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
209
210
211
212
213
214
215
216
217
218
|
dnl SPARC v9 mpn_add_n -- Add two limb vectors of the same length > 0 and
dnl store sum in a third limb vector.
dnl Copyright 2001 Free Software Foundation, Inc.
dnl This file is part of the GNU MP Library.
dnl The GNU MP Library is free software; you can redistribute it and/or modify
dnl it under the terms of the GNU Lesser General Public License as published
dnl by the Free Software Foundation; either version 2.1 of the License, or (at
dnl your option) any later version.
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 Lesser General Public
dnl License for more details.
dnl You should have received a copy of the GNU Lesser General Public License
dnl along with the GNU MP Library; see the file COPYING.LIB. If not, write to
dnl the Free Software Foundation, Inc., 59 Temple Place - Suite 330, Boston,
dnl MA 02111-1307, USA.
include(`../config.m4')
C Compute carry-out from the most significant bits of u,v, and r, where
C r=u+v+carry_in, using logic operations.
C This code runs at 4 cycles/limb on UltraSPARC 1 and 2. It has a 4 insn
C recurrency, and the UltraSPARC 1 and 2 the IE units are 100% saturated.
C Therefore, it seems futile to try to optimize this any further...
C INPUT PARAMETERS
define(`rp',`%i0')
define(`up',`%i1')
define(`vp',`%i2')
define(`n',`%i3')
define(`u0',`%l0')
define(`u1',`%l2')
define(`u2',`%l4')
define(`u3',`%l6')
define(`v0',`%l1')
define(`v1',`%l3')
define(`v2',`%l5')
define(`v3',`%l7')
define(`cy',`%i4')
define(`fanop',`fitod %f0,%f2') dnl A quasi nop running in the FA pipe
define(`fmnop',`fmuld %f0,%f0,%f4') dnl A quasi nop running in the FM pipe
ASM_START()
.register %g2,#scratch
.register %g3,#scratch
PROLOGUE(mpn_add_n)
save %sp,-160,%sp
fitod %f0,%f0 C make sure f0 contains small, quiet number
subcc n,4,%g0
bl,pn %icc,.Loop0
mov 0,cy
ldx [up+0],u0
ldx [vp+0],v0
add up,32,up
ldx [up-24],u1
ldx [vp+8],v1
add vp,32,vp
ldx [up-16],u2
ldx [vp-16],v2
ldx [up-8],u3
ldx [vp-8],v3
subcc n,8,n
add u0,v0,%g1 C main add
add %g1,cy,%g4 C carry add
or u0,v0,%g2
bl,pn %icc,.Lend4567
fanop
b,a .Loop
.align 16
C START MAIN LOOP
.Loop: andn %g2,%g4,%g2
and u0,v0,%g3
ldx [up+0],u0
fanop
C --
or %g3,%g2,%g2
ldx [vp+0],v0
add up,32,up
fanop
C --
srlx %g2,63,cy
add u1,v1,%g1
stx %g4,[rp+0]
fanop
C --
add %g1,cy,%g4
or u1,v1,%g2
fmnop
fanop
C --
andn %g2,%g4,%g2
and u1,v1,%g3
ldx [up-24],u1
fanop
C --
or %g3,%g2,%g2
ldx [vp+8],v1
add vp,32,vp
fanop
C --
srlx %g2,63,cy
add u2,v2,%g1
stx %g4,[rp+8]
fanop
C --
add %g1,cy,%g4
or u2,v2,%g2
fmnop
fanop
C --
andn %g2,%g4,%g2
and u2,v2,%g3
ldx [up-16],u2
fanop
C --
or %g3,%g2,%g2
ldx [vp-16],v2
add rp,32,rp
fanop
C --
srlx %g2,63,cy
add u3,v3,%g1
stx %g4,[rp-16]
fanop
C --
add %g1,cy,%g4
or u3,v3,%g2
fmnop
fanop
C --
andn %g2,%g4,%g2
and u3,v3,%g3
ldx [up-8],u3
fanop
C --
or %g3,%g2,%g2
subcc n,4,n
ldx [vp-8],v3
fanop
C --
srlx %g2,63,cy
add u0,v0,%g1
stx %g4,[rp-8]
fanop
C --
add %g1,cy,%g4
or u0,v0,%g2
bge,pt %icc,.Loop
fanop
C END MAIN LOOP
.Lend4567:
andn %g2,%g4,%g2
and u0,v0,%g3
or %g3,%g2,%g2
srlx %g2,63,cy
add u1,v1,%g1
stx %g4,[rp+0]
add %g1,cy,%g4
or u1,v1,%g2
andn %g2,%g4,%g2
and u1,v1,%g3
or %g3,%g2,%g2
srlx %g2,63,cy
add u2,v2,%g1
stx %g4,[rp+8]
add %g1,cy,%g4
or u2,v2,%g2
andn %g2,%g4,%g2
and u2,v2,%g3
or %g3,%g2,%g2
add rp,32,rp
srlx %g2,63,cy
add u3,v3,%g1
stx %g4,[rp-16]
add %g1,cy,%g4
or u3,v3,%g2
andn %g2,%g4,%g2
and u3,v3,%g3
or %g3,%g2,%g2
srlx %g2,63,cy
stx %g4,[rp-8]
addcc n,4,n
bz,pn %icc,.Lret
fanop
.Loop0: ldx [up],u0
add up,8,up
ldx [vp],v0
add vp,8,vp
add rp,8,rp
subcc n,1,n
add u0,v0,%g1
or u0,v0,%g2
add %g1,cy,%g4
and u0,v0,%g3
andn %g2,%g4,%g2
stx %g4,[rp-8]
or %g3,%g2,%g2
bnz,pt %icc,.Loop0
srlx %g2,63,cy
.Lret: mov cy,%i0
ret
restore
EPILOGUE(mpn_add_n)
|