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
|
dnl AMD64 mpn_copyi optimised for CPUs with fast AVX.
dnl Copyright 2003, 2005, 2007, 2011-2013, 2015 Free Software Foundation, Inc.
dnl Contributed to the GNU project by Torbjörn Granlund.
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 cycles/limb cycles/limb good
C aligned unaligned best seen for cpu?
C AMD K8,K9 n/a
C AMD K10 n/a
C AMD bd1 n/a
C AMD bd2 4.87 4.87 N
C AMD bd3 ? ?
C AMD bd4 0.53 ?
C AMD zn1 0.51 ?
C AMD zn2 0.25 ? Y
C AMD zn3 0.25 ? Y
C AMD bt1 n/a
C AMD bt2 n/a
C Intel P4 n/a
C Intel CNR n/a
C Intel PNR n/a
C Intel NHM n/a
C Intel WSM n/a
C Intel SBR 0.50 0.91 N
C Intel IBR 0.50 0.65 N
C Intel HWL 0.25 0.30 Y
C Intel BWL 0.28 0.37 Y
C Intel SKL 0.27 ? Y
C Intel atom n/a
C Intel SLM n/a
C Intel GLM n/a
C VIA nano n/a
C We try to do as many 32-byte operations as possible. The top-most and
C bottom-most writes might need 8-byte operations. For the bulk copying, we
C write using aligned 32-byte operations, but we read with both aligned and
C unaligned 32-byte operations.
define(`rp', `%rdi')
define(`up', `%rsi')
define(`n', `%rdx')
ABI_SUPPORT(DOS64)
ABI_SUPPORT(STD64)
dnl define(`vmovdqu', vlddqu)
ASM_START()
TEXT
ALIGN(32)
PROLOGUE(mpn_copyi)
FUNC_ENTRY(3)
cmp $7, n
jbe L(bc)
test $8, R8(rp) C is rp 16-byte aligned?
jz L(a2) C jump if rp aligned
mov (up), %rax
lea 8(up), up
mov %rax, (rp)
lea 8(rp), rp
dec n
L(a2): test $16, R8(rp) C is rp 32-byte aligned?
jz L(a3) C jump if rp aligned
vmovdqu (up), %xmm0
lea 16(up), up
vmovdqa %xmm0, (rp)
lea 16(rp), rp
sub $2, n
L(a3): sub $16, n
jc L(sma)
ALIGN(16)
L(top): vmovdqu (up), %ymm0
vmovdqu 32(up), %ymm1
vmovdqu 64(up), %ymm2
vmovdqu 96(up), %ymm3
lea 128(up), up
vmovdqa %ymm0, (rp)
vmovdqa %ymm1, 32(rp)
vmovdqa %ymm2, 64(rp)
vmovdqa %ymm3, 96(rp)
lea 128(rp), rp
L(ali): sub $16, n
jnc L(top)
L(sma): test $8, R8(n)
jz 1f
vmovdqu (up), %ymm0
vmovdqu 32(up), %ymm1
lea 64(up), up
vmovdqa %ymm0, (rp)
vmovdqa %ymm1, 32(rp)
lea 64(rp), rp
1:
test $4, R8(n)
jz 1f
vmovdqu (up), %ymm0
lea 32(up), up
vmovdqa %ymm0, (rp)
lea 32(rp), rp
1:
test $2, R8(n)
jz 1f
vmovdqu (up), %xmm0
lea 16(up), up
vmovdqa %xmm0, (rp)
lea 16(rp), rp
1:
L(end): test $1, R8(n)
jz 1f
mov (up), %r8
mov %r8, (rp)
1:
FUNC_EXIT()
ret
ALIGN(16)
L(bc): test $4, R8(n)
jz 1f
mov (up), %rax
mov 8(up), %rcx
mov 16(up), %r8
mov 24(up), %r9
lea 32(up), up
mov %rax, (rp)
mov %rcx, 8(rp)
mov %r8, 16(rp)
mov %r9, 24(rp)
lea 32(rp), rp
1:
test $2, R8(n)
jz 1f
mov (up), %rax
mov 8(up), %rcx
lea 16(up), up
mov %rax, (rp)
mov %rcx, 8(rp)
lea 16(rp), rp
1:
test $1, R8(n)
jz 1f
mov (up), %rax
mov %rax, (rp)
1:
FUNC_EXIT()
ret
EPILOGUE()
|
