about summary refs log tree commit diff
path: root/sysdeps/aarch64/multiarch/memset_base64.S
blob: c0cccbac648153e18124ccf689ad7d26edf77016 (plain) (blame)
1
2
3
4
5
6
7
8
9
10
11
12
13
14
15
16
17
18
19
20
21
22
23
24
25
26
27
28
29
30
31
32
33
34
35
36
37
38
39
40
41
42
43
44
45
46
47
48
49
50
51
52
53
54
55
56
57
58
59
60
61
62
63
64
65
66
67
68
69
70
71
72
73
74
75
76
77
78
79
80
81
82
83
84
85
86
87
88
89
90
91
92
93
94
95
96
97
98
99
100
101
102
103
104
105
106
107
108
109
110
111
112
113
114
115
116
117
118
119
120
121
122
123
124
125
126
127
128
129
130
131
132
133
134
135
136
137
138
139
140
141
142
143
144
145
146
147
148
149
150
151
152
153
154
155
156
157
158
159
160
161
162
163
164
165
166
167
168
169
170
171
172
173
174
175
176
177
178
179
180
181
182
183
184
185
186
/* Copyright (C) 2018 Free Software Foundation, Inc.

   This file is part of the GNU C Library.

   The GNU C Library is free software; you can redistribute it and/or
   modify it under the terms of the GNU Lesser General Public
   License as published by the Free Software Foundation; either
   version 2.1 of the License, or (at your option) any later version.

   The GNU C Library is distributed in the hope that it will be useful,
   but WITHOUT ANY WARRANTY; without even the implied warranty of
   MERCHANTABILITY or FITNESS FOR A PARTICULAR PURPOSE.  See the GNU
   Lesser General Public License for more details.

   You should have received a copy of the GNU Lesser General Public
   License along with the GNU C Library.  If not, see
   <http://www.gnu.org/licenses/>.  */

#include <sysdep.h>
#include "memset-reg.h"

#ifndef MEMSET
# define MEMSET __memset_base64
#endif

/* To disable DC ZVA, set this threshold to 0. */
#ifndef DC_ZVA_THRESHOLD
# define DC_ZVA_THRESHOLD 512
#endif

/* Assumptions:
 *
 * ARMv8-a, AArch64, unaligned accesses
 *
 */

ENTRY_ALIGN (MEMSET, 6)

	DELOUSE (0)
	DELOUSE (2)

	bfi	valw, valw, 8, 8
	bfi	valw, valw, 16, 16
	bfi	val, val, 32, 32

	add	dstend, dstin, count

	cmp	count, 96
	b.hi	L(set_long)
	cmp	count, 16
	b.hs	L(set_medium)

	/* Set 0..15 bytes.  */
	tbz	count, 3, 1f
	str	val, [dstin]
	str	val, [dstend, -8]
	ret

	.p2align 3
1:	tbz	count, 2, 2f
	str	valw, [dstin]
	str	valw, [dstend, -4]
	ret
2:	cbz	count, 3f
	strb	valw, [dstin]
	tbz	count, 1, 3f
	strh	valw, [dstend, -2]
3:	ret

	.p2align 3
	/* Set 16..96 bytes.  */
L(set_medium):
	stp	val, val, [dstin]
	tbnz	count, 6, L(set96)
	stp	val, val, [dstend, -16]
	tbz	count, 5, 1f
	stp	val, val, [dstin, 16]
	stp	val, val, [dstend, -32]
1:	ret

	.p2align 4
	/* Set 64..96 bytes.  Write 64 bytes from the start and
	   32 bytes from the end.  */
L(set96):
	stp	val, val, [dstin, 16]
	stp	val, val, [dstin, 32]
	stp	val, val, [dstin, 48]
	stp	val, val, [dstend, -32]
	stp	val, val, [dstend, -16]
	ret

	.p2align 4
L(set_long):
	stp	val, val, [dstin]
	bic	dst, dstin, 15
#if DC_ZVA_THRESHOLD
	cmp	count, DC_ZVA_THRESHOLD
	ccmp	val, 0, 0, cs
	b.eq	L(zva_64)
#endif
	/* Small-size or non-zero memset does not use DC ZVA. */
	sub	count, dstend, dst

	/*
	 * Adjust count and bias for loop. By substracting extra 1 from count,
	 * it is easy to use tbz instruction to check whether loop tailing
	 * count is less than 33 bytes, so as to bypass 2 unneccesary stps.
	 */
	sub	count, count, 64+16+1

#if DC_ZVA_THRESHOLD
	/* Align loop on 16-byte boundary, this might be friendly to i-cache. */
	nop
#endif

1:	stp	val, val, [dst, 16]
	stp	val, val, [dst, 32]
	stp	val, val, [dst, 48]
	stp	val, val, [dst, 64]!
	subs	count, count, 64
	b.hs	1b

	tbz	count, 5, 1f	/* Remaining count is less than 33 bytes? */
	stp	val, val, [dst, 16]
	stp	val, val, [dst, 32]
1:	stp	val, val, [dstend, -32]
	stp	val, val, [dstend, -16]
	ret

#if DC_ZVA_THRESHOLD
	.p2align 3
L(zva_64):
	stp	val, val, [dst, 16]
	stp	val, val, [dst, 32]
	stp	val, val, [dst, 48]
	bic	dst, dst, 63

	/*
	 * Previous memory writes might cross cache line boundary, and cause
	 * cache line partially dirty. Zeroing this kind of cache line using
	 * DC ZVA will incur extra cost, for it requires loading untouched
	 * part of the line from memory before zeoring.
	 *
	 * So, write the first 64 byte aligned block using stp to force
	 * fully dirty cache line.
	 */
	stp	val, val, [dst, 64]
	stp	val, val, [dst, 80]
	stp	val, val, [dst, 96]
	stp	val, val, [dst, 112]

	sub	count, dstend, dst
	/*
	 * Adjust count and bias for loop. By substracting extra 1 from count,
	 * it is easy to use tbz instruction to check whether loop tailing
	 * count is less than 33 bytes, so as to bypass 2 unneccesary stps.
	 */
	sub	count, count, 128+64+64+1
	add	dst, dst, 128
	nop

	/* DC ZVA sets 64 bytes each time. */
1:	dc	zva, dst
	add	dst, dst, 64
	subs	count, count, 64
	b.hs	1b

	/*
	 * Write the last 64 byte aligned block using stp to force fully
	 * dirty cache line.
	 */
	stp	val, val, [dst, 0]
	stp	val, val, [dst, 16]
	stp	val, val, [dst, 32]
	stp	val, val, [dst, 48]

	tbz	count, 5, 1f	/* Remaining count is less than 33 bytes? */
	stp	val, val, [dst, 64]
	stp	val, val, [dst, 80]
1:	stp	val, val, [dstend, -32]
	stp	val, val, [dstend, -16]
	ret
#endif

END (MEMSET)
libc_hidden_builtin_def (MEMSET)