gen/libsa/memset.c

   1 /*
   2  * Copyright (c) 1999 Apple Computer, Inc. All rights reserved.
   3  *
   4  * @APPLE_LICENSE_HEADER_START@
   5  *
   6  * Copyright (c) 1999-2003 Apple Computer, Inc.  All Rights Reserved.
   7  *
   8  * This file contains Original Code and/or Modifications of Original Code
   9  * as defined in and that are subject to the Apple Public Source License
  10  * Version 2.0 (the 'License'). You may not use this file except in
  11  * compliance with the License. Please obtain a copy of the License at
  12  * http://www.opensource.apple.com/apsl/ and read it before using this
  13  * file.
  14  *
  15  * The Original Code and all software distributed under the License are
  16  * distributed on an 'AS IS' basis, WITHOUT WARRANTY OF ANY KIND, EITHER
  17  * EXPRESS OR IMPLIED, AND APPLE HEREBY DISCLAIMS ALL SUCH WARRANTIES,
  18  * INCLUDING WITHOUT LIMITATION, ANY WARRANTIES OF MERCHANTABILITY,
  19  * FITNESS FOR A PARTICULAR PURPOSE, QUIET ENJOYMENT OR NON-INFRINGEMENT.
  20  * Please see the License for the specific language governing rights and
  21  * limitations under the License.
  22  *
  23  * @APPLE_LICENSE_HEADER_END@
  24  */
  25 /* Copyright (c) 1991 NeXT Computer, Inc.  All rights reserved.
  26  *
  27  *      File:   libc/i386/ansi/memset.c
  28  *      Author: Bruce Martin, NeXT Computer, Inc.
  29  *
  30  *      This file contains machine dependent code for block compares
  31  *      on NeXT i386-based products.  Currently tuned for the i486.
  32  *
  33  *      This code is loosely based on some work done by Mike DeMoney
  34  *      for the m88k.
  35  *
  36  * HISTORY
  37  * 14-Dec-93  Curtis Galloway
  38  *      Modified for the booter; simplified simple_char_set.
  39  *
  40  * 14-Aug-92  Bruce Martin (Bruce_Martin@NeXT.COM)
  41  *      Created.
  42  */
  43
  44 #define MIN_TO_ALIGN    32
  45
  46 /*
  47  * Convert a void * into a unsigned int so arithmetic can be done
  48  * (Technically, the gnu C compiler is tolerant of arithmetic
  49  * on void *, but ansi isn't; so we do this.)
  50  */
  51 #define UNS(voidp)      ((unsigned int)(voidp))
  52
  53 /*
  54  * Number of bytes addr is past an object of 'align' size.
  55  */
  56 #define BYTES_PAST_ALIGNMENT(voidp, align) \
  57         (UNS(voidp) & (align - 1))
  58
  59 /*
  60  * Bytes moved by an unrolled loop
  61  */
  62 #define LOOP_STRIDE(type, unroll)       ((unroll) * sizeof(type))
  63
  64
  65 /*
  66  * Len modulo LOOP_STRIDE
  67  */
  68 #define MODULO_LOOP_UNROLL(len, type, unroll)   \
  69         ((len) & (LOOP_STRIDE(type, unroll) - 1) & ~(sizeof(type) - 1))
  70
  71
  72 /*
  73  * Convert a void * + offset into char, short, int, or long long reference
  74  * based at that address
  75  */
  76 #define CHAR(voidp, offset)     (*(char *)(UNS(voidp) + (offset)))
  77 #define SHORT(voidp, offset)    (*(short *)(UNS(voidp) + (offset)))
  78 #define INT(voidp, offset)      (*(int *)(UNS(voidp) + (offset)))
  79
  80
  81
  82 static inline void
  83 simple_char_set(char *dst, int val, int len)
  84 {
  85     asm("rep; stosb"
  86         : /* no outputs */
  87         : "&c" (len), "D" (dst), "a" (val)
  88         : "ecx", "edi", "eax");
  89 }
  90
  91
  92
  93 void *memset(void *, int, unsigned long);
  94
  95 int bzero(void *dst, unsigned long ulen)
  96 {
  97     (void) memset(dst, 0, ulen);
  98     return 0;
  99 }
 100
 101
 102 void *memset(void *dst, int val, unsigned long ulen)
 103 {
 104     int len = ulen;
 105     void *orig_dst = dst;
 106     int need_to_set;
 107     int alignment;
 108
 109     val |= (val << 8);
 110     val |= (val << 16);
 111
 112     /* Always do a forward copy */
 113
 114     if (len < MIN_TO_ALIGN) {
 115         simple_char_set(dst, val, len);
 116         return orig_dst;
 117     }
 118
 119     if (need_to_set = BYTES_PAST_ALIGNMENT(dst, 2 * sizeof(int))) {
 120         need_to_set = (2 * sizeof(int)) - need_to_set;
 121         simple_char_set(dst, val, need_to_set);
 122         len -= need_to_set;
 123         UNS(dst) += need_to_set;
 124     }
 125
 126     alignment = MODULO_LOOP_UNROLL(len, int, 8);
 127     UNS(dst) += alignment - LOOP_STRIDE(int, 8);
 128     switch (alignment) {
 129             do {
 130                         INT(dst, 0) = val;
 131         case 28:        INT(dst, 4) = val;
 132         case 24:        INT(dst, 8) = val;
 133         case 20:        INT(dst, 12) = val;
 134         case 16:        INT(dst, 16) = val;
 135         case 12:        INT(dst, 20) = val;
 136         case 8:         INT(dst, 24) = val;
 137         case 4:         INT(dst, 28) = val;
 138         case 0:
 139                         UNS(dst) += LOOP_STRIDE(int, 8);
 140                         len -= LOOP_STRIDE(int, 8);
 141             } while (len >= 0);
 142         len &= sizeof(int) - 1;
 143     }
 144
 145     switch (len) {
 146         case 3: CHAR(dst, 2) = val;
 147         case 2: CHAR(dst, 1) = val;
 148         case 1: CHAR(dst, 0) = val;
 149     }
 150
 151     return orig_dst;
 152 }