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
|
dnl Intel P6 mpn_mod_34lsub1 -- remainder modulo 2^24-1.
dnl Copyright 2000-2002, 2004 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 P6: 2.0 cycles/limb
C TODO
C Experiments with more unrolling indicate that 1.5 c/l is possible on P6-13
C with the current carry handling scheme.
C mp_limb_t mpn_mod_34lsub1 (mp_srcptr src, mp_size_t size)
C
C Groups of three limbs are handled, with carry bits from 0mod3 into 1mod3
C into 2mod3, but at that point going into a separate carries total so we
C don't keep the carry flag live across the loop control. Avoiding decl
C lets us get to 2.0 c/l, as compared to the generic x86 code at 3.66.
C
defframe(PARAM_SIZE, 8)
defframe(PARAM_SRC, 4)
dnl re-use parameter space
define(SAVE_EBX, `PARAM_SIZE')
define(SAVE_ESI, `PARAM_SRC')
TEXT
ALIGN(16)
PROLOGUE(mpn_mod_34lsub1)
deflit(`FRAME',0)
movl PARAM_SIZE, %ecx
movl PARAM_SRC, %edx
subl $2, %ecx C size-2
movl (%edx), %eax C src[0]
ja L(three_or_more)
jb L(one)
C size==2
movl 4(%edx), %ecx C src[1]
movl %eax, %edx C src[0]
shrl $24, %eax C src[0] high
andl $0xFFFFFF, %edx C src[0] low
addl %edx, %eax
movl %ecx, %edx C src[1]
shrl $16, %ecx C src[1] high
andl $0xFFFF, %edx
addl %ecx, %eax
shll $8, %edx C src[1] low
addl %edx, %eax
L(one):
ret
L(three_or_more):
C eax src[0], initial acc 0mod3
C ebx
C ecx size-2
C edx src
C esi
C edi
C ebp
movl %ebx, SAVE_EBX
movl 4(%edx), %ebx C src[1], initial 1mod3
subl $3, %ecx C size-5
movl %esi, SAVE_ESI
movl 8(%edx), %esi C src[2], initial 2mod3
pushl %edi FRAME_pushl()
movl $0, %edi C initial carries 0mod3
jng L(done) C if size < 6
L(top):
C eax acc 0mod3
C ebx acc 1mod3
C ecx counter, limbs
C edx src
C esi acc 2mod3
C edi carrys into 0mod3
C ebp
addl 12(%edx), %eax
adcl 16(%edx), %ebx
adcl 20(%edx), %esi
leal 12(%edx), %edx
adcl $0, %edi
subl $3, %ecx
jg L(top) C at least 3 more to process
L(done):
C ecx is -2, -1 or 0 representing 0, 1 or 2 more limbs respectively
cmpl $-1, %ecx
jl L(done_0) C if -2, meaning 0 more limbs
C 1 or 2 more limbs
movl $0, %ecx
je L(done_1) C if -1, meaning 1 more limb only
movl 16(%edx), %ecx
L(done_1):
addl 12(%edx), %eax C 0mod3
adcl %ecx, %ebx C 1mod3
adcl $0, %esi C 2mod3
adcl $0, %edi C carries 0mod3
L(done_0):
C eax acc 0mod3
C ebx acc 1mod3
C ecx
C edx
C esi acc 2mod3
C edi carries 0mod3
C ebp
movl %eax, %ecx C 0mod3
shrl $24, %eax C 0mod3 high initial total
andl $0xFFFFFF, %ecx C 0mod3 low
movl %edi, %edx C carries
shrl $24, %edi C carries high
addl %ecx, %eax C add 0mod3 low
andl $0xFFFFFF, %edx C carries 0mod3 low
movl %ebx, %ecx C 1mod3
shrl $16, %ebx C 1mod3 high
addl %edi, %eax C add carries high
addl %edx, %eax C add carries 0mod3 low
andl $0xFFFF, %ecx C 1mod3 low mask
addl %ebx, %eax C add 1mod3 high
movl SAVE_EBX, %ebx
shll $8, %ecx C 1mod3 low
movl %esi, %edx C 2mod3
popl %edi FRAME_popl()
shrl $8, %esi C 2mod3 high
andl $0xFF, %edx C 2mod3 low mask
addl %ecx, %eax C add 1mod3 low
shll $16, %edx C 2mod3 low
addl %esi, %eax C add 2mod3 high
movl SAVE_ESI, %esi
addl %edx, %eax C add 2mod3 low
ret
EPILOGUE()
|
