FRAME+1*$SIZE_T`(%sp),%r4
247	$POP	`-$FRAME+2*$SIZE_T`(%sp),%r5
248	$POP	`-$FRAME+3*$SIZE_T`(%sp),%r6
249	$POP	`-$FRAME+4*$SIZE_T`(%sp),%r7
250	$POP	`-$FRAME+5*$SIZE_T`(%sp),%r8
251	$POP	`-$FRAME+6*$SIZE_T`(%sp),%r9
252	$POP	`-$FRAME+7*$SIZE_T`(%sp),%r10
253	$POP	`-$FRAME+8*$SIZE_T`(%sp),%r11
254	$POP	`-$FRAME+9*$SIZE_T`(%sp),%r12
255	$POP	`-$FRAME+10*$SIZE_T`(%sp),%r13
256	$POP	`-$FRAME+11*$SIZE_T`(%sp),%r14
257	$POP	`-$FRAME+12*$SIZE_T`(%sp),%r15
258	$POP	`-$FRAME+13*$SIZE_T`(%sp),%r16
259	bv	(%r2)
260	.EXIT
261	$POPMB	-$FRAME(%sp),%r3
262	.PROCEND
263	.STRINGZ "SHA1 block transform for PA-RISC, CRYPTOGAMS by <appro\@openssl.org>"
264___
265
266if (`$ENV{CC} -Wa,-v -c -o /dev/null -x assembler /dev/null 2>&1`
267	=~ /GNU assembler/) {
268    $gnuas = 1;
269}
270
271foreach(split("\n",$code)) {
272	s/\`([^\`]*)\`/eval $1/ge;
273
274	s/(\.LEVEL\s+2\.0)W/$1w/	if ($gnuas && $SIZE_T==8);
275	s/\.SPACE\s+\$TEXT\$/.text/	if ($gnuas && $SIZE_T==8);
276	s/\.SUBSPA.*//			if ($gnuas && $SIZE_T==8);
277	s/,\*/,/			if ($SIZE_T==4);
278	s/\bbv\b/bve/			if ($SIZE_T==8);
279
280	print $_,"\n";
281}
282close STDOUT or die "error closing STDOUT: $!";
283