211 lines
8.0 KiB
Plaintext
211 lines
8.0 KiB
Plaintext
|
1 # 1 "lib/div64.S"
|
||
|
|
1 /*
|
||
|
|
0
|
||
|
|
0
|
||
|
|
2 * linux/arch/arm/lib/div64.S
|
||
|
|
3 *
|
||
|
|
4 * Optimized computation of 64-bit dividend / 32-bit divisor
|
||
|
|
5 *
|
||
|
|
6 * Author: Nicolas Pitre
|
||
|
|
7 * Created: Oct 5, 2003
|
||
|
|
8 * Copyright: Monta Vista Software, Inc.
|
||
|
|
9 *
|
||
|
|
10 * This program is free software; you can redistribute it and/or modify
|
||
|
|
11 * it under the terms of the GNU General Public License version 2 as
|
||
|
|
12 * published by the Free Software Foundation.
|
||
|
|
13 */
|
||
|
|
14
|
||
|
|
15 #include <asm/linkage.h>
|
||
|
|
1 #ifndef __ASM_LINKAGE_H
|
||
|
|
16
|
||
|
|
17 #ifdef __ARMEB__
|
||
|
|
18 #define xh r0
|
||
|
|
19 #define xl r1
|
||
|
|
20 #define yh r2
|
||
|
|
21 #define yl r3
|
||
|
|
22 #else
|
||
|
|
23 #define xl r0
|
||
|
|
24 #define xh r1
|
||
|
|
25 #define yl r2
|
||
|
|
26 #define yh r3
|
||
|
|
27 #endif
|
||
|
|
28
|
||
|
|
29 /*
|
||
|
|
30 * __do_div64: perform a division with 64-bit dividend and 32-bit divisor.
|
||
|
|
31 *
|
||
|
|
32 * Note: Calling convention is totally non standard for optimal code.
|
||
|
|
33 * This is meant to be used by do_div() from include/asm/div64.h only.
|
||
|
|
34 *
|
||
|
|
35 * Input parameters:
|
||
|
|
36 * xh-xl = dividend (clobbered)
|
||
|
|
37 * r4 = divisor (preserved)
|
||
|
|
38 *
|
||
|
|
39 * Output values:
|
||
|
|
40 * yh-yl = result
|
||
|
|
41 * xh = remainder
|
||
|
|
42 *
|
||
|
|
43 * Clobbered regs: xl, ip
|
||
|
|
44 */
|
||
|
|
45
|
||
|
|
46 ENTRY(__do_div64)
|
||
|
|
47
|
||
|
|
48 @ Test for easy paths first.
|
||
|
|
49 0000 01C054E2 subs ip, r4, #1
|
||
|
|
50 0004 3F00009A bls 9f @ divisor is 0 or 1
|
||
|
|
51 0008 04001CE1 tst ip, r4
|
||
|
|
52 000c 2800000A beq 8f @ divisor is power of 2
|
||
|
|
53
|
||
|
|
54 @ See if we need to handle upper 32-bit result.
|
||
|
|
55 0010 040051E1 cmp xh, r4
|
||
|
|
56 0014 0030A0E3 mov yh, #0
|
||
|
|
57 0018 0C00003A blo 3f
|
||
|
|
58
|
||
|
|
59 @ Align divisor with upper part of dividend.
|
||
|
|
60 @ The aligned divisor is stored in yl preserving the original.
|
||
|
|
61 @ The bit position is stored in ip.
|
||
|
|
62
|
||
|
|
63 #if __LINUX_ARM_ARCH__ >= 5
|
||
|
|
64
|
||
|
|
65 clz yl, r4
|
||
|
|
66 clz ip, xh
|
||
|
|
67 sub yl, yl, ip
|
||
|
|
68 mov ip, #1
|
||
|
|
69 mov ip, ip, lsl yl
|
||
|
|
70 mov yl, r4, lsl yl
|
||
|
|
71
|
||
|
|
72 #else
|
||
|
|
73
|
||
|
|
74 001c 0420A0E1 mov yl, r4
|
||
|
|
75 0020 01C0A0E3 mov ip, #1
|
||
|
|
76 0024 020152E3 1: cmp yl, #0x80000000
|
||
|
|
77 0028 01005231 cmpcc yl, xh
|
||
|
|
78 002c 8220A031 movcc yl, yl, lsl #1
|
||
|
|
79 0030 8CC0A031 movcc ip, ip, lsl #1
|
||
|
|
80 0034 FAFFFF3A bcc 1b
|
||
|
|
81
|
||
|
|
82 #endif
|
||
|
|
83
|
||
|
|
84 @ The division loop for needed upper bit positions.
|
||
|
|
85 @ Break out early if dividend reaches 0.
|
||
|
|
86 0038 020051E1 2: cmp xh, yl
|
||
|
|
87 003c 0C308321 orrcs yh, yh, ip
|
||
|
|
88 0040 02105120 subcss xh, xh, yl
|
||
|
|
89 0044 ACC0B011 movnes ip, ip, lsr #1
|
||
|
|
90 0048 A220A0E1 mov yl, yl, lsr #1
|
||
|
|
91 004c F9FFFF1A bne 2b
|
||
|
|
92
|
||
|
|
93 @ See if we need to handle lower 32-bit result.
|
||
|
|
94 0050 000051E3 3: cmp xh, #0
|
||
|
|
95 0054 0020A0E3 mov yl, #0
|
||
|
|
96 0058 04005001 cmpeq xl, r4
|
||
|
|
97 005c 0010A031 movlo xh, xl
|
||
|
|
98 0060 0EF0A031 movlo pc, lr
|
||
|
|
99
|
||
|
|
100 @ The division loop for lower bit positions.
|
||
|
|
101 @ Here we shift remainer bits leftwards rather than moving the
|
||
|
|
102 @ divisor for comparisons, considering the carry-out bit as well.
|
||
|
|
103 0064 02C1A0E3 mov ip, #0x80000000
|
||
|
|
104 0068 8000B0E1 4: movs xl, xl, lsl #1
|
||
|
|
105 006c 0110B1E0 adcs xh, xh, xh
|
||
|
|
106 0070 0500000A beq 6f
|
||
|
|
107 0074 04005131 cmpcc xh, r4
|
||
|
|
108 0078 0C208221 5: orrcs yl, yl, ip
|
||
|
|
109 007c 04104120 subcs xh, xh, r4
|
||
|
|
110 0080 ACC0B0E1 movs ip, ip, lsr #1
|
||
|
|
111 0084 F7FFFF1A bne 4b
|
||
|
|
112 0088 0EF0A0E1 mov pc, lr
|
||
|
|
113
|
||
|
|
114 @ The top part of remainder became zero. If carry is set
|
||
|
|
115 @ (the 33th bit) this is a false positive so resume the loop.
|
||
|
|
116 @ Otherwise, if lower part is also null then we are done.
|
||
|
|
117 008c F9FFFF2A 6: bcs 5b
|
||
|
|
118 0090 000050E3 cmp xl, #0
|
||
|
|
119 0094 0EF0A001 moveq pc, lr
|
||
|
|
120
|
||
|
|
121 @ We still have remainer bits in the low part. Bring them up.
|
||
|
|
122
|
||
|
|
123 #if __LINUX_ARM_ARCH__ >= 5
|
||
|
|
124
|
||
|
|
125 clz xh, xl @ we know xh is zero here so...
|
||
|
|
126 add xh, xh, #1
|
||
|
|
127 mov xl, xl, lsl xh
|
||
|
|
128 mov ip, ip, lsr xh
|
||
|
|
129
|
||
|
|
130 #else
|
||
|
|
131
|
||
|
|
132 0098 8000B0E1 7: movs xl, xl, lsl #1
|
||
|
|
133 009c ACC0A0E1 mov ip, ip, lsr #1
|
||
|
|
134 00a0 FCFFFF3A bcc 7b
|
||
|
|
135
|
||
|
|
136 #endif
|
||
|
|
137
|
||
|
|
138 @ Current remainder is now 1. It is worthless to compare with
|
||
|
|
139 @ divisor at this point since divisor can not be smaller than 3 here.
|
||
|
|
140 @ If possible, branch for another shift in the division loop.
|
||
|
|
141 @ If no bit position left then we are done.
|
||
|
|
142 00a4 ACC0B0E1 movs ip, ip, lsr #1
|
||
|
|
143 00a8 0110A0E3 mov xh, #1
|
||
|
|
144 00ac EDFFFF1A bne 4b
|
||
|
|
145 00b0 0EF0A0E1 mov pc, lr
|
||
|
|
146
|
||
|
|
147 8: @ Division by a power of 2: determine what that divisor order is
|
||
|
|
148 @ then simply shift values around
|
||
|
|
149
|
||
|
|
150 #if __LINUX_ARM_ARCH__ >= 5
|
||
|
|
151
|
||
|
|
152 clz ip, r4
|
||
|
|
153 rsb ip, ip, #31
|
||
|
|
154
|
||
|
|
155 #else
|
||
|
|
156
|
||
|
|
157 00b4 0420A0E1 mov yl, r4
|
||
|
|
158 00b8 010854E3 cmp r4, #(1 << 16)
|
||
|
|
159 00bc 00C0A0E3 mov ip, #0
|
||
|
|
160 00c0 2228A021 movhs yl, yl, lsr #16
|
||
|
|
161 00c4 10C0A023 movhs ip, #16
|
||
|
|
162
|
||
|
|
163 00c8 010C52E3 cmp yl, #(1 << 8)
|
||
|
|
164 00cc 2224A021 movhs yl, yl, lsr #8
|
||
|
|
165 00d0 08C08C22 addhs ip, ip, #8
|
||
|
|
166
|
||
|
|
167 00d4 100052E3 cmp yl, #(1 << 4)
|
||
|
|
168 00d8 2222A021 movhs yl, yl, lsr #4
|
||
|
|
169 00dc 04C08C22 addhs ip, ip, #4
|
||
|
|
170
|
||
|
|
171 00e0 040052E3 cmp yl, #(1 << 2)
|
||
|
|
172 00e4 03C08C82 addhi ip, ip, #3
|
||
|
|
173 00e8 A2C08C90 addls ip, ip, yl, lsr #1
|
||
|
|
174
|
||
|
|
175 #endif
|
||
|
|
176
|
||
|
|
177 00ec 313CA0E1 mov yh, xh, lsr ip
|
||
|
|
178 00f0 302CA0E1 mov yl, xl, lsr ip
|
||
|
|
179 00f4 20C06CE2 rsb ip, ip, #32
|
||
|
|
180 00f8 112C82E1 orr yl, yl, xh, lsl ip
|
||
|
|
181 00fc 101CA0E1 mov xh, xl, lsl ip
|
||
|
|
182 0100 311CA0E1 mov xh, xh, lsr ip
|
||
|
|
183 0104 0EF0A0E1 mov pc, lr
|
||
|
|
184
|
||
|
|
185 @ eq -> division by 1: obvious enough...
|
||
|
|
186 0108 0020A001 9: moveq yl, xl
|
||
|
|
187 010c 0130A001 moveq yh, xh
|
||
|
|
188 0110 0010A003 moveq xh, #0
|
||
|
|
189 0114 0EF0A001 moveq pc, lr
|
||
|
|
190
|
||
|
|
191 @ Division by 0:
|
||
|
|
192 0118 08E02DE5 str lr, [sp, #-8]!
|
||
|
|
193 011c FEFFFFEB bl __div0
|
||
|
|
194
|
||
|
|
195 @ as wrong as it could be...
|
||
|
|
196 0120 0020A0E3 mov yl, #0
|
||
|
|
197 0124 0030A0E3 mov yh, #0
|
||
|
|
198 0128 0010A0E3 mov xh, #0
|
||
|
|
199 012c 08F09DE4 ldr pc, [sp], #8
|
||
|
|
DEFINED SYMBOLS
|
||
|
|
lib/div64.S:46 .text:00000000 __do_div64
|
||
|
|
lib/div64.S:46 .text:00000000 $a
|
||
|
|
.debug_aranges:0000000c $d
|
||
|
|
|
||
|
|
UNDEFINED SYMBOLS
|
||
|
|
__div0
|