[444] | 1 | /******************************************************************************* |
---|
| 2 | * |
---|
| 3 | * Copyright (c) 1993 Intel Corporation |
---|
| 4 | * |
---|
| 5 | * Intel hereby grants you permission to copy, modify, and distribute this |
---|
| 6 | * software and its documentation. Intel grants this permission provided |
---|
| 7 | * that the above copyright notice appears in all copies and that both the |
---|
| 8 | * copyright notice and this permission notice appear in supporting |
---|
| 9 | * documentation. In addition, Intel grants this permission provided that |
---|
| 10 | * you prominently mark as "not part of the original" any modifications |
---|
| 11 | * made to this software or documentation, and that the name of Intel |
---|
| 12 | * Corporation not be used in advertising or publicity pertaining to |
---|
| 13 | * distribution of the software or the documentation without specific, |
---|
| 14 | * written prior permission. |
---|
| 15 | * |
---|
| 16 | * Intel Corporation provides this AS IS, WITHOUT ANY WARRANTY, EXPRESS OR |
---|
| 17 | * IMPLIED, INCLUDING, WITHOUT LIMITATION, ANY WARRANTY OF MERCHANTABILITY |
---|
| 18 | * OR FITNESS FOR A PARTICULAR PURPOSE. Intel makes no guarantee or |
---|
| 19 | * representations regarding the use of, or the results of the use of, |
---|
| 20 | * the software and documentation in terms of correctness, accuracy, |
---|
| 21 | * reliability, currentness, or otherwise; and you rely on the software, |
---|
| 22 | * documentation and results solely at your own risk. |
---|
| 23 | * |
---|
| 24 | * IN NO EVENT SHALL INTEL BE LIABLE FOR ANY LOSS OF USE, LOSS OF BUSINESS, |
---|
| 25 | * LOSS OF PROFITS, INDIRECT, INCIDENTAL, SPECIAL OR CONSEQUENTIAL DAMAGES |
---|
| 26 | * OF ANY KIND. IN NO EVENT SHALL INTEL'S TOTAL LIABILITY EXCEED THE SUM |
---|
| 27 | * PAID TO INTEL FOR THE PRODUCT LICENSED HEREUNDER. |
---|
| 28 | * |
---|
| 29 | ******************************************************************************/ |
---|
| 30 | |
---|
| 31 | .file "memccpy.s" |
---|
| 32 | #ifdef __PIC |
---|
| 33 | .pic |
---|
| 34 | #endif |
---|
| 35 | #ifdef __PID |
---|
| 36 | .pid |
---|
| 37 | #endif |
---|
| 38 | /* |
---|
| 39 | * (c) copyright 1989,1993 Intel Corp., all rights reserved |
---|
| 40 | */ |
---|
| 41 | /* |
---|
| 42 | procedure memccpy (optimized assembler version for the 80960K series) |
---|
| 43 | |
---|
| 44 | dest_addr = memccpy (dest_addr, src_addr, char, len) |
---|
| 45 | |
---|
| 46 | copy len bytes pointed to by src_addr to the space pointed to by |
---|
| 47 | dest_addr, stopping if char is copied. If char is copied, |
---|
| 48 | return address of byte after char in dest string; else null. |
---|
| 49 | |
---|
| 50 | |
---|
| 51 | Undefined behavior will occur if the end of the source array is in |
---|
| 52 | the last two words of the program's allocated memory space. This |
---|
| 53 | is so because the routine fetches ahead. Disallowing the fetch |
---|
| 54 | ahead would impose a severe performance penalty. |
---|
| 55 | |
---|
| 56 | Undefined behavior will also occur if the source and destination |
---|
| 57 | strings overlap. |
---|
| 58 | |
---|
| 59 | Strategy: |
---|
| 60 | |
---|
| 61 | Fetch the source array by words and store them by words to the |
---|
| 62 | destination array, until there are fewer than three bytes left |
---|
| 63 | to copy. Then, using the last word of the source (the one that |
---|
| 64 | contains the remaining 0, 1, 2, or 3 bytes to be copied), store |
---|
| 65 | a byte at a time until Ldone. |
---|
| 66 | |
---|
| 67 | Tactics: |
---|
| 68 | |
---|
| 69 | 1) Do NOT try to fetch and store the words in a word aligned manner |
---|
| 70 | because, in my judgement, the performance degradation experienced due |
---|
| 71 | to non-aligned accesses does NOT outweigh the time and complexity added |
---|
| 72 | by the preamble and convoluted body that would be necessary to assure |
---|
| 73 | alignment. This is supported by the intuition that most source and |
---|
| 74 | destination arrays (even more true of most big source arrays) will |
---|
| 75 | be word aligned to begin with. |
---|
| 76 | |
---|
| 77 | 2) Rather than decrementing len to zero, |
---|
| 78 | I calculate the address of the byte after the last byte of the |
---|
| 79 | destination array, and quit when the destination byte pointer passes |
---|
| 80 | that. |
---|
| 81 | |
---|
| 82 | */ |
---|
| 83 | |
---|
| 84 | .globl _memccpy |
---|
| 85 | .leafproc _memccpy, __memccpy |
---|
| 86 | .align 2 |
---|
| 87 | _memccpy: |
---|
| 88 | #ifndef __PIC |
---|
| 89 | lda Lrett,g14 |
---|
| 90 | #else |
---|
| 91 | lda Lrett-(.+8)(ip),g14 |
---|
| 92 | #endif |
---|
| 93 | __memccpy: |
---|
| 94 | mov g14, g13 # preserve return address |
---|
| 95 | cmpibge 0,g3,Lexit_char_not_found |
---|
| 96 | |
---|
| 97 | addo g3,g1,g3 # compute beyond end of src |
---|
| 98 | ld (g1), g7 # fetch first word of source |
---|
| 99 | lda 0xff,g5 # mask for char |
---|
| 100 | and g5,g2,g2 # extract only char |
---|
| 101 | shlo 8,g2,g6 |
---|
| 102 | or g2,g6,g6 |
---|
| 103 | shlo 16,g6,g4 |
---|
| 104 | or g6,g4,g6 # word of char |
---|
| 105 | b Lwloop_b |
---|
| 106 | |
---|
| 107 | Lwloop_a: |
---|
| 108 | ld (g1), g7 # fetch ahead next word of source |
---|
| 109 | st g4, (g0) # store word to dest |
---|
| 110 | addo 4, g0, g0 # post-increment dest pointer |
---|
| 111 | Lwloop_b: # word copying loop |
---|
| 112 | addo 4, g1, g1 # pre-increment src pointer |
---|
| 113 | cmpo g3, g1 # is len <= 3 ? |
---|
| 114 | mov g7, g4 # keep a copy of the current word |
---|
| 115 | bl Lcloop_setup # quit word loop if less than 4 bytes |
---|
| 116 | scanbyte g6, g7 # check for char |
---|
| 117 | bno Lwloop_a # continue word loop if char not found. |
---|
| 118 | |
---|
| 119 | Lcloop_setup: |
---|
| 120 | subo 4, g1, g1 # back down src pointer |
---|
| 121 | cmpobe g1, g3, Lexit_char_not_found |
---|
| 122 | |
---|
| 123 | Lcloop_a: # character copying loop (len < 3) |
---|
| 124 | and g5,g4,g7 # check the byte against char |
---|
| 125 | cmpo g7,g2 |
---|
| 126 | stob g7,(g0) # store the byte |
---|
| 127 | addo 1, g0, g0 |
---|
| 128 | be Lexit_char_found |
---|
| 129 | addo 1,g1,g1 |
---|
| 130 | cmpo g1,g3 |
---|
| 131 | shro 8,g4,g4 # position next byte |
---|
| 132 | bne Lcloop_a |
---|
| 133 | |
---|
| 134 | Lexit_char_not_found: |
---|
| 135 | mov 0, g0 |
---|
| 136 | Lexit_char_found: |
---|
| 137 | lda 0,g14 |
---|
| 138 | bx (g13) # g0 = dest array address; g14 = 0 |
---|
| 139 | Lrett: |
---|
| 140 | ret |
---|
| 141 | |
---|
| 142 | |
---|
| 143 | /* end of memccpy */ |
---|