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
|
dnl PowerPC-64 mpn_addlshC_n, mpn_sublshC_n, mpn_rsblshC_n.
dnl Copyright 2003, 2005, 2009, 2010, 2013 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/.
C cycles/limb
C POWER3/PPC630 1.83 (1.5 c/l should be possible)
C POWER4/PPC970 3 (2.0 c/l should be possible)
C POWER5 3
C POWER6 3.5-47
C POWER7 3
C STATUS
C * Try combining upx+up, and vpx+vp.
C * The worst case 47 c/l for POWER6 happens if the 3rd operand for ldx is
C greater than the 2nd operand. Yes, this addition is non-commutative wrt
C performance.
C INPUT PARAMETERS
define(`rp', `r3')
define(`up', `r4')
define(`vp', `r5')
define(`n', `r6')
ifdef(`DO_add', `
define(`ADDSUBC', `addc $1, $2, $3')
define(`ADDSUBE', `adde $1, $2, $3')
define(INITCY, `addic $1, r1, 0')
define(RETVAL, `addze r3, $1')
define(`func', mpn_addlsh`'LSH`'_n)')
ifdef(`DO_sub', `
define(`ADDSUBC', `subfc $1, $2, $3')
define(`ADDSUBE', `subfe $1, $2, $3')
define(INITCY, `addic $1, r1, -1')
define(RETVAL, `subfze r3, $1
neg r3, r3')
define(`func', mpn_sublsh`'LSH`'_n)')
ifdef(`DO_rsb', `
define(`ADDSUBC', `subfc $1, $3, $2')
define(`ADDSUBE', `subfe $1, $3, $2')
define(INITCY, `addic $1, r1, -1')
define(RETVAL, `addme r3, $1')
define(`func', mpn_rsblsh`'LSH`'_n)')
define(`rpx', `r6')
define(`upx', `r7')
define(`vpx', `r12')
define(`s0', `r0') define(`s1', `r9')
define(`u0', `r8')
define(`v0', `r10') define(`v1', `r11')
ASM_START()
PROLOGUE(func)
cmpldi cr0, n, 13
bgt L(big)
mtctr n C copy n in ctr
INITCY( r0) C clear cy
ld v0, 0(vp) C load v limb
ld u0, 0(up) C load u limb
addi up, up, -8 C update up
addi rp, rp, -8 C update rp
sldi s1, v0, LSH
bdz L(ex1) C If done, skip loop
ALIGN(16)
L(lo0): ld v1, 8(vp) C load v limb
ADDSUBE(s1, s1, u0) C add limbs with cy, set cy
ldu u0, 16(up) C load u limb and update up
srdi s0, v0, RSH C shift down previous v limb
std s1, 8(rp) C store result limb
rldimi s0, v1, LSH, 0 C left shift v limb and merge with prev v limb
bdz L(ex0) C decrement ctr and exit if done
ldu v0, 16(vp) C load v limb and update vp
ADDSUBE(s0, s0, u0) C add limbs with cy, set cy
ld u0, 8(up) C load u limb
srdi s1, v1, RSH C shift down previous v limb
stdu s0, 16(rp) C store result limb and update rp
rldimi s1, v0, LSH, 0 C left shift v limb and merge with prev v limb
bdnz L(lo0) C decrement ctr and loop back
L(ex1): ADDSUBE(r7, s1, u0)
std r7, 8(rp) C store last result limb
srdi r0, v0, RSH
RETVAL( r0)
blr
L(ex0): ADDSUBE(r7, s0, u0)
std r7, 16(rp) C store last result limb
srdi r0, v1, RSH
RETVAL( r0)
blr
L(big): rldicl. r0, n, 0,63 C r0 = n & 1, set cr0
addi r6, n, -1 C ...for ctr
srdi r6, r6, 1 C ...for ctr
mtctr r6 C copy count into ctr
beq cr0, L(b0)
L(b1): ld v1, 0(vp)
ld u0, 0(up)
sldi s1, v1, LSH
srdi s0, v1, RSH
ld v0, 8(vp)
ADDSUBC(s1, s1, u0) C add limbs without cy, set cy
addi rpx, rp, -16
addi rp, rp, -8
sub upx, up, rp
sub vpx, vp, rp
sub up, up, rpx
sub vp, vp, rpx
addi up, up, 8
addi upx, upx, 16
addi vp, vp, 16
addi vpx, vpx, 24
b L(mid)
L(b0): ld v0, 0(vp)
ld u0, 0(up)
sldi s0, v0, LSH
srdi s1, v0, RSH
ld v1, 8(vp)
ADDSUBC(s0, s0, u0) C add limbs without cy, set cy
addi rpx, rp, -8
addi rp, rp, -16
sub upx, up, rpx
sub vpx, vp, rpx
sub up, up, rp
sub vp, vp, rp
addi up, up, 8
addi upx, upx, 16
addi vp, vp, 16
addi vpx, vpx, 24
ALIGN(32)
L(top): ldx u0, rp, up
ldx v0, rp, vp
rldimi s1, v1, LSH, 0
stdu s0, 16(rp)
srdi s0, v1, RSH
ADDSUBE(s1, s1, u0) C add limbs with cy, set cy
L(mid): ldx u0, rpx, upx
ldx v1, rpx, vpx
rldimi s0, v0, LSH, 0
stdu s1, 16(rpx)
srdi s1, v0, RSH
ADDSUBE(s0, s0, u0) C add limbs with cy, set cy
bdnz L(top) C decrement CTR and loop back
ldx u0, rp, up
rldimi s1, v1, LSH, 0
std s0, 16(rp)
srdi s0, v1, RSH
ADDSUBE(s1, s1, u0) C add limbs with cy, set cy
std s1, 24(rp)
RETVAL( s0)
blr
EPILOGUE()
|
