Gitiles
Code Review Sign In
review.evervolv.com / android_bionic / refs/heads/jellybean-4.3 / . / libc / arch-arm / bionic / memmove.S
blob: 788ba3c512d326fba2992b1f17ebd47c9c14a7d0 [file] [log] [blame]
athurh2fbb5762013-08-10 22:21:23 +0200[diff] [blame]1/***************************************************************************
2 Copyright (c) 2009-2012 Code Aurora Forum. All rights reserved.
3
4 Redistribution and use in source and binary forms, with or without
5 modification, are permitted provided that the following conditions are met:
6 * Redistributions of source code must retain the above copyright
7 notice, this list of conditions and the following disclaimer.
8 * Redistributions in binary form must reproduce the above copyright
9 notice, this list of conditions and the following disclaimer in the
10 documentation and/or other materials provided with the distribution.
11 * Neither the name of Code Aurora nor the names of its contributors may
12 be used to endorse or promote products derived from this software
13 without specific prior written permission.
14
15 THIS SOFTWARE IS PROVIDED BY THE COPYRIGHT HOLDERS AND CONTRIBUTORS "AS IS"
16 AND ANY EXPRESS OR IMPLIED WARRANTIES, INCLUDING, BUT NOT LIMITED TO, THE
17 IMPLIED WARRANTIES OF MERCHANTABILITY AND FITNESS FOR A PARTICULAR PURPOSE
18 ARE DISCLAIMED. IN NO EVENT SHALL THE COPYRIGHT HOLDER OR CONTRIBUTORS BE
19 LIABLE FOR ANY DIRECT, INDIRECT, INCIDENTAL, SPECIAL, EXEMPLARY, OR
20 CONSEQUENTIAL DAMAGES (INCLUDING, BUT NOT LIMITED TO, PROCUREMENT OF
21 SUBSTITUTE GOODS OR SERVICES; LOSS OF USE, DATA, OR PROFITS; OR BUSINESS
22 INTERRUPTION) HOWEVER CAUSED AND ON ANY THEORY OF LIABILITY, WHETHER IN
23 CONTRACT, STRICT LIABILITY, OR TORT (INCLUDING NEGLIGENCE OR OTHERWISE)
24 ARISING IN ANY WAY OUT OF THE USE OF THIS SOFTWARE, EVEN IF ADVISED OF THE
25 POSSIBILITY OF SUCH DAMAGE.
26 ***************************************************************************/
27
28/***************************************************************************
29 * Neon memmove: Attempts to do a memmove with Neon registers if possible,
30 * Inputs:
31 * dest: The destination buffer
32 * src: The source buffer
33 * n: The size of the buffer to transfer
34 * Outputs:
35 *
36 ***************************************************************************/
37
38#include <machine/cpu-features.h>
39
40#ifndef PLDOFFS
41#define PLDOFFS (10)
42#endif
43#ifndef PLDTHRESH
44#define PLDTHRESH (PLDOFFS)
45#endif
46#if (PLDOFFS < 5)
47#error Routine does not support offsets less than 5
48#endif
49#if (PLDTHRESH < PLDOFFS)
50#error PLD threshold must be greater than or equal to the PLD offset
51#endif
52#ifndef PLDSIZE
53#define PLDSIZE (64)
54#endif
55#define NOP_OPCODE (0xe320f000)
56
57 .code 32
58 .align 5
59 .global memmove
60 .type memmove, %function
61
62 .global bcopy
63 .type bcopy, %function
64
65bcopy:
66 mov r12, r0
67 mov r0, r1
68 mov r1, r12
69 .balignl 64, NOP_OPCODE, 4*2
70memmove:
71.Lneon_memmove_cmf:
72 subs r12, r0, r1
73 bxeq lr
74 cmphi r2, r12
75 bls memcpy /* Use memcpy for non-overlapping areas */
76
77 push {r0}
78
79.Lneon_back_to_front_copy:
80 add r0, r0, r2
81 add r1, r1, r2
82 cmp r2, #4
83 bgt .Lneon_b2f_gt4
84 cmp r2, #0
85.Lneon_b2f_smallcopy_loop:
86 beq .Lneon_memmove_done
87 ldrb r12, [r1, #-1]!
88 subs r2, r2, #1
89 strb r12, [r0, #-1]!
90 b .Lneon_b2f_smallcopy_loop
91.Lneon_b2f_gt4:
92 sub r3, r0, r1
93 cmp r2, r3
94 movle r12, r2
95 movgt r12, r3
96 cmp r12, #64
97 bge .Lneon_b2f_copy_64
98 cmp r12, #32
99 bge .Lneon_b2f_copy_32
100 cmp r12, #8
101 bge .Lneon_b2f_copy_8
102 cmp r12, #4
103 bge .Lneon_b2f_copy_4
104 b .Lneon_b2f_copy_1
105.Lneon_b2f_copy_64:
106 sub r1, r1, #64 /* Predecrement */
107 sub r0, r0, #64
108 movs r12, r2, lsr #6
109 cmp r12, #PLDTHRESH
110 ble .Lneon_b2f_copy_64_loop_nopld
111 sub r12, #PLDOFFS
112 pld [r1, #-(PLDOFFS-5)*PLDSIZE]
113 pld [r1, #-(PLDOFFS-4)*PLDSIZE]
114 pld [r1, #-(PLDOFFS-3)*PLDSIZE]
115 pld [r1, #-(PLDOFFS-2)*PLDSIZE]
116 pld [r1, #-(PLDOFFS-1)*PLDSIZE]
117 .balignl 64, NOP_OPCODE, 4*2
118.Lneon_b2f_copy_64_loop_outer:
119 pld [r1, #-(PLDOFFS)*PLDSIZE]
120 vld1.32 {q0, q1}, [r1]!
121 vld1.32 {q2, q3}, [r1]
122 subs r12, r12, #1
123 vst1.32 {q0, q1}, [r0]!
124 sub r1, r1, #96 /* Post-fixup and predecrement */
125 vst1.32 {q2, q3}, [r0]
126 sub r0, r0, #96
127 bne .Lneon_b2f_copy_64_loop_outer
128 mov r12, #PLDOFFS
129 .balignl 64, NOP_OPCODE, 4*2
130.Lneon_b2f_copy_64_loop_nopld:
131 vld1.32 {q8, q9}, [r1]!
132 vld1.32 {q10, q11}, [r1]
133 subs r12, r12, #1
134 vst1.32 {q8, q9}, [r0]!
135 sub r1, r1, #96 /* Post-fixup and predecrement */
136 vst1.32 {q10, q11}, [r0]
137 sub r0, r0, #96
138 bne .Lneon_b2f_copy_64_loop_nopld
139 ands r2, r2, #0x3f
140 beq .Lneon_memmove_done
141 add r1, r1, #64 /* Post-fixup */
142 add r0, r0, #64
143 cmp r2, #32
144 blt .Lneon_b2f_copy_finish
145.Lneon_b2f_copy_32:
146 mov r12, r2, lsr #5
147.Lneon_b2f_copy_32_loop:
148 sub r1, r1, #32 /* Predecrement */
149 sub r0, r0, #32
150 vld1.32 {q0,q1}, [r1]
151 subs r12, r12, #1
152 vst1.32 {q0,q1}, [r0]
153 bne .Lneon_b2f_copy_32_loop
154 ands r2, r2, #0x1f
155 beq .Lneon_memmove_done
156.Lneon_b2f_copy_finish:
157.Lneon_b2f_copy_8:
158 movs r12, r2, lsr #0x3
159 beq .Lneon_b2f_copy_4
160 .balignl 64, NOP_OPCODE, 4*2
161.Lneon_b2f_copy_8_loop:
162 sub r1, r1, #8 /* Predecrement */
163 sub r0, r0, #8
164 vld1.32 {d0}, [r1]
165 subs r12, r12, #1
166 vst1.32 {d0}, [r0]
167 bne .Lneon_b2f_copy_8_loop
168 ands r2, r2, #0x7
169 beq .Lneon_memmove_done
170.Lneon_b2f_copy_4:
171 movs r12, r2, lsr #0x2
172 beq .Lneon_b2f_copy_1
173.Lneon_b2f_copy_4_loop:
174 ldr r3, [r1, #-4]!
175 subs r12, r12, #1
176 str r3, [r0, #-4]!
177 bne .Lneon_b2f_copy_4_loop
178 ands r2, r2, #0x3
179.Lneon_b2f_copy_1:
180 cmp r2, #0
181 beq .Lneon_memmove_done
182 .balignl 64, NOP_OPCODE, 4*2
183.Lneon_b2f_copy_1_loop:
184 ldrb r12, [r1, #-1]!
185 subs r2, r2, #1
186 strb r12, [r0, #-1]!
187 bne .Lneon_b2f_copy_1_loop
188
189.Lneon_memmove_done:
190 pop {r0}
191 bx lr
192
193 .end