1 // Copyright (c) 2010-2011, Linaro Limited
2 // All rights reserved.
4 // Redistribution and use in source and binary forms, with or without
5 // modification, are permitted provided that the following conditions
8 // * Redistributions of source code must retain the above copyright
9 // 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.
15 // * Neither the name of Linaro Limited nor the names of its
16 // contributors may be used to endorse or promote products derived
17 // from this software without specific prior written permission.
19 // THIS SOFTWARE IS PROVIDED BY THE COPYRIGHT HOLDERS AND CONTRIBUTORS
20 // "AS IS" AND ANY EXPRESS OR IMPLIED WARRANTIES, INCLUDING, BUT NOT
21 // LIMITED TO, THE IMPLIED WARRANTIES OF MERCHANTABILITY AND FITNESS FOR
22 // A PARTICULAR PURPOSE ARE DISCLAIMED. IN NO EVENT SHALL THE COPYRIGHT
23 // HOLDER OR CONTRIBUTORS BE LIABLE FOR ANY DIRECT, INDIRECT, INCIDENTAL,
24 // SPECIAL, EXEMPLARY, OR CONSEQUENTIAL DAMAGES (INCLUDING, BUT NOT
25 // LIMITED TO, PROCUREMENT OF SUBSTITUTE GOODS OR SERVICES; LOSS OF USE,
26 // DATA, OR PROFITS; OR BUSINESS INTERRUPTION) HOWEVER CAUSED AND ON ANY
27 // THEORY OF LIABILITY, WHETHER IN CONTRACT, STRICT LIABILITY, OR TORT
28 // (INCLUDING NEGLIGENCE OR OTHERWISE) ARISING IN ANY WAY OUT OF THE USE
29 // OF THIS SOFTWARE, EVEN IF ADVISED OF THE POSSIBILITY OF SUCH DAMAGE.
33 // Written by Dave Gilbert <david.gilbert@linaro.org>
35 // This memchr routine is optimised on a Cortex-A9 and should work on
36 // all ARMv7 processors. It has a fast past for short sizes, and has
37 // an optimised path for large data sets; the worst case is finding the
38 // match early in a large data set.
42 // 2011-02-07 david.gilbert@linaro.org
43 // Extracted from local git a5b438d861
44 // 2011-07-14 david.gilbert@linaro.org
45 // Import endianness fix from local git ea786f1b
46 // 2011-12-07 david.gilbert@linaro.org
47 // Removed unneeded cbz from align loop
49 // this lets us check a flag in a 00/ff byte easily in either endianness
50 #define CHARTSTMASK(c) 1<<(c*8)
56 .type ASM_PFX(InternalMemScanMem8), %function
57 ASM_GLOBAL ASM_PFX(InternalMemScanMem8)
58 ASM_PFX(InternalMemScanMem8):
59 // r0 = start of memory to scan
61 // r2 = character to look for
62 // returns r0 = pointer to character or NULL if not found
63 uxtb r2, r2 // Don't think we can trust the caller to actually pass a char
65 cmp r1, #16 // If it's short don't bother with anything clever
68 tst r0, #7 // If it's already aligned skip the next bit
71 // Work up to an aligned point
76 beq 50f // If it matches exit found
78 bne 5b // If not aligned yet then do next byte
81 // At this point, we are aligned, we know we have at least 8 bytes to work with
83 orr r2, r2, r2, lsl #8 // expand the match word across to all bytes
84 orr r2, r2, r2, lsl #16
85 bic r4, r1, #7 // Number of double words to work with
86 mvns r7, #0 // all F's
92 eor r5, r5, r2 // Get it so that r5,r6 have 00's where the bytes match the target
94 uadd8 r5, r5, r7 // Parallel add 0xff - sets the GE bits for anything that wasn't 0
95 sel r5, r3, r7 // bytes are 00 for none-00 bytes, or ff for 00 bytes - NOTE INVERSION
96 uadd8 r6, r6, r7 // Parallel add 0xff - sets the GE bits for anything that wasn't 0
97 sel r6, r5, r7 // chained....bytes are 00 for none-00 bytes, or ff for 00 bytes - NOTE INVERSION
99 bne 15b // (Flags from the subs above) If not run out of bytes then go around again
102 and r2, r2, #0xff // Get r2 back to a single character from the expansion above
103 and r1, r1, #7 // Leave the count remaining as the number after the double words have been done
106 cbz r1, 40f // 0 length or hit the end already then not found
108 21: // Post aligned section, or just a short call
111 eor r3, r3, r2 // r3 = 0 if match - doesn't break flags from sub
113 bne 21b // on r1 flags
116 movs r0, #0 // not found
120 subs r0, r0, #1 // found
123 60: // We're here because the fast path found a hit - now we have to track down exactly which word it was
124 // r0 points to the start of the double word after the one that was tested
125 // r5 has the 00/ff pattern for the first word, r6 has the chained value
129 moveq.n r5, r6 // the end is in the 2nd word
131 subne.n r0, r0, #4 // or 2nd byte of 1st word
133 // r0 currently points to the 3rd byte of the word containing the hit
134 tst r5, #CHARTSTMASK(0) // 1st character
137 tst r5, #CHARTSTMASK(1) // 2nd character
140 tst r5, #(3 << 15) // 2nd & 3rd character
141 // If not the 3rd must be the last one