Richard Kuo | 7567746 | 2011-10-31 18:48:07 -0500 | [diff] [blame] | 1 | /* |
| 2 | * User string length functions for kernel |
| 3 | * |
| 4 | * Copyright (c) 2010-2011, Code Aurora Forum. All rights reserved. |
| 5 | * |
| 6 | * This program is free software; you can redistribute it and/or modify |
| 7 | * it under the terms of the GNU General Public License version 2 and |
| 8 | * only version 2 as published by the Free Software Foundation. |
| 9 | * |
| 10 | * This program is distributed in the hope that it will be useful, |
| 11 | * but WITHOUT ANY WARRANTY; without even the implied warranty of |
| 12 | * MERCHANTABILITY or FITNESS FOR A PARTICULAR PURPOSE. See the |
| 13 | * GNU General Public License for more details. |
| 14 | * |
| 15 | * You should have received a copy of the GNU General Public License |
| 16 | * along with this program; if not, write to the Free Software |
| 17 | * Foundation, Inc., 51 Franklin Street, Fifth Floor, Boston, MA |
| 18 | * 02110-1301, USA. |
| 19 | */ |
| 20 | |
| 21 | #define isrc r0 |
| 22 | #define max r1 /* Do not change! */ |
| 23 | |
| 24 | #define end r2 |
| 25 | #define tmp1 r3 |
| 26 | |
| 27 | #define obo r6 /* off-by-one */ |
| 28 | #define start r7 |
| 29 | #define mod8 r8 |
| 30 | #define dbuf r15:14 |
| 31 | #define dcmp r13:12 |
| 32 | |
| 33 | /* |
| 34 | * The vector mask version of this turned out *really* badly. |
| 35 | * The hardware loop version also turned out *really* badly. |
| 36 | * Seems straight pointer arithmetic basically wins here. |
| 37 | */ |
| 38 | |
| 39 | #define fname __strnlen_user |
| 40 | |
| 41 | .text |
| 42 | .global fname |
| 43 | .type fname, @function |
| 44 | .p2align 5 /* why? */ |
| 45 | fname: |
| 46 | { |
| 47 | mod8 = and(isrc,#7); |
| 48 | end = add(isrc,max); |
| 49 | start = isrc; |
| 50 | } |
| 51 | { |
| 52 | P0 = cmp.eq(mod8,#0); |
| 53 | mod8 = and(end,#7); |
| 54 | dcmp = #0; |
| 55 | if (P0.new) jump:t dw_loop; /* fire up the oven */ |
| 56 | } |
| 57 | |
| 58 | alignment_loop: |
| 59 | fail_1: { |
| 60 | tmp1 = memb(start++#1); |
| 61 | } |
| 62 | { |
| 63 | P0 = cmp.eq(tmp1,#0); |
| 64 | if (P0.new) jump:nt exit_found; |
| 65 | P1 = cmp.gtu(end,start); |
| 66 | mod8 = and(start,#7); |
| 67 | } |
| 68 | { |
| 69 | if (!P1) jump exit_error; /* hit the end */ |
| 70 | P0 = cmp.eq(mod8,#0); |
| 71 | } |
| 72 | { |
| 73 | if (!P0) jump alignment_loop; |
| 74 | } |
| 75 | |
| 76 | |
| 77 | |
| 78 | dw_loop: |
| 79 | fail_2: { |
| 80 | dbuf = memd(start); |
| 81 | obo = add(start,#1); |
| 82 | } |
| 83 | { |
| 84 | P0 = vcmpb.eq(dbuf,dcmp); |
| 85 | } |
| 86 | { |
| 87 | tmp1 = P0; |
| 88 | P0 = cmp.gtu(end,start); |
| 89 | } |
| 90 | { |
| 91 | tmp1 = ct0(tmp1); |
| 92 | mod8 = and(end,#7); |
| 93 | if (!P0) jump end_check; |
| 94 | } |
| 95 | { |
| 96 | P0 = cmp.eq(tmp1,#32); |
| 97 | if (!P0.new) jump:nt exit_found; |
| 98 | if (!P0.new) start = add(obo,tmp1); |
| 99 | } |
| 100 | { |
| 101 | start = add(start,#8); |
| 102 | jump dw_loop; |
| 103 | } /* might be nice to combine these jumps... */ |
| 104 | |
| 105 | |
| 106 | end_check: |
| 107 | { |
| 108 | P0 = cmp.gt(tmp1,mod8); |
| 109 | if (P0.new) jump:nt exit_error; /* neverfound! */ |
| 110 | start = add(obo,tmp1); |
| 111 | } |
| 112 | |
| 113 | exit_found: |
| 114 | { |
| 115 | R0 = sub(start,isrc); |
| 116 | jumpr R31; |
| 117 | } |
| 118 | |
| 119 | exit_error: |
| 120 | { |
| 121 | R0 = add(max,#1); |
| 122 | jumpr R31; |
| 123 | } |
| 124 | |
| 125 | /* Uh, what does the "fixup" return here? */ |
| 126 | .falign |
| 127 | fix_1: |
| 128 | { |
| 129 | R0 = #0; |
| 130 | jumpr R31; |
| 131 | } |
| 132 | |
| 133 | .size fname,.-fname |
| 134 | |
| 135 | |
| 136 | .section __ex_table,"a" |
| 137 | .long fail_1,fix_1 |
| 138 | .long fail_2,fix_1 |
| 139 | .previous |