blob: d721b6299b1bd37f30489958a0f01dec4dc29926 [file] [log] [blame]
Stephen Canon5abb5c12011-03-18 16:35:02 +00001/*===-- udivsi3.S - 32-bit unsigned integer divide ------------------------===//
2 *
3 * The LLVM Compiler Infrastructure
4 *
5 * This file is dual licensed under the MIT and the University of Illinois Open
6 * Source Licenses. See LICENSE.TXT for details.
7 *
8 *===----------------------------------------------------------------------===//
9 *
10 * This file implements the __udivsi3 (32-bit unsigned integer divide)
11 * function for the ARM architecture. A naive digit-by-digit computation is
12 * employed for simplicity.
13 *
14 *===----------------------------------------------------------------------===*/
15
16#include "../assembly.h"
17
18#define ESTABLISH_FRAME \
19 push {r7, lr} ;\
20 mov r7, sp
21#define CLEAR_FRAME_AND_RETURN \
22 pop {r7, pc}
23
24#define a r0
25#define b r1
26#define r r2
27#define i r3
28#define q ip
29#define one lr
30
31.syntax unified
32.align 3
33DEFINE_COMPILERRT_FUNCTION(__udivsi3)
34// We use a simple digit by digit algorithm; before we get into the actual
35// divide loop, we must calculate the left-shift amount necessary to align
36// the MSB of the divisor with that of the dividend (If this shift is
37// negative, then the result is zero, and we early out). We also conjure a
38// bit mask of 1 to use in constructing the quotient, and initialize the
39// quotient to zero.
40 ESTABLISH_FRAME
41 clz r2, a
42 tst b, b // detect divide-by-zero
43 clz r3, b
44 mov q, #0
45 beq L_return // return 0 if b is zero.
46 mov one, #1
47 subs i, r3, r2
48 blt L_return // return 0 if MSB(a) < MSB(b)
49
50L_mainLoop:
51// This loop basically implements the following:
52//
53// do {
54// if (a >= b << i) {
55// a -= b << i;
56// q |= 1 << i;
57// if (a == 0) break;
58// }
59// } while (--i)
60//
61// Note that this does not perform the final iteration (i == 0); by doing it
62// this way, we can merge the two branches which is a substantial win for
63// such a tight loop on current ARM architectures.
64 subs r, a, b, lsl i
65 orrhs q, q,one, lsl i
66 movhs a, r
67 subsne i, i, #1
68 bhi L_mainLoop
69
70// Do the final test subtraction and update of quotient (i == 0), as it is
71// not performed in the main loop.
72 subs r, a, b
73 orrhs q, #1
74
75L_return:
76// Move the quotient to r0 and return.
77 mov r0, q
78 CLEAR_FRAME_AND_RETURN