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
163
164
165
166
167
168
169
170
171
172
173
174
175
176
177
178
179
180
181
182
183
184
185
186
187
188
189
190
191
192
193
194
195
196
197
198
199
200
201
202
203
204
205
206
207
208
|
dnl AMD K7 mpn_divexact_1 -- mpn by limb exact division.
dnl Copyright 2001, 2002, 2004, 2007 Free Software Foundation, Inc.
dnl This file is part of the GNU MP Library.
dnl
dnl The GNU MP Library is free software; you can redistribute it and/or modify
dnl it under the terms of either:
dnl
dnl * the GNU Lesser General Public License as published by the Free
dnl Software Foundation; either version 3 of the License, or (at your
dnl option) any later version.
dnl
dnl or
dnl
dnl * the GNU General Public License as published by the Free Software
dnl Foundation; either version 2 of the License, or (at your option) any
dnl later version.
dnl
dnl or both in parallel, as here.
dnl
dnl The GNU MP Library is distributed in the hope that it will be useful, but
dnl WITHOUT ANY WARRANTY; without even the implied warranty of MERCHANTABILITY
dnl or FITNESS FOR A PARTICULAR PURPOSE. See the GNU General Public License
dnl for more details.
dnl
dnl You should have received copies of the GNU General Public License and the
dnl GNU Lesser General Public License along with the GNU MP Library. If not,
dnl see https://www.gnu.org/licenses/.
include(`../config.m4')
C cycles/limb
C Athlon: 11.0
C Hammer: 9.0
C void mpn_divexact_1 (mp_ptr dst, mp_srcptr src, mp_size_t size,
C mp_limb_t divisor);
C
C The dependent chain is mul+imul+sub for 11 cycles and that speed is
C achieved with no special effort. The load and shrld latencies are hidden
C by out of order execution.
C
C It's a touch faster on size==1 to use the mul-by-inverse than divl.
defframe(PARAM_DIVISOR,16)
defframe(PARAM_SIZE, 12)
defframe(PARAM_SRC, 8)
defframe(PARAM_DST, 4)
defframe(SAVE_EBX, -4)
defframe(SAVE_ESI, -8)
defframe(SAVE_EDI, -12)
defframe(SAVE_EBP, -16)
defframe(VAR_INVERSE, -20)
defframe(VAR_DST_END, -24)
deflit(STACK_SPACE, 24)
TEXT
ALIGN(16)
PROLOGUE(mpn_divexact_1)
deflit(`FRAME',0)
movl PARAM_DIVISOR, %eax
subl $STACK_SPACE, %esp deflit(`FRAME',STACK_SPACE)
movl $-1, %ecx C shift count
movl %ebp, SAVE_EBP
movl PARAM_SIZE, %ebp
movl %esi, SAVE_ESI
movl %edi, SAVE_EDI
C If there's usually only one or two trailing zero bits then this
C should be faster than bsfl.
L(strip_twos):
incl %ecx
shrl %eax
jnc L(strip_twos)
movl %ebx, SAVE_EBX
leal 1(%eax,%eax), %ebx C d without twos
andl $127, %eax C d/2, 7 bits
ifdef(`PIC',`
LEA( binvert_limb_table, %edx)
movzbl (%eax,%edx), %eax C inv 8 bits
',`
movzbl binvert_limb_table(%eax), %eax C inv 8 bits
')
leal (%eax,%eax), %edx C 2*inv
movl %ebx, PARAM_DIVISOR C d without twos
imull %eax, %eax C inv*inv
movl PARAM_SRC, %esi
movl PARAM_DST, %edi
imull %ebx, %eax C inv*inv*d
subl %eax, %edx C inv = 2*inv - inv*inv*d
leal (%edx,%edx), %eax C 2*inv
imull %edx, %edx C inv*inv
leal (%esi,%ebp,4), %esi C src end
leal (%edi,%ebp,4), %edi C dst end
negl %ebp C -size
imull %ebx, %edx C inv*inv*d
subl %edx, %eax C inv = 2*inv - inv*inv*d
ASSERT(e,` C expect d*inv == 1 mod 2^GMP_LIMB_BITS
pushl %eax FRAME_pushl()
imull PARAM_DIVISOR, %eax
cmpl $1, %eax
popl %eax FRAME_popl()')
movl %eax, VAR_INVERSE
movl (%esi,%ebp,4), %eax C src[0]
incl %ebp
jz L(one)
movl (%esi,%ebp,4), %edx C src[1]
shrdl( %cl, %edx, %eax)
movl %edi, VAR_DST_END
xorl %ebx, %ebx
jmp L(entry)
ALIGN(8)
L(top):
C eax q
C ebx carry bit, 0 or 1
C ecx shift
C edx
C esi src end
C edi dst end
C ebp counter, limbs, negative
mull PARAM_DIVISOR C carry limb in edx
movl -4(%esi,%ebp,4), %eax
movl (%esi,%ebp,4), %edi
shrdl( %cl, %edi, %eax)
subl %ebx, %eax C apply carry bit
setc %bl
movl VAR_DST_END, %edi
subl %edx, %eax C apply carry limb
adcl $0, %ebx
L(entry):
imull VAR_INVERSE, %eax
movl %eax, -4(%edi,%ebp,4)
incl %ebp
jnz L(top)
mull PARAM_DIVISOR C carry limb in edx
movl -4(%esi), %eax C src high limb
shrl %cl, %eax
movl SAVE_ESI, %esi
subl %ebx, %eax C apply carry bit
movl SAVE_EBX, %ebx
movl SAVE_EBP, %ebp
subl %edx, %eax C apply carry limb
imull VAR_INVERSE, %eax
movl %eax, -4(%edi)
movl SAVE_EDI, %edi
addl $STACK_SPACE, %esp
ret
L(one):
shrl %cl, %eax
movl SAVE_ESI, %esi
movl SAVE_EBX, %ebx
imull VAR_INVERSE, %eax
movl SAVE_EBP, %ebp
movl %eax, -4(%edi)
movl SAVE_EDI, %edi
addl $STACK_SPACE, %esp
ret
EPILOGUE()
ASM_END()
|
