1 | /* Copyright (C) 1996-2022 Free Software Foundation, Inc. |
2 | This file is part of the GNU C Library. |
3 | |
4 | The GNU C Library is free software; you can redistribute it and/or |
5 | modify it under the terms of the GNU Lesser General Public |
6 | License as published by the Free Software Foundation; either |
7 | version 2.1 of the License, or (at your option) any later version. |
8 | |
9 | The GNU C Library is distributed in the hope that it will be useful, |
10 | but WITHOUT ANY WARRANTY; without even the implied warranty of |
11 | MERCHANTABILITY or FITNESS FOR A PARTICULAR PURPOSE. See the GNU |
12 | Lesser General Public License for more details. |
13 | |
14 | You should have received a copy of the GNU Lesser General Public |
15 | License along with the GNU C Library. If not, see |
16 | <https://www.gnu.org/licenses/>. */ |
17 | |
18 | /* Finds length of a 0-terminated string. Optimized for the Alpha |
19 | architecture: |
20 | |
21 | - memory accessed as aligned quadwords only |
22 | - uses cmpbge to compare 8 bytes in parallel |
23 | - does binary search to find 0 byte in last quadword (HAKMEM |
24 | needed 12 instructions to do this instead of the 8 instructions |
25 | that the binary search needs). |
26 | */ |
27 | |
28 | #include <sysdep.h> |
29 | |
30 | .set noreorder |
31 | .set noat |
32 | |
33 | ENTRY(strlen) |
34 | #ifdef PROF |
35 | ldgp gp, 0(pv) |
36 | lda AT, _mcount |
37 | jsr AT, (AT), _mcount |
38 | .prologue 1 |
39 | #else |
40 | .prologue 0 |
41 | #endif |
42 | |
43 | ldq_u t0, 0(a0) # load first quadword (a0 may be misaligned) |
44 | lda t1, -1(zero) |
45 | insqh t1, a0, t1 |
46 | andnot a0, 7, v0 |
47 | or t1, t0, t0 |
48 | nop # dual issue the next two on ev5 |
49 | cmpbge zero, t0, t1 # t1 <- bitmask: bit i == 1 <==> i-th byte == 0 |
50 | bne t1, $found |
51 | |
52 | $loop: ldq t0, 8(v0) |
53 | addq v0, 8, v0 # addr += 8 |
54 | cmpbge zero, t0, t1 |
55 | beq t1, $loop |
56 | |
57 | $found: negq t1, t2 # clear all but least set bit |
58 | and t1, t2, t1 |
59 | |
60 | and t1, 0xf0, t2 # binary search for that set bit |
61 | and t1, 0xcc, t3 |
62 | and t1, 0xaa, t4 |
63 | cmovne t2, 4, t2 |
64 | cmovne t3, 2, t3 |
65 | cmovne t4, 1, t4 |
66 | addq t2, t3, t2 |
67 | addq v0, t4, v0 |
68 | addq v0, t2, v0 |
69 | nop # dual issue next two on ev4 and ev5 |
70 | |
71 | subq v0, a0, v0 |
72 | ret |
73 | |
74 | END(strlen) |
75 | libc_hidden_builtin_def (strlen) |
76 | |