1 /* Optimized version of the standard strcpy() function.
2 This file is part of the GNU C Library.
3 Copyright (C) 2000-2023 Free Software Foundation, Inc.
5 The GNU C Library is free software; you can redistribute it and/or
6 modify it under the terms of the GNU Lesser General Public
7 License as published by the Free Software Foundation; either
8 version 2.1 of the License, or (at your option) any later version.
10 The GNU C Library is distributed in the hope that it will be useful,
11 but WITHOUT ANY WARRANTY; without even the implied warranty of
12 MERCHANTABILITY or FITNESS FOR A PARTICULAR PURPOSE. See the GNU
13 Lesser General Public License for more details.
15 You should have received a copy of the GNU Lesser General Public
16 License along with the GNU C Library; if not, see
17 <https://www.gnu.org/licenses/>. */
25 In this form, it assumes little endian mode. For big endian mode,
26 the two shifts in .l2 must be inverted:
28 shl value = r[1], sh1 // value = w0 << sh1
29 shr.u tmp = r[0], sh2 // tmp = w1 >> sh2
54 alloc r2 = ar.pfs, 2, 0, 30, 32
60 mov ret0 = in0 // return value = dest
62 mov saved_pr = pr // save the predicate registers
64 mov saved_lc = ar.lc // save the loop counter
66 sub tmp = r0, in0 ;; // tmp = -dest
67 mov dest = in0 // dest
69 and loopcnt = 7, tmp ;; // loopcnt = -dest % 8
70 cmp.eq p6, p0 = loopcnt, r0
71 adds loopcnt = -1, loopcnt // --loopcnt
72 (p6) br.cond.sptk .dest_aligned ;;
74 .l1: // copy -dest % 8 bytes
75 ld1 c = [src], 1 // c = *src++
77 st1 [dest] = c, 1 // *dest++ = c
79 (p6) br.cond.dpnt .restore_and_exit
82 and sh1 = 7, src // sh1 = src % 8
83 mov ar.lc = -1 // "infinite" loop
84 and asrc = -8, src ;; // asrc = src & -OPSIZ -- align src
86 mov pr.rot = 1 << 16 // set rotating predicates
87 cmp.ne p7, p0 = r0, r0 // clear p7
88 shl sh1 = sh1, 3 ;; // sh1 = 8 * (src % 8)
89 sub sh2 = 64, sh1 // sh2 = 64 - sh1
90 cmp.eq p6, p0 = sh1, r0 // is the src aligned?
91 (p6) br.cond.sptk .src_aligned ;;
92 ld8 r[1] = [asrc],8 ;;
96 ld8.s r[0] = [asrc], 8
97 shr.u value = r[1], sh1 ;; // value = w0 >> sh1
98 czx1.r pos = value ;; // do we have an "early" zero
99 cmp.lt p7, p0 = pos, thresh // in w0 >> sh1?
100 (p7) br.cond.dpnt .found0
101 chk.s r[0], .recovery2 // it is safe to do that only
102 .back2: // after the previous test
103 shl tmp = r[0], sh2 // tmp = w1 << sh2
105 or value = value, tmp ;; // value |= tmp
106 czx1.r pos = value ;;
107 cmp.ne p7, p0 = 8, pos
108 (p7) br.cond.dpnt .found0
109 st8 [dest] = value, 8 // store val to dest
113 (p[0]) ld8.s r[0] = [src], 8
114 (p[MEMLAT]) chk.s r[MEMLAT], .recovery3
116 (p[MEMLAT]) mov value = r[MEMLAT]
117 (p[MEMLAT]) czx1.r pos = r[MEMLAT] ;;
118 (p[MEMLAT]) cmp.ne p7, p0 = 8, pos
119 (p7) br.cond.dpnt .found0
120 (p[MEMLAT]) st8 [dest] = r[MEMLAT], 8
125 extr.u c = value, 0, 8 // c = value & 0xff
126 shr.u value = value, 8
131 mov ar.lc = saved_lc // restore the loop counter
132 mov pr = saved_pr, -1 // restore the predicate registers
135 add tmp = -8, asrc ;;
139 add tmp = -(MEMLAT + 1) * 8, src ;;
140 ld8 r[MEMLAT] = [tmp]
143 libc_hidden_builtin_def (strcpy)