]>
Commit | Line | Data |
---|---|---|
a37f6605 AB |
1 | ;\r |
2 | ; Copyright (c) 2013 - 2016, Linaro Limited\r | |
3 | ; All rights reserved.\r | |
4 | ;\r | |
5 | ; Redistribution and use in source and binary forms, with or without\r | |
6 | ; modification, are permitted provided that the following conditions are met:\r | |
7 | ; * Redistributions of source code must retain the above copyright\r | |
8 | ; notice, this list of conditions and the following disclaimer.\r | |
9 | ; * Redistributions in binary form must reproduce the above copyright\r | |
10 | ; notice, this list of conditions and the following disclaimer in the\r | |
11 | ; documentation and/or other materials provided with the distribution.\r | |
12 | ; * Neither the name of the Linaro nor the\r | |
13 | ; names of its contributors may be used to endorse or promote products\r | |
14 | ; derived from this software without specific prior written permission.\r | |
15 | ;\r | |
16 | ; THIS SOFTWARE IS PROVIDED BY THE COPYRIGHT HOLDERS AND CONTRIBUTORS\r | |
17 | ; "AS IS" AND ANY EXPRESS OR IMPLIED WARRANTIES, INCLUDING, BUT NOT\r | |
18 | ; LIMITED TO, THE IMPLIED WARRANTIES OF MERCHANTABILITY AND FITNESS FOR\r | |
19 | ; A PARTICULAR PURPOSE ARE DISCLAIMED. IN NO EVENT SHALL THE COPYRIGHT\r | |
20 | ; HOLDER OR CONTRIBUTORS BE LIABLE FOR ANY DIRECT, INDIRECT, INCIDENTAL,\r | |
21 | ; SPECIAL, EXEMPLARY, OR CONSEQUENTIAL DAMAGES (INCLUDING, BUT NOT\r | |
22 | ; LIMITED TO, PROCUREMENT OF SUBSTITUTE GOODS OR SERVICES; LOSS OF USE,\r | |
23 | ; DATA, OR PROFITS; OR BUSINESS INTERRUPTION) HOWEVER CAUSED AND ON ANY\r | |
24 | ; THEORY OF LIABILITY, WHETHER IN CONTRACT, STRICT LIABILITY, OR TORT\r | |
25 | ; (INCLUDING NEGLIGENCE OR OTHERWISE) ARISING IN ANY WAY OUT OF THE USE\r | |
26 | ; OF THIS SOFTWARE, EVEN IF ADVISED OF THE POSSIBILITY OF SUCH DAMAGE.\r | |
27 | ;\r | |
28 | \r | |
29 | ; Parameters and result.\r | |
30 | #define src1 r0\r | |
31 | #define src2 r1\r | |
32 | #define limit r2\r | |
33 | #define result r0\r | |
34 | \r | |
35 | ; Internal variables.\r | |
36 | #define data1 r3\r | |
37 | #define data2 r4\r | |
38 | #define limit_wd r5\r | |
39 | #define diff r6\r | |
40 | #define tmp1 r7\r | |
41 | #define tmp2 r12\r | |
42 | #define pos r8\r | |
43 | #define mask r14\r | |
44 | \r | |
45 | EXPORT InternalMemCompareMem\r | |
46 | THUMB\r | |
47 | AREA CompareMem, CODE, READONLY\r | |
48 | \r | |
49 | InternalMemCompareMem\r | |
50 | push {r4-r8, lr}\r | |
51 | eor tmp1, src1, src2\r | |
52 | tst tmp1, #3\r | |
53 | bne Lmisaligned4\r | |
54 | ands tmp1, src1, #3\r | |
55 | bne Lmutual_align\r | |
56 | add limit_wd, limit, #3\r | |
57 | nop.w\r | |
58 | lsr limit_wd, limit_wd, #2\r | |
59 | \r | |
60 | ; Start of performance-critical section -- one 32B cache line.\r | |
61 | Lloop_aligned\r | |
62 | ldr data1, [src1], #4\r | |
63 | ldr data2, [src2], #4\r | |
64 | Lstart_realigned\r | |
65 | subs limit_wd, limit_wd, #1\r | |
66 | eor diff, data1, data2 ; Non-zero if differences found.\r | |
67 | cbnz diff, L0\r | |
68 | bne Lloop_aligned\r | |
69 | ; End of performance-critical section -- one 32B cache line.\r | |
70 | \r | |
71 | ; Not reached the limit, must have found a diff.\r | |
72 | L0\r | |
73 | cbnz limit_wd, Lnot_limit\r | |
74 | \r | |
75 | // Limit % 4 == 0 => all bytes significant.\r | |
76 | ands limit, limit, #3\r | |
77 | beq Lnot_limit\r | |
78 | \r | |
79 | lsl limit, limit, #3 // Bits -> bytes.\r | |
80 | mov mask, #~0\r | |
81 | lsl mask, mask, limit\r | |
82 | bic data1, data1, mask\r | |
83 | bic data2, data2, mask\r | |
84 | \r | |
85 | orr diff, diff, mask\r | |
86 | \r | |
87 | Lnot_limit\r | |
88 | rev diff, diff\r | |
89 | rev data1, data1\r | |
90 | rev data2, data2\r | |
91 | \r | |
92 | ; The MS-non-zero bit of DIFF marks either the first bit\r | |
93 | ; that is different, or the end of the significant data.\r | |
94 | ; Shifting left now will bring the critical information into the\r | |
95 | ; top bits.\r | |
96 | clz pos, diff\r | |
97 | lsl data1, data1, pos\r | |
98 | lsl data2, data2, pos\r | |
99 | \r | |
100 | ; But we need to zero-extend (char is unsigned) the value and then\r | |
101 | ; perform a signed 32-bit subtraction.\r | |
102 | lsr data1, data1, #28\r | |
103 | sub result, data1, data2, lsr #28\r | |
104 | pop {r4-r8, pc}\r | |
105 | \r | |
106 | Lmutual_align\r | |
107 | ; Sources are mutually aligned, but are not currently at an\r | |
108 | ; alignment boundary. Round down the addresses and then mask off\r | |
109 | ; the bytes that precede the start point.\r | |
110 | bic src1, src1, #3\r | |
111 | bic src2, src2, #3\r | |
112 | add limit, limit, tmp1 ; Adjust the limit for the extra.\r | |
113 | lsl tmp1, tmp1, #2 ; Bytes beyond alignment -> bits.\r | |
114 | ldr data1, [src1], #4\r | |
115 | neg tmp1, tmp1 ; Bits to alignment -32.\r | |
116 | ldr data2, [src2], #4\r | |
117 | mov tmp2, #~0\r | |
118 | \r | |
119 | ; Little-endian. Early bytes are at LSB.\r | |
120 | lsr tmp2, tmp2, tmp1 ; Shift (tmp1 & 31).\r | |
121 | add limit_wd, limit, #3\r | |
122 | orr data1, data1, tmp2\r | |
123 | orr data2, data2, tmp2\r | |
124 | lsr limit_wd, limit_wd, #2\r | |
125 | b Lstart_realigned\r | |
126 | \r | |
127 | Lmisaligned4\r | |
128 | sub limit, limit, #1\r | |
129 | L1\r | |
130 | // Perhaps we can do better than this.\r | |
131 | ldrb data1, [src1], #1\r | |
132 | ldrb data2, [src2], #1\r | |
133 | subs limit, limit, #1\r | |
134 | it cs\r | |
135 | cmpcs data1, data2\r | |
136 | beq L1\r | |
137 | sub result, data1, data2\r | |
138 | pop {r4-r8, pc}\r | |
139 | \r | |
140 | END\r |