/* * Copyright (c) 2003 Apple Computer, Inc. All rights reserved. * * @APPLE_LICENSE_HEADER_START@ * * The contents of this file constitute Original Code as defined in and * are subject to the Apple Public Source License Version 1.1 (the * "License"). You may not use this file except in compliance with the * License. Please obtain a copy of the License at * http://www.apple.com/publicsource and read it before using this file. * * This Original Code and all software distributed under the License are * distributed on an "AS IS" basis, WITHOUT WARRANTY OF ANY KIND, EITHER * EXPRESS OR IMPLIED, AND APPLE HEREBY DISCLAIMS ALL SUCH WARRANTIES, * INCLUDING WITHOUT LIMITATION, ANY WARRANTIES OF MERCHANTABILITY, * FITNESS FOR A PARTICULAR PURPOSE OR NON-INFRINGEMENT. Please see the * License for the specific language governing rights and limitations * under the License. * * @APPLE_LICENSE_HEADER_END@ */ /* We use mode-independent "g" opcodes such as "srgi". These expand * into word operations when targeting __ppc__, and into doubleword * operations when targeting __ppc64__. */ #include <architecture/ppc/mode_independent_asm.h> #include <mach/ppc/asm.h> #define __APPLE_API_PRIVATE #include <machine/cpu_capabilities.h> #undef __APPLE_API_PRIVATE // Strlen, optimized for PPC. We use an inobvious but very efficient // word-parallel test for 0-bytes: // // y = dataWord + 0xFEFEFEFF // z = ~dataWord & 0x80808080 // if ( y & z ) = 0 then all bytes in dataWord are non-zero // // The test maps any non-zero byte to zeros and any zero byte to 0x80, // with one exception: 0x01 bytes preceeding the first zero are also // mapped to 0x80. Using altivec is another possibility, but it turns // out that the overhead of maintaining VRSAVE and dealing with edge // cases pushes the crossover point out to around 30 bytes... longer // the the "typical" operand length. // // In 64-bit mode, the algorithm is doubleword parallel. .text .align 5 .globl EXT(strlen) LEXT(strlen) // int strlen(ptr) clrrgi r9,r3,LOG2_GPR_BYTES// align pointer by zeroing right LOG2_GPR_BYTES bits li r7,-1 // get 0xFFs lg r8,0(r9) // get word or doubleword with 1st operand byte rlwinm r4,r3,3,(GPR_BYTES-1)*8 // get starting bit position of operand #if defined(__ppc__) lis r5,hi16(0xFEFEFEFF) // start to generate 32-bit magic constants lis r6,hi16(0x80808080) srw r7,r7,r4 // create a mask of 0xFF bytes for operand in r8 ori r5,r5,lo16(0xFEFEFEFF) ori r6,r6,lo16(0x80808080) #else ld r5,_COMM_PAGE_MAGIC_FE(0) // get 0xFEFEFEFE FEFEFEFF from commpage ld r6,_COMM_PAGE_MAGIC_80(0) // get 0x80808080 80808080 from commpage srd r7,r7,r4 // create a mask of 0xFF bytes for operand in r8 #endif orc r8,r8,r7 // make sure bytes preceeding operand are 0xFF b Lloop1 // enter loop // Loop over words or doublewords. // r3 = original address // r5 = 0xFEFEFEFE FEFEFEFF // r6 = 0x80808080 80808080 // r9 = address (aligned) .align 5 Lloop: lgu r8,GPR_BYTES(r9) // get next word or doubleword Lloop1: // initial entry add r4,r5,r8 // r4 = data + 0xFEFEFEFF andc r7,r6,r8 // r7 = ~data & 0x80808080 and. r4,r4,r7 // r4 = r4 & r7 beq Lloop // if r4 is zero, then all bytes are non-zero // Now we know one of the bytes in r8 is zero, we just have to figure out which one. // We have mapped 0 bytes to 0x80, and nonzero bytes to 0x00, with one exception: // 0x01 bytes preceeding the first zero are also mapped to 0x80. So we have to mask // out the 0x80s caused by 0x01s before searching for the 0x80 byte. slgi r5,r8,7 // move 0x01 bits to 0x80 position sub r3,r9,r3 // start to compute string length andc r4,r4,r5 // turn off false hits from 0x0100 worst case cntlzg r7,r4 // now we can count leading 0s srwi r7,r7,3 // convert 0,8,16,24 to 0,1,2,3, etc add r3,r3,r7 // add in nonzero bytes in last word blr