1/* 2 * CDDL HEADER START 3 * 4 * The contents of this file are subject to the terms of the 5 * Common Development and Distribution License (the "License"). 6 * You may not use this file except in compliance with the License. 7 * 8 * You can obtain a copy of the license at usr/src/OPENSOLARIS.LICENSE 9 * or http://www.opensolaris.org/os/licensing. 10 * See the License for the specific language governing permissions 11 * and limitations under the License. 12 * 13 * When distributing Covered Code, include this CDDL HEADER in each 14 * file and include the License file at usr/src/OPENSOLARIS.LICENSE. 15 * If applicable, add the following below this CDDL HEADER, with the 16 * fields enclosed by brackets "[]" replaced with your own identifying 17 * information: Portions Copyright [yyyy] [name of copyright owner] 18 * 19 * CDDL HEADER END 20 */ 21 22/* 23 * Copyright 2008 Sun Microsystems, Inc. All rights reserved. 24 * Use is subject to license terms. 25 */ 26 27 .file "strcmp.s" 28 29/* strcmp(s1, s2) 30 * 31 * Compare strings: s1>s2: >0 s1==s2: 0 s1<s2: <0 32 * 33 * Fast assembler language version of the following C-program for strcmp 34 * which represents the `standard' for the C-library. 35 * 36 * int 37 * strcmp(s1, s2) 38 * register const char *s1; 39 * register const char *s2; 40 * { 41 * 42 * if(s1 == s2) 43 * return(0); 44 * while(*s1 == *s2++) 45 * if(*s1++ == '\0') 46 * return(0); 47 * return(*s1 - s2[-1]); 48 * } 49 */ 50 51#include <sys/asm_linkage.h> 52 53 ! This strcmp implementation first determines whether s1 is aligned. 54 ! If it is not, it attempts to align it and then checks the 55 ! alignment of the destination string. If it is possible to 56 ! align s2, this also happens and then the compare begins. Otherwise, 57 ! a different compare for non-aligned strings is used. 58 59 ENTRY(strcmp) 60 61 .align 32 62 63 subcc %o0, %o1, %o2 ! s1 == s2 ? 64 bz,pn %xcc, .stringsequal ! yup, same string, done 65 sethi %hi(0x01010101), %o5 ! magic2<31:13> 66 andcc %o0, 7, %o3 ! s1 sword-aligned ? 67 or %o5, %lo(0x01010101),%o5! magic2<31:0> 68 bz,pn %xcc, .s1aligned ! yup 69 sllx %o5, 32, %o4 ! magic2<63:32> 70 sub %o3, 8, %o3 ! number of bytes till s1 aligned 71 72.aligns1: 73 ldub [%o1 + %o2], %o0 ! s1[] 74 ldub [%o1], %g1 ! s2[] 75 subcc %o0, %g1, %o0 ! s1[] != s2[] ? 76 bne,pn %xcc, .done ! yup, done 77 addcc %o0, %g1, %g0 ! s1[] == 0 ? 78 bz,pn %xcc, .done ! yup, done 79 inccc %o3 ! s1 aligned yet? 80 bnz,pt %xcc, .aligns1 ! nope, compare another pair of bytes 81 inc %o1 ! s1++, s2++ 82 83.s1aligned: 84 andcc %o1, 7, %o3 ! s2 dword aligned ? 85 or %o5, %o4, %o5 ! magic2<63:0> 86 bz,pn %xcc, .s2aligned ! yup 87 sllx %o5, 7, %o4 ! magic1 88 sllx %o3, 3, %g5 ! leftshift = 8*ofs 89 orn %g0, %g0, %g1 ! all ones 90 and %o1, -8, %o1 ! round s1 down to next aligned dword 91 srlx %g1, %g5, %g1 ! mask for fixing up bytes 92 ldx [%o1], %o0 ! new lower dword in s2 93 orn %o0, %g1, %o0 ! force start bytes to non-zero 94 sub %g0, %g5, %g4 ! rightshift = -(8*ofs) mod 64 95 sllx %o0, %g5, %g1 ! partial unaligned word from s2 96 add %o2, %o3, %o2 ! adjust pointers 97 nop ! align loop to 16-byte boundary 98 nop ! align loop to 16-byte boundary 99 100.cmp: 101 andn %o4, %o0, %o3 ! ~word & 0x80808080 102 sub %o0, %o5, %o0 ! word - 0x01010101 103 andcc %o0, %o3, %g0 ! (word - 0x01010101) & ~word & 0x80808080 104 bz,a,pt %xcc, .doload ! no null byte in previous word from s2 105 ldx [%o1+8], %o0 ! next aligned word in s2 106.doload: 107 srlx %o0, %g4, %o3 ! bytes from aligned word from s2 108 or %g1, %o3, %g1 ! merge to get unaligned word from s2 109 ldx [%o1 + %o2], %o3 ! word from s1 110 cmp %o3, %g1 ! *s1 != *s2 ? 111 bne,pn %xcc, .wordsdiffer ! yup, find the byte that is different 112 add %o1, 8, %o1 ! s1+=8, s2+=8 113 andn %o4, %o3, %g1 ! ~word & 0x80808080 114 sub %o3, %o5, %o3 ! word - 0x01010101 115 andcc %o3, %g1, %g0 ! (word - 0x01010101) & ~word & 0x80808080 116 bz,pt %xcc, .cmp ! no null-byte in s1 yet 117 sllx %o0, %g5, %g1 ! partial unaligned word from s2 118 119 ! words are equal but the end of s1 has been reached 120 ! this means the strings must be equal 121.strequal: 122 retl ! return from leaf function 123 mov %g0, %o0 ! return 0, i.e. strings are equal 124 nop 125 126.s2aligned: 127 ldx [%o1 + %o2], %o3 ! load word from s1 128 129.cmpaligned: 130 ldx [%o1], %g1 ! load word from s2 131 cmp %o3, %g1 ! *scr1 == *src2 ? 132 bne,pn %xcc, .wordsdiffer ! nope, find mismatching character 133 add %o1, 8, %o1 ! src1 += 8, src2 += 8 134 andn %o4, %o3, %o0 ! ~word & 0x80808080 135 sub %o3, %o5, %o3 ! word - 0x01010101 136 andcc %o3, %o0, %g0 ! (word - 0x01010101) & ~word & 0x80808080 137 bz,a,pt %xcc, .cmpaligned ! no null-byte in s1 yet 138 ldx [%o1 + %o2], %o3 ! load word from s1 139 140 ! words are equal but the end of s1 has been reached 141 ! this means the strings must be equal 142 143.stringsequal: 144 retl ! return from leaf function 145 mov %g0, %o0 ! return 0, i.e. strings are equal 146 nop ! align loop on 16-byte boundary 147 nop ! align loop on 16-byte boundary 148 nop ! align loop on 16-byte boundary 149 150.wordsdiffer: 151 mov 56, %o4 ! initial shift count 152 srlx %g1, %o4, %o2 ! first byte of mismatching word in s2 153.cmpbytes: 154 srlx %o3, %o4, %o1 ! first byte of mismatching word in s1 155 subcc %o1, %o2, %o0 ! *s1-*s2 156 bnz,pn %xcc, .done ! bytes differ, return difference 157 nop 158 andcc %o1, 0xff, %o0 ! *s1 == 0 ? 159 bz,pn %xcc, .done ! yup, strings match 160 subcc %o4, 8, %o4 ! shift_count -= 8 161 bpos,pt %xcc, .cmpbytes ! until all bytes processed 162 srlx %g1, %o4, %o2 ! first byte of mismatching word in s2 163 164.done: 165 retl ! return from leaf routine 166 nop ! padding 167 168 SET_SIZE(strcmp) 169