memset.S revision eb847bc8666842a3cfc9c06e8458ad1abebebaf0 
1/*
2 * Copyright (C) 2013 The Android Open Source Project
3 * All rights reserved.
4 *
5 * Redistribution and use in source and binary forms, with or without
6 * modification, are permitted provided that the following conditions
7 * are met:
8 *  * Redistributions of source code must retain the above copyright
9 *    notice, this list of conditions and the following disclaimer.
10 *  * Redistributions in binary form must reproduce the above copyright
11 *    notice, this list of conditions and the following disclaimer in
12 *    the documentation and/or other materials provided with the
13 *    distribution.
14 *
15 * THIS SOFTWARE IS PROVIDED BY THE COPYRIGHT HOLDERS AND CONTRIBUTORS
16 * "AS IS" AND ANY EXPRESS OR IMPLIED WARRANTIES, INCLUDING, BUT NOT
17 * LIMITED TO, THE IMPLIED WARRANTIES OF MERCHANTABILITY AND FITNESS
18 * FOR A PARTICULAR PURPOSE ARE DISCLAIMED. IN NO EVENT SHALL THE
19 * COPYRIGHT OWNER OR CONTRIBUTORS BE LIABLE FOR ANY DIRECT, INDIRECT,
20 * INCIDENTAL, SPECIAL, EXEMPLARY, OR CONSEQUENTIAL DAMAGES (INCLUDING,
21 * BUT NOT LIMITED TO, PROCUREMENT OF SUBSTITUTE GOODS OR SERVICES; LOSS
22 * OF USE, DATA, OR PROFITS; OR BUSINESS INTERRUPTION) HOWEVER CAUSED
23 * AND ON ANY THEORY OF LIABILITY, WHETHER IN CONTRACT, STRICT LIABILITY,
24 * OR TORT (INCLUDING NEGLIGENCE OR OTHERWISE) ARISING IN ANY WAY OUT
25 * OF THE USE OF THIS SOFTWARE, EVEN IF ADVISED OF THE POSSIBILITY OF
26 * SUCH DAMAGE.
27 */
28
29#include <machine/cpu-features.h>
30#include <machine/asm.h>
31#include "private/libc_events.h"
32
33        /*
34         * Optimized memset() for ARM.
35         *
36         * memset() returns its first argument.
37         */
38
39        .fpu        neon
40        .syntax     unified
41
42ENTRY(__memset_chk)
43        .cfi_startproc
44        cmp         r2, r3
45        bls         .L_done
46
47        // Preserve lr for backtrace.
48        .save       {lr}
49        push        {lr}
50        .cfi_def_cfa_offset 4
51        .cfi_rel_offset lr, 0
52
53        ldr         r0, error_message
54        ldr         r1, error_code
551:
56        add         r0, pc
57        bl          __fortify_chk_fail
58error_code:
59        .word       BIONIC_EVENT_MEMSET_BUFFER_OVERFLOW
60error_message:
61        .word       error_string-(1b+8)
62
63        .cfi_endproc
64END(__memset_chk)
65
66ENTRY(bzero)
67        .cfi_startproc
68        mov         r2, r1
69        mov         r1, #0
70.L_done:
71        // Fall through to memset...
72        .cfi_endproc
73END(bzero)
74
75ENTRY(memset)
76        .cfi_startproc
77        .save       {r0}
78        stmfd       sp!, {r0}
79        .cfi_def_cfa_offset 4
80        .cfi_rel_offset r0, 0
81
82        // The new algorithm is slower for copies < 16 so use the old
83        // neon code in that case.
84        cmp         r2, #16
85        blo         .L_set_less_than_16_unknown_align
86
87        // Use strd which requires an even and odd register so move the
88        // values so that:
89        //   r0 and r1 contain the memset value
90        //   r2 is the number of bytes to set
91        //   r3 is the destination pointer
92        mov         r3, r0
93
94        // Copy the byte value in every byte of r1.
95        mov         r1, r1, lsl #24
96        orr         r1, r1, r1, lsr #8
97        orr         r1, r1, r1, lsr #16
98
99.L_check_alignment:
100        // Align destination to a double word to avoid the strd crossing
101        // a cache line boundary.
102        ands        ip, r3, #7
103        bne         .L_do_double_word_align
104
105.L_double_word_aligned:
106        mov         r0, r1
107
108        subs        r2, #64
109        blo         .L_set_less_than_64
110
1111:      // Main loop sets 64 bytes at a time.
112        .irp        offset, #0, #8, #16, #24, #32, #40, #48, #56
113        strd        r0, r1, [r3, \offset]
114        .endr
115
116        add         r3, #64
117        subs        r2, #64
118        bge         1b
119
120.L_set_less_than_64:
121        // Restore r2 to the count of bytes left to set.
122        add         r2, #64
123        lsls        ip, r2, #27
124        bcc         .L_set_less_than_32
125        // Set 32 bytes.
126        .irp        offset, #0, #8, #16, #24
127        strd        r0, r1, [r3, \offset]
128        .endr
129        add         r3, #32
130
131.L_set_less_than_32:
132        bpl         .L_set_less_than_16
133        // Set 16 bytes.
134        .irp        offset, #0, #8
135        strd        r0, r1, [r3, \offset]
136        .endr
137        add         r3, #16
138
139.L_set_less_than_16:
140        // Less than 16 bytes to set.
141        lsls        ip, r2, #29
142        bcc         .L_set_less_than_8
143
144        // Set 8 bytes.
145        strd        r0, r1, [r3], #8
146
147.L_set_less_than_8:
148        bpl         .L_set_less_than_4
149        // Set 4 bytes
150        str         r1, [r3], #4
151
152.L_set_less_than_4:
153        lsls        ip, r2, #31
154        it          ne
155        strbne      r1, [r3], #1
156        itt         cs
157        strbcs      r1, [r3], #1
158        strbcs      r1, [r3]
159
160        ldmfd       sp!, {r0}
161        bx          lr
162
163.L_do_double_word_align:
164        rsb         ip, ip, #8
165        sub         r2, r2, ip
166        movs        r0, ip, lsl #31
167        it          mi
168        strbmi      r1, [r3], #1
169        itt         cs
170        strbcs      r1, [r3], #1
171        strbcs      r1, [r3], #1
172
173        // Dst is at least word aligned by this point.
174        cmp         ip, #4
175        blo         .L_double_word_aligned
176        str         r1, [r3], #4
177        b           .L_double_word_aligned
178
179.L_set_less_than_16_unknown_align:
180        // Set up to 15 bytes.
181        vdup.8      d0, r1
182        movs        ip, r2, lsl #29
183        bcc         1f
184        vst1.8      {d0}, [r0]!
1851:      bge         2f
186        vst1.32     {d0[0]}, [r0]!
1872:      movs        ip, r2, lsl #31
188        it          mi
189        strbmi      r1, [r0], #1
190        itt         cs
191        strbcs      r1, [r0], #1
192        strbcs      r1, [r0], #1
193        ldmfd       sp!, {r0}
194        bx          lr
195        .cfi_endproc
196END(memset)
197
198        .data
199error_string:
200        .string     "memset prevented write past end of buffer"
201