1/* 2 strchr - find a character in a string 3 4 Copyright (c) 2014, ARM Limited 5 All rights Reserved. 6 7 Redistribution and use in source and binary forms, with or without 8 modification, are permitted provided that the following conditions are met: 9 * Redistributions of source code must retain the above copyright 10 notice, this list of conditions and the following disclaimer. 11 * Redistributions in binary form must reproduce the above copyright 12 notice, this list of conditions and the following disclaimer in the 13 documentation and/or other materials provided with the distribution. 14 * Neither the name of the company nor the names of its contributors 15 may be used to endorse or promote products derived from this 16 software without specific prior written permission. 17 18 THIS SOFTWARE IS PROVIDED BY THE COPYRIGHT HOLDERS AND CONTRIBUTORS 19 "AS IS" AND ANY EXPRESS OR IMPLIED WARRANTIES, INCLUDING, BUT NOT 20 LIMITED TO, THE IMPLIED WARRANTIES OF MERCHANTABILITY AND FITNESS FOR 21 A PARTICULAR PURPOSE ARE DISCLAIMED. IN NO EVENT SHALL THE COPYRIGHT 22 HOLDER OR CONTRIBUTORS BE LIABLE FOR ANY DIRECT, INDIRECT, INCIDENTAL, 23 SPECIAL, EXEMPLARY, OR CONSEQUENTIAL DAMAGES (INCLUDING, BUT NOT 24 LIMITED TO, PROCUREMENT OF SUBSTITUTE GOODS OR SERVICES; LOSS OF USE, 25 DATA, OR PROFITS; OR BUSINESS INTERRUPTION) HOWEVER CAUSED AND ON ANY 26 THEORY OF LIABILITY, WHETHER IN CONTRACT, STRICT LIABILITY, OR TORT 27 (INCLUDING NEGLIGENCE OR OTHERWISE) ARISING IN ANY WAY OUT OF THE USE 28 OF THIS SOFTWARE, EVEN IF ADVISED OF THE POSSIBILITY OF SUCH DAMAGE. */ 29 30/* Assumptions: 31 * 32 * ARMv8-a, AArch64 33 * Neon Available. 34 */ 35 36/* Arguments and results. */ 37#define srcin x0 38#define chrin w1 39 40#define result x0 41 42#define src x2 43#define tmp1 x3 44#define wtmp2 w4 45#define tmp3 x5 46 47#define vrepchr v0 48#define vdata1 v1 49#define vdata2 v2 50#define vhas_nul1 v3 51#define vhas_nul2 v4 52#define vhas_chr1 v5 53#define vhas_chr2 v6 54#define vrepmask_0 v7 55#define vrepmask_c v16 56#define vend1 v17 57#define vend2 v18 58 59/* Core algorithm. 60 61 For each 32-byte hunk we calculate a 64-bit syndrome value, with 62 two bits per byte (LSB is always in bits 0 and 1, for both big 63 and little-endian systems). For each tuple, bit 0 is set iff 64 the relevant byte matched the requested character; bit 1 is set 65 iff the relevant byte matched the NUL end of string (we trigger 66 off bit0 for the special case of looking for NUL). Since the bits 67 in the syndrome reflect exactly the order in which things occur 68 in the original string a count_trailing_zeros() operation will 69 identify exactly which byte is causing the termination, and why. */ 70 71/* Locals and temporaries. */ 72 73 .macro def_fn f p2align=0 74 .text 75 .p2align \p2align 76 .global \f 77 .type \f, %function 78\f: 79 .endm 80 81def_fn strchr 82 /* Magic constant 0x40100401 to allow us to identify which lane 83 matches the requested byte. Magic constant 0x80200802 used 84 similarly for NUL termination. */ 85 mov wtmp2, #0x0401 86 movk wtmp2, #0x4010, lsl #16 87 dup vrepchr.16b, chrin 88 bic src, srcin, #31 /* Work with aligned 32-byte hunks. */ 89 dup vrepmask_c.4s, wtmp2 90 ands tmp1, srcin, #31 91 add vrepmask_0.4s, vrepmask_c.4s, vrepmask_c.4s /* equiv: lsl #1 */ 92 b.eq .Lloop 93 94 /* Input string is not 32-byte aligned. Rather than forcing 95 the padding bytes to a safe value, we calculate the syndrome 96 for all the bytes, but then mask off those bits of the 97 syndrome that are related to the padding. */ 98 ld1 {vdata1.16b, vdata2.16b}, [src], #32 99 neg tmp1, tmp1 100 cmeq vhas_nul1.16b, vdata1.16b, #0 101 cmeq vhas_chr1.16b, vdata1.16b, vrepchr.16b 102 cmeq vhas_nul2.16b, vdata2.16b, #0 103 cmeq vhas_chr2.16b, vdata2.16b, vrepchr.16b 104 and vhas_nul1.16b, vhas_nul1.16b, vrepmask_0.16b 105 and vhas_nul2.16b, vhas_nul2.16b, vrepmask_0.16b 106 and vhas_chr1.16b, vhas_chr1.16b, vrepmask_c.16b 107 and vhas_chr2.16b, vhas_chr2.16b, vrepmask_c.16b 108 orr vend1.16b, vhas_nul1.16b, vhas_chr1.16b 109 orr vend2.16b, vhas_nul2.16b, vhas_chr2.16b 110 lsl tmp1, tmp1, #1 111 addp vend1.16b, vend1.16b, vend2.16b // 256->128 112 mov tmp3, #~0 113 addp vend1.16b, vend1.16b, vend2.16b // 128->64 114 lsr tmp1, tmp3, tmp1 115 116 mov tmp3, vend1.2d[0] 117 bic tmp1, tmp3, tmp1 // Mask padding bits. 118 cbnz tmp1, .Ltail 119 120.Lloop: 121 ld1 {vdata1.16b, vdata2.16b}, [src], #32 122 cmeq vhas_nul1.16b, vdata1.16b, #0 123 cmeq vhas_chr1.16b, vdata1.16b, vrepchr.16b 124 cmeq vhas_nul2.16b, vdata2.16b, #0 125 cmeq vhas_chr2.16b, vdata2.16b, vrepchr.16b 126 /* Use a fast check for the termination condition. */ 127 orr vend1.16b, vhas_nul1.16b, vhas_chr1.16b 128 orr vend2.16b, vhas_nul2.16b, vhas_chr2.16b 129 orr vend1.16b, vend1.16b, vend2.16b 130 addp vend1.2d, vend1.2d, vend1.2d 131 mov tmp1, vend1.2d[0] 132 cbz tmp1, .Lloop 133 134 /* Termination condition found. Now need to establish exactly why 135 we terminated. */ 136 and vhas_nul1.16b, vhas_nul1.16b, vrepmask_0.16b 137 and vhas_nul2.16b, vhas_nul2.16b, vrepmask_0.16b 138 and vhas_chr1.16b, vhas_chr1.16b, vrepmask_c.16b 139 and vhas_chr2.16b, vhas_chr2.16b, vrepmask_c.16b 140 orr vend1.16b, vhas_nul1.16b, vhas_chr1.16b 141 orr vend2.16b, vhas_nul2.16b, vhas_chr2.16b 142 addp vend1.16b, vend1.16b, vend2.16b // 256->128 143 addp vend1.16b, vend1.16b, vend2.16b // 128->64 144 145 mov tmp1, vend1.2d[0] 146.Ltail: 147 /* Count the trailing zeros, by bit reversing... */ 148 rbit tmp1, tmp1 149 /* Re-bias source. */ 150 sub src, src, #32 151 clz tmp1, tmp1 /* And counting the leading zeros. */ 152 /* Tmp1 is even if the target charager was found first. Otherwise 153 we've found the end of string and we weren't looking for NUL. */ 154 tst tmp1, #1 155 add result, src, tmp1, lsr #1 156 csel result, result, xzr, eq 157 ret 158 159 .size strchr, . - strchr 160