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
|
dnl Intel Pentium-4 mpn_modexact_1_odd -- mpn by limb exact remainder.
dnl Copyright 2001, 2002, 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 P4: 19.0 cycles/limb
C mp_limb_t mpn_modexact_1_odd (mp_srcptr src, mp_size_t size,
C mp_limb_t divisor);
C mp_limb_t mpn_modexact_1c_odd (mp_srcptr src, mp_size_t size,
C mp_limb_t divisor, mp_limb_t carry);
C
defframe(PARAM_CARRY, 16)
defframe(PARAM_DIVISOR,12)
defframe(PARAM_SIZE, 8)
defframe(PARAM_SRC, 4)
TEXT
ALIGN(16)
PROLOGUE(mpn_modexact_1c_odd)
deflit(`FRAME',0)
movd PARAM_CARRY, %mm1
jmp L(start_1c)
EPILOGUE()
ALIGN(16)
PROLOGUE(mpn_modexact_1_odd)
deflit(`FRAME',0)
pxor %mm1, %mm1 C carry limb
L(start_1c):
movl PARAM_DIVISOR, %eax
movd PARAM_DIVISOR, %mm7
shrl %eax
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
')
C
movd %eax, %mm6 C inv
movd %eax, %mm0 C inv
pmuludq %mm6, %mm6 C inv*inv
C
pmuludq %mm7, %mm6 C inv*inv*d
paddd %mm0, %mm0 C 2*inv
C
psubd %mm6, %mm0 C inv = 2*inv - inv*inv*d
pxor %mm6, %mm6
paddd %mm0, %mm6
pmuludq %mm0, %mm0 C inv*inv
C
pmuludq %mm7, %mm0 C inv*inv*d
paddd %mm6, %mm6 C 2*inv
movl PARAM_SRC, %eax
movl PARAM_SIZE, %ecx
C
psubd %mm0, %mm6 C inv = 2*inv - inv*inv*d
ASSERT(e,` C expect d*inv == 1 mod 2^GMP_LIMB_BITS
pushl %eax FRAME_pushl()
movd %mm6, %eax
imul PARAM_DIVISOR, %eax
cmpl $1, %eax
popl %eax FRAME_popl()')
pxor %mm0, %mm0 C carry bit
C The dependent chain here is as follows.
C
C latency
C psubq s = (src-cbit) - climb 2
C pmuludq q = s*inverse 8
C pmuludq prod = q*divisor 8
C psrlq climb = high(prod) 2
C --
C 20
C
C Yet the loop measures 19.0 c/l, so obviously there's something gained
C there over a straight reading of the chip documentation.
L(top):
C eax src, incrementing
C ebx
C ecx counter, limbs
C edx
C
C mm0 carry bit
C mm1 carry limb
C mm6 inverse
C mm7 divisor
movd (%eax), %mm2
addl $4, %eax
psubq %mm0, %mm2 C src - cbit
psubq %mm1, %mm2 C src - cbit - climb
movq %mm2, %mm0
psrlq $63, %mm0 C new cbit
pmuludq %mm6, %mm2 C s*inverse
movq %mm7, %mm1
pmuludq %mm2, %mm1 C q*divisor
psrlq $32, %mm1 C new climb
subl $1, %ecx
jnz L(top)
L(done):
paddq %mm1, %mm0
movd %mm0, %eax
emms
ret
EPILOGUE()
ASM_END()
|
