]> git.ipfire.org Git - thirdparty/glibc.git/blob - sysdeps/ia64/strncpy.S
40c9309ead003da150d4598136c4aa7bd04bf53f
[thirdparty/glibc.git] / sysdeps / ia64 / strncpy.S
1 /* Optimized version of the standard strncpy() function.
2 This file is part of the GNU C Library.
3 Copyright (C) 2000-2019 Free Software Foundation, Inc.
4 Contributed by Dan Pop <Dan.Pop@cern.ch>
5 and Jakub Jelinek <jakub@redhat.com>.
6
7 The GNU C Library is free software; you can redistribute it and/or
8 modify it under the terms of the GNU Lesser General Public
9 License as published by the Free Software Foundation; either
10 version 2.1 of the License, or (at your option) any later version.
11
12 The GNU C Library is distributed in the hope that it will be useful,
13 but WITHOUT ANY WARRANTY; without even the implied warranty of
14 MERCHANTABILITY or FITNESS FOR A PARTICULAR PURPOSE. See the GNU
15 Lesser General Public License for more details.
16
17 You should have received a copy of the GNU Lesser General Public
18 License along with the GNU C Library; if not, see
19 <http://www.gnu.org/licenses/>. */
20
21 /* Return: dest
22
23 Inputs:
24 in0: dest
25 in1: src
26 in2: len
27
28 In this form, it assumes little endian mode.
29 */
30
31 #include <sysdep.h>
32 #undef ret
33
34 #define saved_lc r15
35 #define saved_pr r16
36 #define thresh r17
37 #define dest r18
38 #define dest2 r19
39 #define src r20
40 #define len r21
41 #define asrc r22
42 #define tmp r23
43 #define pos r24
44 #define w0 r25
45 #define w1 r26
46 #define c r27
47 #define sh2 r28
48 #define sh1 r29
49 #define loopcnt r30
50 #define value r31
51
52 ENTRY(strncpy)
53 .prologue
54 alloc r2 = ar.pfs, 3, 0, 29, 32
55
56 #define MEMLAT 2
57 .rotr r[MEMLAT + 2]
58 .rotp p[MEMLAT + 1]
59
60 mov ret0 = in0 // return value = dest
61 .save pr, saved_pr
62 mov saved_pr = pr // save the predicate registers
63 .save ar.lc, saved_lc
64 mov saved_lc = ar.lc // save the loop counter
65 mov ar.ec = 0 // ec is not guaranteed to
66 // be zero upon function entry
67 .body
68 cmp.geu p6, p5 = 24, in2
69 (p6) br.cond.spnt .short_len
70 sub tmp = r0, in0 ;; // tmp = -dest
71 mov len = in2 // len
72 mov dest = in0 // dest
73 mov src = in1 // src
74 and tmp = 7, tmp ;; // loopcnt = -dest % 8
75 cmp.eq p6, p7 = tmp, r0
76 adds loopcnt = -1, tmp // --loopcnt
77 (p6) br.cond.sptk .dest_aligned ;;
78 sub len = len, tmp // len -= -dest % 8
79 mov ar.lc = loopcnt
80 .l1: // copy -dest % 8 bytes
81 (p5) ld1 c = [src], 1 // c = *src++
82 ;;
83 st1 [dest] = c, 1 // *dest++ = c
84 cmp.ne p5, p7 = c, r0
85 br.cloop.dptk .l1 ;;
86 (p7) br.cond.dpnt .found0_align
87
88 .dest_aligned: // p7 should be cleared here
89 shr.u c = len, 3 // c = len / 8
90 and sh1 = 7, src // sh1 = src % 8
91 and asrc = -8, src ;; // asrc = src & -OPSIZ -- align src
92 adds c = (MEMLAT-1), c // c = (len / 8) + MEMLAT - 1
93 sub thresh = 8, sh1
94 mov pr.rot = 1 << 16 // set rotating predicates
95 shl sh1 = sh1, 3 ;; // sh1 = 8 * (src % 8)
96 mov ar.lc = c // "infinite" loop
97 sub sh2 = 64, sh1 // sh2 = 64 - sh1
98 cmp.eq p6, p0 = sh1, r0 // is the src aligned?
99 (p6) br.cond.sptk .src_aligned
100 adds c = -(MEMLAT-1), c ;; // c = (len / 8)
101 ld8 r[1] = [asrc],8
102 mov ar.lc = c ;;
103
104 .align 32
105 .l2:
106 (p6) st8 [dest] = value, 8 // store val to dest
107 ld8.s r[0] = [asrc], 8
108 shr.u value = r[1], sh1 ;; // value = w0 >> sh1
109 czx1.r pos = value ;; // do we have an "early" zero
110 cmp.lt p7, p0 = pos, thresh // in w0 >> sh1?
111 adds len = -8, len // len -= 8
112 (p7) br.cond.dpnt .nonalign_found0
113 chk.s r[0], .recovery2 // it is safe to do that only
114 .back2: // after the previous test
115 shl tmp = r[0], sh2 // tmp = w1 << sh2
116 ;;
117 or value = value, tmp ;; // value |= tmp
118 czx1.r pos = value ;;
119 cmp.ne p7, p6 = 8, pos
120 (p7) br.cond.dpnt .nonalign_found0
121 br.ctop.dptk .l2 ;;
122 adds len = 8, len
123 br.cond.sptk .not_found0 ;;
124 .nonalign_found0:
125 cmp.gtu p6, p0 = -8, len
126 (p6) br.cond.dptk .found0
127 adds len = 8, len
128 br.cond.sptk .not_found0 ;;
129
130 .align 32
131 .src_aligned:
132 .l3:
133 (p[0]) ld8.s r[0] = [src], 8
134 (p[MEMLAT]) chk.s r[MEMLAT], .recovery3
135 .back3:
136 (p[MEMLAT]) mov value = r[MEMLAT]
137 (p[MEMLAT]) czx1.r pos = r[MEMLAT] ;;
138 (p[MEMLAT]) cmp.ne p7, p0 = 8, pos
139 (p[MEMLAT]) adds len = -8, len // len -= 8
140 (p7) br.cond.dpnt .found0
141 (p[MEMLAT]) st8 [dest] = r[MEMLAT], 8
142 br.ctop.dptk .l3 ;;
143
144 chk.s r[MEMLAT-1], .recovery4
145 .back4:
146 mov value = r[MEMLAT-1]
147
148 .not_found0:
149 cmp.eq p5, p6 = len, r0
150 adds len = -1, len
151 (p5) br.cond.dptk .restore_and_exit ;;
152 mov ar.lc = len
153 .l4:
154 (p6) extr.u c = value, 0, 8 // c = value & 0xff
155 (p6) shr.u value = value, 8 ;;
156 st1 [dest] = c, 1
157 cmp.ne p6, p0 = c, r0
158 br.cloop.dptk .l4
159 br.cond.sptk .restore_and_exit
160
161 .found0_align:
162 mov pos = 0
163 adds len = -8, len
164 mov value = 0 ;;
165 .found0:
166 shl tmp = pos, 3
167 shr.u loopcnt = len, 4 // loopcnt = len / 16
168 mov c = -1 ;;
169 cmp.eq p6, p0 = loopcnt, r0
170 adds loopcnt = -1, loopcnt
171 shl c = c, tmp ;;
172 and len = 0xf, len
173 andcm value = value, c
174 mov ar.lc = loopcnt ;;
175 cmp.le p7, p0 = 8, len
176 adds dest2 = 16, dest
177 st8 [dest] = value, 8
178 and len = 0x7, len
179 (p6) br.cond.dpnt .l6 ;;
180 .l5:
181 st8 [dest] = r0, 16
182 st8 [dest2] = r0, 16
183 br.cloop.dptk .l5 ;;
184 .l6:
185 (p7) st8 [dest] = r0, 8
186 cmp.eq p5, p0 = len, r0
187 adds len = -1, len
188 (p5) br.cond.dptk .restore_and_exit ;;
189 mov ar.lc = len ;;
190 .l7:
191 st1 [dest] = r0, 1
192 br.cloop.dptk .l7 ;;
193 .restore_and_exit:
194 mov ar.lc = saved_lc // restore the loop counter
195 mov pr = saved_pr, -1 // restore the predicate registers
196 br.ret.sptk.many b0
197
198 .short_len:
199 cmp.eq p5, p0 = in2, r0
200 adds loopcnt = -1, in2
201 (p5) br.cond.spnt .restore_and_exit ;;
202 mov ar.lc = loopcnt // p6 should be set when we get here
203 .l8:
204 (p6) ld1 c = [in1], 1 // c = *src++
205 ;;
206 st1 [in0] = c, 1 // *dest++ = c
207 (p6) cmp.ne p6, p0 = c, r0
208 br.cloop.dptk .l8
209 ;;
210 mov ar.lc = saved_lc // restore the loop counter
211 mov pr = saved_pr, -1 // restore the predicate registers
212 br.ret.sptk.many b0
213 .recovery2:
214 add c = 8, len
215 add tmp = -8, asrc ;;
216 cmp.gtu p8, p5 = c, thresh ;;
217 (p8) ld8 r[0] = [tmp]
218 (p5) mov r[0] = r0
219 br.cond.sptk .back2
220 .recovery3:
221 add tmp = -(MEMLAT + 1) * 8, src ;;
222 ld8 r[MEMLAT] = [tmp]
223 br.cond.sptk .back3
224 .recovery4:
225 cmp.eq p5, p6 = len, r0
226 add tmp = -MEMLAT * 8, src ;;
227 (p6) ld8 r[MEMLAT - 1] = [tmp]
228 (p5) mov r[MEMLAT - 1] = r0
229 br.cond.sptk .back4
230 END(strncpy)
231 libc_hidden_builtin_def (strncpy)