| 1 | /* Optimized strlen implementation using basic LoongArch instructions. |
|---|
| 2 | Copyright (C) 2023-2024 Free Software Foundation, Inc. |
|---|
| 3 | |
|---|
| 4 | This file is part of the GNU C Library. |
|---|
| 5 | |
|---|
| 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. |
|---|
| 10 | |
|---|
| 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. |
|---|
| 15 | |
|---|
| 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 | <https://www.gnu.org/licenses/>. */ |
|---|
| 19 | |
|---|
| 20 | #include <sysdep.h> |
|---|
| 21 | #include <sys/regdef.h> |
|---|
| 22 | #include <sys/asm.h> |
|---|
| 23 | |
|---|
| 24 | #if IS_IN (libc) |
|---|
| 25 | # define STRLEN __strlen_aligned |
|---|
| 26 | #else |
|---|
| 27 | # define STRLEN strlen |
|---|
| 28 | #endif |
|---|
| 29 | |
|---|
| 30 | LEAF(STRLEN, 6) |
|---|
| 31 | move a1, a0 |
|---|
| 32 | bstrins.d a0, zero, 2, 0 |
|---|
| 33 | lu12i.w a2, 0x01010 |
|---|
| 34 | li.w t0, -1 |
|---|
| 35 | |
|---|
| 36 | ld.d t2, a0, 0 |
|---|
| 37 | andi t1, a1, 0x7 |
|---|
| 38 | ori a2, a2, 0x101 |
|---|
| 39 | slli.d t1, t1, 3 |
|---|
| 40 | |
|---|
| 41 | bstrins.d a2, a2, 63, 32 |
|---|
| 42 | sll.d t1, t0, t1 |
|---|
| 43 | slli.d t3, a2, 7 |
|---|
| 44 | nor a3, zero, t3 |
|---|
| 45 | |
|---|
| 46 | orn t2, t2, t1 |
|---|
| 47 | sub.d t0, t2, a2 |
|---|
| 48 | nor t1, t2, a3 |
|---|
| 49 | and t0, t0, t1 |
|---|
| 50 | |
|---|
| 51 | |
|---|
| 52 | bnez t0, L(count_pos) |
|---|
| 53 | addi.d a0, a0, 8 |
|---|
| 54 | L(loop_16_7bit): |
|---|
| 55 | ld.d t2, a0, 0 |
|---|
| 56 | sub.d t1, t2, a2 |
|---|
| 57 | |
|---|
| 58 | and t0, t1, t3 |
|---|
| 59 | bnez t0, L(more_check) |
|---|
| 60 | ld.d t2, a0, 8 |
|---|
| 61 | sub.d t1, t2, a2 |
|---|
| 62 | |
|---|
| 63 | and t0, t1, t3 |
|---|
| 64 | addi.d a0, a0, 16 |
|---|
| 65 | beqz t0, L(loop_16_7bit) |
|---|
| 66 | addi.d a0, a0, -8 |
|---|
| 67 | |
|---|
| 68 | L(more_check): |
|---|
| 69 | nor t0, t2, a3 |
|---|
| 70 | and t0, t1, t0 |
|---|
| 71 | bnez t0, L(count_pos) |
|---|
| 72 | addi.d a0, a0, 8 |
|---|
| 73 | |
|---|
| 74 | |
|---|
| 75 | L(loop_16_8bit): |
|---|
| 76 | ld.d t2, a0, 0 |
|---|
| 77 | sub.d t1, t2, a2 |
|---|
| 78 | nor t0, t2, a3 |
|---|
| 79 | and t0, t0, t1 |
|---|
| 80 | |
|---|
| 81 | bnez t0, L(count_pos) |
|---|
| 82 | ld.d t2, a0, 8 |
|---|
| 83 | addi.d a0, a0, 16 |
|---|
| 84 | sub.d t1, t2, a2 |
|---|
| 85 | |
|---|
| 86 | nor t0, t2, a3 |
|---|
| 87 | and t0, t0, t1 |
|---|
| 88 | beqz t0, L(loop_16_8bit) |
|---|
| 89 | addi.d a0, a0, -8 |
|---|
| 90 | |
|---|
| 91 | L(count_pos): |
|---|
| 92 | ctz.d t1, t0 |
|---|
| 93 | sub.d a0, a0, a1 |
|---|
| 94 | srli.d t1, t1, 3 |
|---|
| 95 | add.d a0, a0, t1 |
|---|
| 96 | |
|---|
| 97 | jr ra |
|---|
| 98 | END(STRLEN) |
|---|
| 99 | |
|---|
| 100 | libc_hidden_builtin_def (STRLEN) |
|---|
| 101 | |
|---|
