1 | ; $Id: x86-aulldvrm-core.mac 98509 2023-02-08 16:01:15Z vboxsync $
|
---|
2 | ;; @file
|
---|
3 | ; IPRT - Visual C++ Compiler - unsigned 64-bit division support, x86.
|
---|
4 | ;
|
---|
5 |
|
---|
6 | ;
|
---|
7 | ; Copyright (C) 2023 Oracle and/or its affiliates.
|
---|
8 | ;
|
---|
9 | ; This file is part of VirtualBox base platform packages, as
|
---|
10 | ; available from https://www.virtualbox.org.
|
---|
11 | ;
|
---|
12 | ; This program is free software; you can redistribute it and/or
|
---|
13 | ; modify it under the terms of the GNU General Public License
|
---|
14 | ; as published by the Free Software Foundation, in version 3 of the
|
---|
15 | ; License.
|
---|
16 | ;
|
---|
17 | ; This program is distributed in the hope that it will be useful, but
|
---|
18 | ; WITHOUT ANY WARRANTY; without even the implied warranty of
|
---|
19 | ; MERCHANTABILITY or FITNESS FOR A PARTICULAR PURPOSE. See the GNU
|
---|
20 | ; General Public License for more details.
|
---|
21 | ;
|
---|
22 | ; You should have received a copy of the GNU General Public License
|
---|
23 | ; along with this program; if not, see <https://www.gnu.org/licenses>.
|
---|
24 | ;
|
---|
25 | ; The contents of this file may alternatively be used under the terms
|
---|
26 | ; of the Common Development and Distribution License Version 1.0
|
---|
27 | ; (CDDL), a copy of it is provided in the "COPYING.CDDL" file included
|
---|
28 | ; in the VirtualBox distribution, in which case the provisions of the
|
---|
29 | ; CDDL are applicable instead of those of the GPL.
|
---|
30 | ;
|
---|
31 | ; You may elect to license modified versions of this file under the
|
---|
32 | ; terms and conditions of either the GPL or the CDDL or both.
|
---|
33 | ;
|
---|
34 | ; SPDX-License-Identifier: GPL-3.0-only OR CDDL-1.0
|
---|
35 | ;
|
---|
36 |
|
---|
37 |
|
---|
38 | ;*********************************************************************************************************************************
|
---|
39 | ;* Header Files *
|
---|
40 | ;*********************************************************************************************************************************
|
---|
41 | %include "iprt/asmdefs.mac"
|
---|
42 |
|
---|
43 |
|
---|
44 | ;;
|
---|
45 | ; 64-bit unsigned division.
|
---|
46 | ;
|
---|
47 | ; @param 1 Function name.
|
---|
48 | ; @param 2 dvrm=0, div=1, rem=1
|
---|
49 | ;
|
---|
50 | %macro rtVccUnsignedDivision 2
|
---|
51 |
|
---|
52 | ;;
|
---|
53 | ; Division of unsigned 64-bit values, returning both quotient and remainder.
|
---|
54 | ;
|
---|
55 | ; @returns Quotient in edx:eax, remainder in ebx:ecx.
|
---|
56 | ; @param [ebp+08h] Dividend (64-bit)
|
---|
57 | ; @param [ebp+10h] Divisor (64-bit)
|
---|
58 | ;
|
---|
59 | ; @note The remainder registers are swapped compared to Watcom's I8D and U8D.
|
---|
60 | ;
|
---|
61 | BEGINPROC_RAW %1
|
---|
62 | %if 1
|
---|
63 | ;
|
---|
64 | ; This code is duplicated in in x86-aulldiv.asm and x86-aulldvrm.asm.
|
---|
65 | ; Apply fixes everywhere.
|
---|
66 | ;
|
---|
67 | push ebp
|
---|
68 | mov ebp, esp
|
---|
69 | %if %2 > 0
|
---|
70 | push ebx
|
---|
71 | %endif
|
---|
72 |
|
---|
73 | %define DIVIDEND_LO ebp + 08h
|
---|
74 | %define DIVIDEND_HI ebp + 0ch
|
---|
75 | %define DIVISOR_LO ebp + 10h
|
---|
76 | %define DIVISOR_HI ebp + 14h
|
---|
77 |
|
---|
78 | ;
|
---|
79 | ; If the divisor is only 32-bit wide as we can do a two-step division on 32-bit units.
|
---|
80 | ;
|
---|
81 | mov ebx, [DIVISOR_HI]
|
---|
82 | or ebx, ebx
|
---|
83 | jnz .full_64_bit_divisor
|
---|
84 |
|
---|
85 | ; step 1: dividend_hi / divisor
|
---|
86 | mov ebx, [DIVISOR_LO]
|
---|
87 | mov eax, [DIVIDEND_HI]
|
---|
88 | xor edx, edx
|
---|
89 | div ebx
|
---|
90 | %if %2 <= 1
|
---|
91 | mov ecx, eax ; high quotient bits.
|
---|
92 | %endif
|
---|
93 |
|
---|
94 | ; step 2: (dividend_lo + step_1_remainder) / divisor
|
---|
95 | mov eax, [DIVIDEND_LO] ; edx contains the remainder from the first step.
|
---|
96 | div ebx ; -> eax = low quotient, edx = remainder.
|
---|
97 |
|
---|
98 | %if %2 <= 1
|
---|
99 | xchg edx, ecx ; ecx = (low) remainder, edx = saved high quotient from step 1
|
---|
100 | %if %2 == 0
|
---|
101 | xor ebx, ebx ; ebx = high remainder is zero, since divisor is 32-bit.
|
---|
102 | %endif
|
---|
103 | %else
|
---|
104 | mov eax, edx
|
---|
105 | xor edx, edx ; edx:eax = remainder
|
---|
106 | %endif
|
---|
107 |
|
---|
108 | %if %2 > 0
|
---|
109 | pop ebx
|
---|
110 | %endif
|
---|
111 | leave
|
---|
112 | ret 10h
|
---|
113 |
|
---|
114 | ;
|
---|
115 | ; The divisor is larger than 32 bits.
|
---|
116 | ;
|
---|
117 | ; We can approximate the quotient by reducing the divisor to 32 bits
|
---|
118 | ; (reducing the dividend accordingly) and perform a 32-bit division.
|
---|
119 | ; The result will be at most one off.
|
---|
120 | ;
|
---|
121 | ; The remainder has to be calculated using multiplication and
|
---|
122 | ; subtraction.
|
---|
123 | ;
|
---|
124 | .full_64_bit_divisor:
|
---|
125 | push edi
|
---|
126 |
|
---|
127 | ; Find the shift count needed to reduce the divisor to 32-bit.
|
---|
128 | bsr ecx, ebx
|
---|
129 | inc cl
|
---|
130 | test cl, ~31
|
---|
131 | jnz .shift_32
|
---|
132 |
|
---|
133 | ; Shift the divisor into edi.
|
---|
134 | mov edi, [DIVISOR_LO]
|
---|
135 | shrd edi, ebx, cl ; edi = reduced divisor
|
---|
136 |
|
---|
137 | ; Shift the dividend into edx:eax.
|
---|
138 | mov eax, [DIVIDEND_LO]
|
---|
139 | mov edx, [DIVIDEND_HI]
|
---|
140 | shrd eax, edx, cl
|
---|
141 | shr edx, cl
|
---|
142 | jmp .shifted
|
---|
143 |
|
---|
144 | .shift_32: ; simplified version.
|
---|
145 | mov edi, ebx
|
---|
146 | mov eax, [DIVIDEND_HI]
|
---|
147 | xor edx, edx
|
---|
148 | .shifted:
|
---|
149 |
|
---|
150 | ; Divide and save the approximate quotient (Qapprox) in edi.
|
---|
151 | div edi
|
---|
152 | mov edi, eax ; edi = Qapprox
|
---|
153 |
|
---|
154 | ; Now multiply Qapprox with the divisor.
|
---|
155 | mul dword [DIVISOR_HI]
|
---|
156 | mov ecx, eax ; temporary storage
|
---|
157 | mov eax, [DIVISOR_LO]
|
---|
158 | mul edi
|
---|
159 | add edx, ecx ; edx:eax = QapproxDividend = Qapprox * divisor
|
---|
160 |
|
---|
161 | ; Preload the dividend into ebx:ecx for remainder calculation and for adjusting Qapprox.
|
---|
162 | mov ecx, [DIVIDEND_LO]
|
---|
163 | mov ebx, [DIVIDEND_HI]
|
---|
164 |
|
---|
165 | ; If carry is set, the result overflowed 64 bits, so the quotient must be too large.
|
---|
166 | jc .quotient_is_one_above_and_calc_remainder
|
---|
167 |
|
---|
168 | ; Calculate the remainder, if this overflows (CF) it means Qapprox is
|
---|
169 | ; one above and we need to reduce it and the adjust the remainder.
|
---|
170 | sub ecx, eax
|
---|
171 | sbb ebx, edx
|
---|
172 | jc .quotient_is_one_above
|
---|
173 | .done:
|
---|
174 | %if %2 <= 1
|
---|
175 | mov eax, edi
|
---|
176 | xor edx, edx ; edx:eax = quotient
|
---|
177 | %else
|
---|
178 | mov eax, ecx
|
---|
179 | mov edx, ebx ; edx:eax = remainder
|
---|
180 | %endif
|
---|
181 |
|
---|
182 | pop edi
|
---|
183 | %if %2 > 0
|
---|
184 | pop ebx
|
---|
185 | %endif
|
---|
186 | leave
|
---|
187 | ret 10h
|
---|
188 |
|
---|
189 | .quotient_is_one_above_and_calc_remainder:
|
---|
190 | %if %2 != 1
|
---|
191 | sub ecx, eax
|
---|
192 | sbb ebx, edx
|
---|
193 | %endif
|
---|
194 | .quotient_is_one_above:
|
---|
195 | %if %2 != 1
|
---|
196 | add ecx, [DIVISOR_LO]
|
---|
197 | adc ebx, [DIVISOR_HI]
|
---|
198 | %endif
|
---|
199 | %if %2 != 2
|
---|
200 | dec edi
|
---|
201 | %endif
|
---|
202 | jmp .done
|
---|
203 |
|
---|
204 | %else
|
---|
205 | ;
|
---|
206 | ; Fall back on a rather slow C implementation.
|
---|
207 | ;
|
---|
208 | push ebp
|
---|
209 | mov ebp, esp
|
---|
210 |
|
---|
211 | ; Call RTVccUInt64Div(RTUINT64U const *paDividendDivisor, RTUINT64U *paQuotientRemainder)
|
---|
212 | sub esp, 10h ; space for quotient and remainder.
|
---|
213 | mov edx, esp
|
---|
214 | push edx
|
---|
215 | lea ecx, [ebp + 8]
|
---|
216 | push ecx
|
---|
217 | extern NAME(RTVccUInt64Div)
|
---|
218 | call NAME(RTVccUInt64Div)
|
---|
219 |
|
---|
220 | ; Load the result.
|
---|
221 | %if %2 != 2
|
---|
222 | mov eax, [ebp - 10h]
|
---|
223 | mov edx, [ebp - 10h + 4]
|
---|
224 |
|
---|
225 | %if %2 != 1
|
---|
226 | mov ecx, [ebp - 08h]
|
---|
227 | mov ebx, [ebp - 08h + 4]
|
---|
228 | %endif
|
---|
229 | %else
|
---|
230 | mov eax, [ebp - 08h]
|
---|
231 | mov edx, [ebp - 08h + 4]
|
---|
232 | %endif
|
---|
233 | leave
|
---|
234 | ret 10h
|
---|
235 | %endif
|
---|
236 | ENDPROC_RAW %1
|
---|
237 | %endm
|
---|