blob: c0b5c5c1cfa09ab4aaad734b8c900ca7fe7f716a [file] [log] [blame]
Austin Schuhdace2a62020-08-18 10:56:48 -07001dnl IA-64 mpn_popcount -- mpn population count.
2
3dnl Contributed to the GNU project by Torbjorn Granlund.
4
5dnl Copyright 2000-2005 Free Software Foundation, Inc.
6
7dnl This file is part of the GNU MP Library.
8dnl
9dnl The GNU MP Library is free software; you can redistribute it and/or modify
10dnl it under the terms of either:
11dnl
12dnl * the GNU Lesser General Public License as published by the Free
13dnl Software Foundation; either version 3 of the License, or (at your
14dnl option) any later version.
15dnl
16dnl or
17dnl
18dnl * the GNU General Public License as published by the Free Software
19dnl Foundation; either version 2 of the License, or (at your option) any
20dnl later version.
21dnl
22dnl or both in parallel, as here.
23dnl
24dnl The GNU MP Library is distributed in the hope that it will be useful, but
25dnl WITHOUT ANY WARRANTY; without even the implied warranty of MERCHANTABILITY
26dnl or FITNESS FOR A PARTICULAR PURPOSE. See the GNU General Public License
27dnl for more details.
28dnl
29dnl You should have received copies of the GNU General Public License and the
30dnl GNU Lesser General Public License along with the GNU MP Library. If not,
31dnl see https://www.gnu.org/licenses/.
32
33include(`../config.m4')
34
35C cycles/limb
36C Itanium: 1.5
37C Itanium 2: 1
38
39C INPUT PARAMETERS
40define(`up', `r32')
41define(`n', `r33')
42
43define(`u0',`r16') define(`u1',`r17') define(`u2',`r18') define(`u3',`r19')
44define(`c0',`r28') define(`c1',`r29') define(`c2',`r30') define(`c3',`r31')
45define(`s',`r8')
46
47
48ASM_START()
49PROLOGUE(mpn_popcount)
50 .prologue
51ifdef(`HAVE_ABI_32',
52` addp4 up = 0, up C M I
53 nop.m 0
54 zxt4 n = n C I
55 ;;
56')
57
58 {.mmi; add r9 = 512, up C prefetch pointer M I
59 ld8 r10 = [up], 8 C load first limb M01
60 mov.i r2 = ar.lc C save ar.lc I0
61}{.mmi; and r14 = 3, n C M I
62 cmp.lt p15, p14 = 4, n C small count? M I
63 add n = -5, n C M I
64 ;;
65}{.mmi; cmp.eq p6, p0 = 1, r14 C M I
66 cmp.eq p7, p0 = 2, r14 C M I
67 cmp.eq p8, p0 = 3, r14 C M I
68}{.bbb
69 (p6) br.dptk .Lb01 C B
70 (p7) br.dptk .Lb10 C B
71 (p8) br.dptk .Lb11 C B
72}
73
74
75.Lb00: ld8 u1 = [up], 8 C M01
76 shr.u n = n, 2 C I0
77 mov s = 0 C M I
78 ;;
79 ld8 u2 = [up], 8 C M01
80 popcnt c0 = r10 C I0
81 mov.i ar.lc = n C I0
82 ;;
83 ld8 u3 = [up], 8 C M01
84 popcnt c1 = u1 C I0
85 (p15) br.cond.dptk .grt4 C B
86 ;;
87 nop.m 0 C -
88 nop.m 0 C -
89 popcnt c2 = u2 C I0
90 ;;
91 mov s = c0 C M I
92 popcnt c3 = u3 C I0
93 br .Lcj4 C B
94
95.grt4: ld8 u0 = [up], 8 C M01
96 popcnt c2 = u2 C I0
97 br .LL00 C B
98
99
100.Lb01:
101 popcnt s = r10 C I0
102 (p14) br.ret.sptk.many b0 C B
103
104.grt1: ld8 u0 = [up], 8 C M01
105 shr.u n = n, 2 C I0
106 ;;
107 ld8 u1 = [up], 8 C M01
108 mov.i ar.lc = n C I0
109 ;;
110 ld8 u2 = [up], 8 C M01
111 popcnt c0 = u0 C I0
112 mov c3 = 0 C I0
113
114 ;;
115 ld8 u3 = [up], 8 C M01
116 popcnt c1 = u1 C I0
117 br.cloop.dptk .Loop C B
118 br .Lend C B
119
120
121.Lb10: ld8 u3 = [up], 8 C M01
122 shr.u n = n, 2 C I0
123 (p15) br.cond.dptk .grt2 C B
124
125 popcnt s = r10 C I0
126 ;;
127 popcnt c3 = u3 C I0
128 br .Lcj2 C B
129
130.grt2: ld8 u0 = [up], 8 C M01
131 mov.i ar.lc = n C I0
132 popcnt c2 = r10 C I0
133 ;;
134 ld8 u1 = [up], 8 C M01
135 popcnt c3 = u3 C I0
136 mov s = 0 C M I
137 ;;
138 ld8 u2 = [up], 8 C M01
139 popcnt c0 = u0 C I0
140 br .LL10 C B
141
142
143.Lb11: ld8 u2 = [up], 8 C M01
144 shr.u n = n, 2 C I0
145 mov s = 0 C M I
146 ;;
147 ld8 u3 = [up], 8 C M01
148 popcnt s = r10 C I0
149 (p15) br.cond.dptk .grt3 C B
150
151 popcnt c2 = u2 C I0
152 ;;
153 popcnt c3 = u3 C I0
154 br .Lcj3 C B
155
156.grt3: ld8 u0 = [up], 8 C M01
157 popcnt c2 = u2 C I0
158 mov.i ar.lc = n C I0
159 mov c1 = 0
160 ;;
161 ld8 u1 = [up], 8 C M01
162 popcnt c3 = u3 C I0
163 br .LL11 C B
164
165
166.Loop: ld8 u0 = [up], 8 C M01
167 popcnt c2 = u2 C I0
168 add s = s, c3 C M I
169 ;;
170.LL00: ld8 u1 = [up], 8 C M01
171 popcnt c3 = u3 C I0
172 add s = s, c0 C M I
173 ;;
174.LL11: ld8 u2 = [up], 8 C M01
175 popcnt c0 = u0 C I0
176 add s = s, c1 C M I
177 ;;
178.LL10: ld8 u3 = [up], 8 C M01
179 popcnt c1 = u1 C I0
180 add s = s, c2 C M I
181 lfetch [r9], 32 C M01
182 nop.m 0 C -
183 br.cloop.dptk .Loop C B
184 ;;
185
186.Lend: popcnt c2 = u2 C I0
187 add s = s, c3 C M I
188 ;;
189 popcnt c3 = u3 C I0
190 add s = s, c0 C M I
191 ;;
192.Lcj4: add s = s, c1 C M I
193 ;;
194.Lcj3: add s = s, c2 C M I
195 ;;
196.Lcj2: add s = s, c3 C M I
197 mov.i ar.lc = r2 C I0
198 br.ret.sptk.many b0 C B
199EPILOGUE()
200ASM_END()