aboutsummaryrefslogtreecommitdiffstats
path: root/contrib/libs/cxxsupp/builtins/i386/divdi3.S
blob: 2cb0ddd4c29f1099b083610badecdb7f8b7476f7 (plain) (blame)
1
2
3
4
5
6
7
8
9
10
11
12
13
14
15
16
17
18
19
20
21
22
23
24
25
26
27
28
29
30
31
32
33
34
35
36
37
38
39
40
41
42
43
44
45
46
47
48
49
50
51
52
53
54
55
56
57
58
59
60
61
62
63
64
65
66
67
68
69
70
71
72
73
74
75
76
77
78
79
80
81
82
83
84
85
86
87
88
89
90
91
92
93
94
95
96
97
98
99
100
101
102
103
104
105
106
107
108
109
110
111
112
113
114
115
116
117
118
119
120
121
122
123
124
125
126
127
128
129
130
131
132
133
134
135
136
137
138
139
140
141
142
143
144
145
146
147
148
149
150
151
152
153
154
155
156
157
158
159
160
161
162
// This file is dual licensed under the MIT and the University of Illinois Open
// Source Licenses. See LICENSE.TXT for details.

#include "../assembly.h"

// di_int __divdi3(di_int a, di_int b);

// result = a / b.
// both inputs and the output are 64-bit signed integers.
// This will do whatever the underlying hardware is set to do on division by zero.
// No other exceptions are generated, as the divide cannot overflow.
//
// This is targeted at 32-bit x86 *only*, as this can be done directly in hardware
// on x86_64.  The performance goal is ~40 cycles per divide, which is faster than
// currently possible via simulation of integer divides on the x87 unit.
//
// Stephen Canon, December 2008

#ifdef __i386__

.text
.balign 4
DEFINE_COMPILERRT_FUNCTION(__divdi3)

/* This is currently implemented by wrapping the unsigned divide up in an absolute
   value, then restoring the correct sign at the end of the computation.  This could
   certainly be improved upon. */

	pushl		%esi
	movl	 20(%esp),			%edx	// high word of b
	movl	 16(%esp),			%eax	// low word of b
	movl		%edx,			%ecx
	sarl		$31,			%ecx	// (b < 0) ? -1 : 0
	xorl		%ecx,			%eax
	xorl		%ecx,			%edx	// EDX:EAX = (b < 0) ? not(b) : b
	subl		%ecx,			%eax
	sbbl		%ecx,			%edx	// EDX:EAX = abs(b)
	movl		%edx,		 20(%esp)
	movl		%eax,		 16(%esp)	// store abs(b) back to stack
	movl		%ecx,			%esi	// set aside sign of b
	
	movl	 12(%esp),			%edx	// high word of b
	movl	  8(%esp),			%eax	// low word of b
	movl		%edx,			%ecx
	sarl		$31,			%ecx	// (a < 0) ? -1 : 0
	xorl		%ecx,			%eax
	xorl		%ecx,			%edx	// EDX:EAX = (a < 0) ? not(a) : a
	subl		%ecx,			%eax
	sbbl		%ecx,			%edx	// EDX:EAX = abs(a)
	movl		%edx,		 12(%esp)
	movl		%eax,		  8(%esp)	// store abs(a) back to stack
	xorl		%ecx,			%esi	// sign of result = (sign of a) ^ (sign of b)

	pushl		%ebx
	movl	 24(%esp),			%ebx	// Find the index i of the leading bit in b.
	bsrl		%ebx,			%ecx	// If the high word of b is zero, jump to
	jz			9f						// the code to handle that special case [9].
	
	/* High word of b is known to be non-zero on this branch */
	
	movl	 20(%esp),			%eax	// Construct bhi, containing bits [1+i:32+i] of b
	
	shrl		%cl,			%eax	// Practically, this means that bhi is given by:
	shrl		%eax					//
	notl		%ecx					//		bhi = (high word of b) << (31 - i) |
	shll		%cl,			%ebx	//			  (low word of b) >> (1 + i)
	orl			%eax,			%ebx	//
	movl	 16(%esp),			%edx	// Load the high and low words of a, and jump
	movl	 12(%esp),			%eax	// to [1] if the high word is larger than bhi
	cmpl		%ebx,			%edx	// to avoid overflowing the upcoming divide.
	jae			1f						
		
	/* High word of a is greater than or equal to (b >> (1 + i)) on this branch */
	
	divl		%ebx					// eax <-- qs, edx <-- r such that ahi:alo = bs*qs + r

	pushl		%edi
	notl		%ecx
	shrl		%eax
	shrl		%cl,			%eax	// q = qs >> (1 + i)
	movl		%eax,			%edi
	mull	 24(%esp)					// q*blo
	movl	 16(%esp),			%ebx
	movl	 20(%esp),			%ecx	// ECX:EBX = a
	subl		%eax,			%ebx
	sbbl		%edx,			%ecx	// ECX:EBX = a - q*blo
	movl	 28(%esp),			%eax
	imull		%edi,			%eax	// q*bhi
	subl		%eax,			%ecx	// ECX:EBX = a - q*b
	sbbl		$0,				%edi	// decrement q if remainder is negative
	xorl		%edx,			%edx
	movl		%edi,			%eax
	
	addl		%esi,			%eax	// Restore correct sign to result
	adcl		%esi,			%edx
	xorl		%esi,			%eax
	xorl		%esi,			%edx
	popl		%edi					// Restore callee-save registers
	popl		%ebx
	popl		%esi
	retl								// Return


1:	/* High word of a is greater than or equal to (b >> (1 + i)) on this branch */
	 
	subl		%ebx,			%edx	// subtract bhi from ahi so that divide will not
	divl		%ebx					// overflow, and find q and r such that
										//
										//		ahi:alo = (1:q)*bhi + r
										//
										// Note that q is a number in (31-i).(1+i)
										// fix point.

	pushl		%edi
	notl		%ecx
	shrl		%eax
	orl			$0x80000000,	%eax
	shrl		%cl,			%eax	// q = (1:qs) >> (1 + i)
	movl		%eax,			%edi
	mull	 24(%esp)					// q*blo
	movl	 16(%esp),			%ebx
	movl	 20(%esp),			%ecx	// ECX:EBX = a
	subl		%eax,			%ebx
	sbbl		%edx,			%ecx	// ECX:EBX = a - q*blo
	movl	 28(%esp),			%eax
	imull		%edi,			%eax	// q*bhi
	subl		%eax,			%ecx	// ECX:EBX = a - q*b
	sbbl		$0,				%edi	// decrement q if remainder is negative
	xorl		%edx,			%edx
	movl		%edi,			%eax
	
	addl		%esi,			%eax	// Restore correct sign to result
	adcl		%esi,			%edx
	xorl		%esi,			%eax
	xorl		%esi,			%edx
	popl		%edi					// Restore callee-save registers
	popl		%ebx
	popl		%esi
	retl								// Return

	
9:	/* High word of b is zero on this branch */

	movl	 16(%esp),			%eax	// Find qhi and rhi such that
	movl	 20(%esp),			%ecx	//
	xorl		%edx,			%edx	//		ahi = qhi*b + rhi	with	0 ≤ rhi < b
	divl		%ecx					//
	movl		%eax,			%ebx	//
	movl	 12(%esp),			%eax	// Find qlo such that
	divl		%ecx					//
	movl		%ebx,			%edx	//		rhi:alo = qlo*b + rlo  with 0 ≤ rlo < b
	
	addl		%esi,			%eax	// Restore correct sign to result
	adcl		%esi,			%edx
	xorl		%esi,			%eax
	xorl		%esi,			%edx
	popl		%ebx					// Restore callee-save registers
	popl		%esi
	retl								// Return
END_COMPILERRT_FUNCTION(__divdi3)

#endif // __i386__