1/* 2 * memchr - scan memory for a character 3 * 4 * Part of the LLVM Project, under the Apache License v2.0 with LLVM Exceptions. 5 * See https://llvm.org/LICENSE.txt for license information. 6 * SPDX-License-Identifier: Apache-2.0 WITH LLVM-exception 7 */ 8 9/* 10 Written by Dave Gilbert <david.gilbert@linaro.org> 11 12 This __memchr_arm routine is optimised on a Cortex-A9 and should work on 13 all ARMv7 processors. It has a fast past for short sizes, and has 14 an optimised path for large data sets; the worst case is finding the 15 match early in a large data set. 16 17 */ 18 19@ 2011-02-07 david.gilbert@linaro.org 20@ Extracted from local git a5b438d861 21@ 2011-07-14 david.gilbert@linaro.org 22@ Import endianness fix from local git ea786f1b 23@ 2011-12-07 david.gilbert@linaro.org 24@ Removed unneeded cbz from align loop 25 26 .syntax unified 27 .arch armv7-a 28 29@ this lets us check a flag in a 00/ff byte easily in either endianness 30#ifdef __ARMEB__ 31#define CHARTSTMASK(c) 1<<(31-(c*8)) 32#else 33#define CHARTSTMASK(c) 1<<(c*8) 34#endif 35 .text 36 .thumb 37 38@ --------------------------------------------------------------------------- 39 .thumb_func 40 .align 2 41 .p2align 4,,15 42 .global __memchr_arm 43 .type __memchr_arm,%function 44__memchr_arm: 45 @ r0 = start of memory to scan 46 @ r1 = character to look for 47 @ r2 = length 48 @ returns r0 = pointer to character or NULL if not found 49 and r1,r1,#0xff @ Don't think we can trust the caller to actually pass a char 50 51 cmp r2,#16 @ If it's short don't bother with anything clever 52 blt 20f 53 54 tst r0, #7 @ If it's already aligned skip the next bit 55 beq 10f 56 57 @ Work up to an aligned point 585: 59 ldrb r3, [r0],#1 60 subs r2, r2, #1 61 cmp r3, r1 62 beq 50f @ If it matches exit found 63 tst r0, #7 64 bne 5b @ If not aligned yet then do next byte 65 6610: 67 @ At this point, we are aligned, we know we have at least 8 bytes to work with 68 push {r4,r5,r6,r7} 69 orr r1, r1, r1, lsl #8 @ expand the match word across to all bytes 70 orr r1, r1, r1, lsl #16 71 bic r4, r2, #7 @ Number of double words to work with 72 mvns r7, #0 @ all F's 73 movs r3, #0 74 7515: 76 ldmia r0!,{r5,r6} 77 subs r4, r4, #8 78 eor r5,r5, r1 @ Get it so that r5,r6 have 00's where the bytes match the target 79 eor r6,r6, r1 80 uadd8 r5, r5, r7 @ Parallel add 0xff - sets the GE bits for anything that wasn't 0 81 sel r5, r3, r7 @ bytes are 00 for none-00 bytes, or ff for 00 bytes - NOTE INVERSION 82 uadd8 r6, r6, r7 @ Parallel add 0xff - sets the GE bits for anything that wasn't 0 83 sel r6, r5, r7 @ chained....bytes are 00 for none-00 bytes, or ff for 00 bytes - NOTE INVERSION 84 cbnz r6, 60f 85 bne 15b @ (Flags from the subs above) If not run out of bytes then go around again 86 87 pop {r4,r5,r6,r7} 88 and r1,r1,#0xff @ Get r1 back to a single character from the expansion above 89 and r2,r2,#7 @ Leave the count remaining as the number after the double words have been done 90 9120: 92 cbz r2, 40f @ 0 length or hit the end already then not found 93 9421: @ Post aligned section, or just a short call 95 ldrb r3,[r0],#1 96 subs r2,r2,#1 97 eor r3,r3,r1 @ r3 = 0 if match - doesn't break flags from sub 98 cbz r3, 50f 99 bne 21b @ on r2 flags 100 10140: 102 movs r0,#0 @ not found 103 bx lr 104 10550: 106 subs r0,r0,#1 @ found 107 bx lr 108 10960: @ We're here because the fast path found a hit - now we have to track down exactly which word it was 110 @ r0 points to the start of the double word after the one that was tested 111 @ r5 has the 00/ff pattern for the first word, r6 has the chained value 112 cmp r5, #0 113 itte eq 114 moveq r5, r6 @ the end is in the 2nd word 115 subeq r0,r0,#3 @ Points to 2nd byte of 2nd word 116 subne r0,r0,#7 @ or 2nd byte of 1st word 117 118 @ r0 currently points to the 3rd byte of the word containing the hit 119 tst r5, # CHARTSTMASK(0) @ 1st character 120 bne 61f 121 adds r0,r0,#1 122 tst r5, # CHARTSTMASK(1) @ 2nd character 123 ittt eq 124 addeq r0,r0,#1 125 tsteq r5, # (3<<15) @ 2nd & 3rd character 126 @ If not the 3rd must be the last one 127 addeq r0,r0,#1 128 12961: 130 pop {r4,r5,r6,r7} 131 subs r0,r0,#1 132 bx lr 133 134 .size __memchr_arm, . - __memchr_arm 135