1 | // Part of the LLVM Project, under the Apache License v2.0 with LLVM Exceptions. |
2 | // See https://llvm.org/LICENSE.txt for license information. |
3 | // SPDX-License-Identifier: Apache-2.0 WITH LLVM-exception |
4 | |
5 | #include "../assembly.h" |
6 | |
7 | // di_int __divdi3(di_int a, di_int b); |
8 | |
9 | // result = a / b. |
10 | // both inputs and the output are 64-bit signed integers. |
11 | // This will do whatever the underlying hardware is set to do on division by zero. |
12 | // No other exceptions are generated, as the divide cannot overflow. |
13 | // |
14 | // This is targeted at 32-bit x86 *only*, as this can be done directly in hardware |
15 | // on x86_64. The performance goal is ~40 cycles per divide, which is faster than |
16 | // currently possible via simulation of integer divides on the x87 unit. |
17 | // |
18 | // Stephen Canon, December 2008 |
19 | |
20 | #ifdef __i386__ |
21 | |
22 | .text |
23 | .balign 4 |
24 | DEFINE_COMPILERRT_FUNCTION(__divdi3) |
25 | |
26 | // This is currently implemented by wrapping the unsigned divide up in an absolute |
27 | // value, then restoring the correct sign at the end of the computation. This could |
28 | // certainly be improved upon. |
29 | |
30 | pushl %esi |
31 | movl 20(%esp), %edx // high word of b |
32 | movl 16(%esp), %eax // low word of b |
33 | movl %edx, %ecx |
34 | sarl $31, %ecx // (b < 0) ? -1 : 0 |
35 | xorl %ecx, %eax |
36 | xorl %ecx, %edx // EDX:EAX = (b < 0) ? not(b) : b |
37 | subl %ecx, %eax |
38 | sbbl %ecx, %edx // EDX:EAX = abs(b) |
39 | movl %edx, 20(%esp) |
40 | movl %eax, 16(%esp) // store abs(b) back to stack |
41 | movl %ecx, %esi // set aside sign of b |
42 | |
43 | movl 12(%esp), %edx // high word of b |
44 | movl 8(%esp), %eax // low word of b |
45 | movl %edx, %ecx |
46 | sarl $31, %ecx // (a < 0) ? -1 : 0 |
47 | xorl %ecx, %eax |
48 | xorl %ecx, %edx // EDX:EAX = (a < 0) ? not(a) : a |
49 | subl %ecx, %eax |
50 | sbbl %ecx, %edx // EDX:EAX = abs(a) |
51 | movl %edx, 12(%esp) |
52 | movl %eax, 8(%esp) // store abs(a) back to stack |
53 | xorl %ecx, %esi // sign of result = (sign of a) ^ (sign of b) |
54 | |
55 | pushl %ebx |
56 | movl 24(%esp), %ebx // Find the index i of the leading bit in b. |
57 | bsrl %ebx, %ecx // If the high word of b is zero, jump to |
58 | jz 9f // the code to handle that special case [9]. |
59 | |
60 | // High word of b is known to be non-zero on this branch |
61 | |
62 | movl 20(%esp), %eax // Construct bhi, containing bits [1+i:32+i] of b |
63 | |
64 | shrl %cl, %eax // Practically, this means that bhi is given by: |
65 | shrl %eax // |
66 | notl %ecx // bhi = (high word of b) << (31 - i) | |
67 | shll %cl, %ebx // (low word of b) >> (1 + i) |
68 | orl %eax, %ebx // |
69 | movl 16(%esp), %edx // Load the high and low words of a, and jump |
70 | movl 12(%esp), %eax // to [1] if the high word is larger than bhi |
71 | cmpl %ebx, %edx // to avoid overflowing the upcoming divide. |
72 | jae 1f |
73 | |
74 | // High word of a is greater than or equal to (b >> (1 + i)) on this branch |
75 | |
76 | divl %ebx // eax <-- qs, edx <-- r such that ahi:alo = bs*qs + r |
77 | |
78 | pushl %edi |
79 | notl %ecx |
80 | shrl %eax |
81 | shrl %cl, %eax // q = qs >> (1 + i) |
82 | movl %eax, %edi |
83 | mull 24(%esp) // q*blo |
84 | movl 16(%esp), %ebx |
85 | movl 20(%esp), %ecx // ECX:EBX = a |
86 | subl %eax, %ebx |
87 | sbbl %edx, %ecx // ECX:EBX = a - q*blo |
88 | movl 28(%esp), %eax |
89 | imull %edi, %eax // q*bhi |
90 | subl %eax, %ecx // ECX:EBX = a - q*b |
91 | sbbl $0, %edi // decrement q if remainder is negative |
92 | xorl %edx, %edx |
93 | movl %edi, %eax |
94 | |
95 | addl %esi, %eax // Restore correct sign to result |
96 | adcl %esi, %edx |
97 | xorl %esi, %eax |
98 | xorl %esi, %edx |
99 | popl %edi // Restore callee-save registers |
100 | popl %ebx |
101 | popl %esi |
102 | retl // Return |
103 | |
104 | |
105 | 1: // High word of a is greater than or equal to (b >> (1 + i)) on this branch |
106 | |
107 | subl %ebx, %edx // subtract bhi from ahi so that divide will not |
108 | divl %ebx // overflow, and find q and r such that |
109 | // |
110 | // ahi:alo = (1:q)*bhi + r |
111 | // |
112 | // Note that q is a number in (31-i).(1+i) |
113 | // fix point. |
114 | |
115 | pushl %edi |
116 | notl %ecx |
117 | shrl %eax |
118 | orl $0x80000000, %eax |
119 | shrl %cl, %eax // q = (1:qs) >> (1 + i) |
120 | movl %eax, %edi |
121 | mull 24(%esp) // q*blo |
122 | movl 16(%esp), %ebx |
123 | movl 20(%esp), %ecx // ECX:EBX = a |
124 | subl %eax, %ebx |
125 | sbbl %edx, %ecx // ECX:EBX = a - q*blo |
126 | movl 28(%esp), %eax |
127 | imull %edi, %eax // q*bhi |
128 | subl %eax, %ecx // ECX:EBX = a - q*b |
129 | sbbl $0, %edi // decrement q if remainder is negative |
130 | xorl %edx, %edx |
131 | movl %edi, %eax |
132 | |
133 | addl %esi, %eax // Restore correct sign to result |
134 | adcl %esi, %edx |
135 | xorl %esi, %eax |
136 | xorl %esi, %edx |
137 | popl %edi // Restore callee-save registers |
138 | popl %ebx |
139 | popl %esi |
140 | retl // Return |
141 | |
142 | |
143 | 9: // High word of b is zero on this branch |
144 | |
145 | movl 16(%esp), %eax // Find qhi and rhi such that |
146 | movl 20(%esp), %ecx // |
147 | xorl %edx, %edx // ahi = qhi*b + rhi with 0 ≤ rhi < b |
148 | divl %ecx // |
149 | movl %eax, %ebx // |
150 | movl 12(%esp), %eax // Find qlo such that |
151 | divl %ecx // |
152 | movl %ebx, %edx // rhi:alo = qlo*b + rlo with 0 ≤ rlo < b |
153 | |
154 | addl %esi, %eax // Restore correct sign to result |
155 | adcl %esi, %edx |
156 | xorl %esi, %eax |
157 | xorl %esi, %edx |
158 | popl %ebx // Restore callee-save registers |
159 | popl %esi |
160 | retl // Return |
161 | END_COMPILERRT_FUNCTION(__divdi3) |
162 | |
163 | #endif // __i386__ |
164 | |
165 | NO_EXEC_STACK_DIRECTIVE |
166 | |
167 | |