1 /* Optimized memset implementation for PowerPC64/POWER7.
2 Copyright (C) 2010-2019 Free Software Foundation, Inc.
3 Contributed by Luis Machado <luisgpm@br.ibm.com>.
4 This file is part of the GNU C Library.
6 The GNU C Library is free software; you can redistribute it and/or
7 modify it under the terms of the GNU Lesser General Public
8 License as published by the Free Software Foundation; either
9 version 2.1 of the License, or (at your option) any later version.
11 The GNU C Library is distributed in the hope that it will be useful,
12 but WITHOUT ANY WARRANTY; without even the implied warranty of
13 MERCHANTABILITY or FITNESS FOR A PARTICULAR PURPOSE. See the GNU
14 Lesser General Public License for more details.
16 You should have received a copy of the GNU Lesser General Public
17 License along with the GNU C Library; if not, see
18 <http://www.gnu.org/licenses/>. */
22 /* void * [r3] memset (void *s [r3], int c [r4], size_t n [r5]));
26 # define MEMSET memset
29 ENTRY_TOCLESS (MEMSET, 5)
37 /* Replicate byte to word. */
40 ble cr6,L(small) /* If length <= 8, use short copy code. */
43 ble cr7,L(medium) /* If length < 32, use medium copy code. */
45 andi. 11,10,7 /* Check alignment of SRC. */
46 insrdi 4,4,32,0 /* Replicate word to double word. */
55 /* Get DST aligned to 8 bytes. */
64 4: bf 29,L(big_aligned)
76 srdi 9,5,3 /* Number of full doublewords remaining. */
81 /* From this point on, we'll copy 32+ bytes and the value
82 isn't 0 (so we can't use dcbz). */
90 /* Copy 1~3 doublewords so the main loop starts
91 at a multiple of 32 bytes. */
103 blt cr1,L(tail_bytes)
107 1: /* Copy 1 doubleword. */
113 /* Main aligned copy loop. Copies 32-bytes at a time and
114 ping-pong through r10 and r12 to avoid AGEN delays. */
137 /* Check for tail bytes. */
143 /* At this point we have a tail of 0-7 bytes and we know that the
144 destination is doubleword-aligned. */
145 4: /* Copy 4 bytes. */
150 2: /* Copy 2 bytes. */
155 1: /* Copy 1 byte. */
161 /* Special case when value is 0 and we have a long length to deal
162 with. Use dcbz to zero out 128-bytes at a time. Before using
163 dcbz though, we need to get the destination 128-bytes aligned. */
175 /* Get DST aligned to 128 bytes. */
202 1: bf 31,L(huge_aligned)
220 /* Check how many bytes are still left. */
229 /* We have a tail o 1~127 bytes. Copy up to 15 doublewords for
230 speed. We'll handle the resulting tail bytes later. */
264 /* Handle the rest of the tail bytes here. */
284 /* Expanded tree to copy tail bytes without increments. */
319 /* Handle copies of 9~31 bytes. */
322 /* At least 9 bytes to go. */
325 beq L(medium_aligned)
327 /* Force 4-bytes alignment for DST. */
330 1: /* Copy 1 byte. */
335 2: /* Copy 2 bytes. */
336 bf 30,L(medium_aligned)
343 /* At least 6 bytes to go, and DST is word-aligned. */
354 8: /* Copy 8 bytes. */
360 4: /* Copy 4 bytes. */
365 2: /* Copy 2-3 bytes. */
370 1: /* Copy 1 byte. */
376 /* Handles copies of 0~8 bytes. */
386 END_GEN_TB (MEMSET,TB_TOCLESS)
387 libc_hidden_builtin_def (memset)
389 /* Copied from bzero.S to prevent the linker from inserting a stub
390 between bzero and memset. */
391 ENTRY_TOCLESS (__bzero)
398 weak_alias (__bzero, bzero)