]>
Commit | Line | Data |
---|---|---|
3402aac7 | 1 | #------------------------------------------------------------------------------\r |
d39eb83c | 2 | #\r |
3 | # CopyMem() worker for ARM\r | |
4 | #\r | |
5 | # This file started out as C code that did 64 bit moves if the buffer was\r | |
6 | # 32-bit aligned, else it does a byte copy. It also does a byte copy for\r | |
3402aac7 | 7 | # any trailing bytes. It was updated to do 32-byte copies using stm/ldm.\r |
d39eb83c | 8 | #\r |
d6ebcab7 HT |
9 | # Copyright (c) 2008 - 2010, Apple Inc. All rights reserved.<BR>\r |
10 | # This program and the accompanying materials\r | |
d39eb83c | 11 | # are licensed and made available under the terms and conditions of the BSD License\r |
12 | # which accompanies this distribution. The full text of the license may be found at\r | |
13 | # http://opensource.org/licenses/bsd-license.php\r | |
14 | #\r | |
15 | # THE PROGRAM IS DISTRIBUTED UNDER THE BSD LICENSE ON AN "AS IS" BASIS,\r | |
16 | # WITHOUT WARRANTIES OR REPRESENTATIONS OF ANY KIND, EITHER EXPRESS OR IMPLIED.\r | |
17 | #\r | |
18 | #------------------------------------------------------------------------------\r | |
19 | \r | |
20 | /**\r | |
21 | Copy Length bytes from Source to Destination. Overlap is OK.\r | |
22 | \r | |
3402aac7 | 23 | This implementation\r |
d39eb83c | 24 | \r |
25 | @param Destination Target of copy\r | |
26 | @param Source Place to copy from\r | |
27 | @param Length Number of bytes to copy\r | |
28 | \r | |
29 | @return Destination\r | |
30 | \r | |
31 | \r | |
32 | VOID *\r | |
33 | EFIAPI\r | |
34 | InternalMemCopyMem (\r | |
35 | OUT VOID *DestinationBuffer,\r | |
36 | IN CONST VOID *SourceBuffer,\r | |
37 | IN UINTN Length\r | |
38 | )\r | |
39 | **/\r | |
40 | .text\r | |
41 | .align 2\r | |
1bfda055 | 42 | GCC_ASM_EXPORT(InternalMemCopyMem)\r |
d39eb83c | 43 | \r |
44 | ASM_PFX(InternalMemCopyMem):\r | |
2ac288f9 | 45 | stmfd sp!, {r4-r11, lr}\r |
643ec0e7 | 46 | // Save the input parameters in extra registers (r11 = destination, r14 = source, r12 = length)\r |
2ac288f9 | 47 | mov r11, r0\r |
48 | mov r10, r0\r | |
643ec0e7 | 49 | mov r12, r2\r |
50 | mov r14, r1\r | |
3402aac7 | 51 | \r |
643ec0e7 | 52 | memcopy_check_overlapped:\r |
53 | cmp r11, r1\r | |
54 | // If (dest < source)\r | |
55 | bcc memcopy_check_optim_default\r | |
56 | // If (dest <= source). But with the previous condition -> If (dest == source)\r | |
57 | bls memcopy_end\r | |
58 | \r | |
59 | // If (source + length < dest)\r | |
60 | rsb r3, r1, r11\r | |
61 | cmp r12, r3\r | |
62 | bcc memcopy_check_optim_default\r | |
63 | \r | |
3402aac7 | 64 | // If (length == 0)\r |
643ec0e7 | 65 | cmp r12, #0\r |
66 | beq memcopy_end\r | |
3402aac7 | 67 | \r |
643ec0e7 | 68 | b memcopy_check_optim_overlap\r |
69 | \r | |
70 | memcopy_check_optim_default:\r | |
71 | // Check if we can use an optimized path ((length >= 32) && destination word-aligned && source word-aligned) for the memcopy (optimized path if r0 == 1)\r | |
72 | tst r0, #0xF\r | |
2ac288f9 | 73 | movne r0, #0\r |
643ec0e7 | 74 | bne memcopy_default\r |
75 | tst r1, #0xF\r | |
2ac288f9 | 76 | movne r3, #0\r |
77 | moveq r3, #1\r | |
78 | cmp r2, #31\r | |
79 | movls r0, #0\r | |
80 | andhi r0, r3, #1\r | |
643ec0e7 | 81 | b memcopy_default\r |
3402aac7 | 82 | \r |
643ec0e7 | 83 | memcopy_check_optim_overlap:\r |
84 | // r10 = dest_end, r14 = source_end\r | |
85 | add r10, r11, r12\r | |
86 | add r14, r12, r1\r | |
87 | \r | |
88 | // Are we in the optimized case ((length >= 32) && dest_end word-aligned && source_end word-aligned)\r | |
89 | cmp r2, #31\r | |
2ac288f9 | 90 | movls r0, #0\r |
643ec0e7 | 91 | movhi r0, #1\r |
92 | tst r10, #0xF\r | |
93 | movne r0, #0\r | |
94 | tst r14, #0xF\r | |
95 | movne r0, #0\r | |
96 | b memcopy_overlapped\r | |
3402aac7 | 97 | \r |
643ec0e7 | 98 | memcopy_overlapped_non_optim:\r |
99 | // We read 1 byte from the end of the source buffer\r | |
100 | sub r3, r14, #1\r | |
101 | sub r12, r12, #1\r | |
3402aac7 | 102 | ldrb r3, [r3, #0]\r |
2ac288f9 | 103 | sub r2, r10, #1\r |
643ec0e7 | 104 | cmp r12, #0\r |
105 | // We write 1 byte at the end of the dest buffer\r | |
2ac288f9 | 106 | sub r10, r10, #1\r |
643ec0e7 | 107 | sub r14, r14, #1\r |
2ac288f9 | 108 | strb r3, [r2, #0]\r |
643ec0e7 | 109 | bne memcopy_overlapped_non_optim\r |
110 | b memcopy_end\r | |
111 | \r | |
112 | // r10 = dest_end, r14 = source_end\r | |
113 | memcopy_overlapped:\r | |
114 | // Are we in the optimized case ?\r | |
115 | cmp r0, #0\r | |
116 | beq memcopy_overlapped_non_optim\r | |
3402aac7 | 117 | \r |
643ec0e7 | 118 | // Optimized Overlapped - Read 32 bytes\r |
119 | sub r14, r14, #32\r | |
120 | sub r12, r12, #32\r | |
121 | cmp r12, #31\r | |
122 | ldmia r14, {r2-r9}\r | |
3402aac7 | 123 | \r |
643ec0e7 | 124 | // If length is less than 32 then disable optim\r |
125 | movls r0, #0\r | |
3402aac7 | 126 | \r |
643ec0e7 | 127 | cmp r12, #0\r |
3402aac7 RC |
128 | \r |
129 | // Optimized Overlapped - Write 32 bytes\r | |
643ec0e7 | 130 | sub r10, r10, #32\r |
131 | stmia r10, {r2-r9}\r | |
3402aac7 | 132 | \r |
643ec0e7 | 133 | // while (length != 0)\r |
134 | bne memcopy_overlapped\r | |
135 | b memcopy_end\r | |
3402aac7 | 136 | \r |
643ec0e7 | 137 | memcopy_default_non_optim:\r |
138 | // Byte copy\r | |
3402aac7 | 139 | ldrb r3, [r14], #1\r |
643ec0e7 | 140 | sub r12, r12, #1\r |
2ac288f9 | 141 | strb r3, [r10], #1\r |
3402aac7 | 142 | \r |
643ec0e7 | 143 | memcopy_default:\r |
144 | cmp r12, #0\r | |
145 | beq memcopy_end\r | |
3402aac7 | 146 | \r |
643ec0e7 | 147 | // r10 = dest, r14 = source\r |
148 | memcopy_default_loop:\r | |
2ac288f9 | 149 | cmp r0, #0\r |
643ec0e7 | 150 | beq memcopy_default_non_optim\r |
3402aac7 | 151 | \r |
643ec0e7 | 152 | // Optimized memcopy - Read 32 Bytes\r |
153 | sub r12, r12, #32\r | |
154 | cmp r12, #31\r | |
155 | ldmia r14!, {r2-r9}\r | |
3402aac7 | 156 | \r |
643ec0e7 | 157 | // If length is less than 32 then disable optim\r |
2ac288f9 | 158 | movls r0, #0\r |
3402aac7 | 159 | \r |
643ec0e7 | 160 | cmp r12, #0\r |
3402aac7 | 161 | \r |
643ec0e7 | 162 | // Optimized memcopy - Write 32 Bytes\r |
2ac288f9 | 163 | stmia r10!, {r2-r9}\r |
643ec0e7 | 164 | \r |
165 | // while (length != 0)\r | |
166 | bne memcopy_default_loop\r | |
3402aac7 | 167 | \r |
643ec0e7 | 168 | memcopy_end:\r |
2ac288f9 | 169 | mov r0, r11\r |
170 | ldmfd sp!, {r4-r11, pc}\r | |
3402aac7 | 171 | \r |