osfmk/ppc/commpage/memset_64.s

   1 /*
   2  * Copyright (c) 2003 Apple Computer, Inc. All rights reserved.
   3  *
   4  * @APPLE_OSREFERENCE_LICENSE_HEADER_START@
   5  *
   6  * This file contains Original Code and/or Modifications of Original Code
   7  * as defined in and that are subject to the Apple Public Source License
   8  * Version 2.0 (the 'License'). You may not use this file except in
   9  * compliance with the License. The rights granted to you under the License
  10  * may not be used to create, or enable the creation or redistribution of,
  11  * unlawful or unlicensed copies of an Apple operating system, or to
  12  * circumvent, violate, or enable the circumvention or violation of, any
  13  * terms of an Apple operating system software license agreement.
  14  *
  15  * Please obtain a copy of the License at
  16  * http://www.opensource.apple.com/apsl/ and read it before using this file.
  17  *
  18  * The Original Code and all software distributed under the License are
  19  * distributed on an 'AS IS' basis, WITHOUT WARRANTY OF ANY KIND, EITHER
  20  * EXPRESS OR IMPLIED, AND APPLE HEREBY DISCLAIMS ALL SUCH WARRANTIES,
  21  * INCLUDING WITHOUT LIMITATION, ANY WARRANTIES OF MERCHANTABILITY,
  22  * FITNESS FOR A PARTICULAR PURPOSE, QUIET ENJOYMENT OR NON-INFRINGEMENT.
  23  * Please see the License for the specific language governing rights and
  24  * limitations under the License.
  25  *
  26  * @APPLE_OSREFERENCE_LICENSE_HEADER_END@
  27  */
  28
  29 #include <sys/appleapiopts.h>
  30 #include <ppc/asm.h>
  31 #include <machine/cpu_capabilities.h>
  32 #include <machine/commpage.h>
  33
  34 /*
  35  * WARNING: this code is written for 32-bit mode, and ported by the kernel if necessary
  36  * to 64-bit mode for use in the 64-bit commpage.  This "port" consists of the following
  37  * simple transformations:
  38  *      - all word compares are changed to doubleword
  39  *      - all "srwi[.]" opcodes are changed to "srdi[.]"
  40  * Nothing else is done.  For this to work, the following rules must be
  41  * carefully followed:
  42  *      - do not use carry or overflow
  43  *      - only use record mode if you are sure the results are mode-invariant
  44  *        for example, all "andi." and almost all "rlwinm." are fine
  45  *      - do not use "slwi", "slw", or "srw"
  46  * An imaginative programmer could break the porting model in other ways, but the above
  47  * are the most likely problem areas.  It is perhaps surprising how well in practice
  48  * this simple method works.
  49  */
  50
  51         .text
  52         .align  2
  53
  54
  55 /* *********************
  56  * * M E M S E T _ 6 4 *
  57  * *********************
  58  *
  59  * This is a subroutine called by Libc memset and _memset_pattern for large nonzero
  60  * operands (zero operands are funneled into bzero.)  This version is for a
  61  * hypothetic processor that is 64-bit but not Altivec.
  62  * It is not optimized, since it would only be used during bringup.
  63  *
  64  * Registers at entry:
  65  *              r4 = count of bytes to store (must be >= 32)
  66  *      r8 = ptr to the 1st byte to store (16-byte aligned)
  67  *      r9 = ptr to 16-byte pattern to store (16-byte aligned)
  68  * When we return:
  69  *              r3 = not changed, since memset returns it
  70  *      r4 = bytes remaining to store (will be <32)
  71  *      r7 = not changed
  72  *      r8 = ptr to next byte to store (still 16-byte aligned)
  73  *     r12 = not changed (holds return value for memset)
  74  */
  75
  76 memset_64:
  77         srwi    r0,r4,5                 // get number of 32-byte chunks (>0)
  78         ld      r10,0(r9)               // load pattern
  79         ld      r11,8(r9)
  80         rlwinm  r4,r4,0,0x1F            // mask down count
  81         mtctr   r0                      // set up loop count
  82
  83         // Loop over 32-byte chunks.
  84 1:
  85         std     r10,0(r8)
  86         std     r11,8(r8)
  87         std     r10,16(r8)
  88         std     r11,24(r8)
  89         addi    r8,r8,32
  90         bdnz++  1b
  91
  92         blr
  93
  94
  95         COMMPAGE_DESCRIPTOR(memset_64,_COMM_PAGE_MEMSET_PATTERN,k64Bit,kHasAltivec, \
  96                                 kCommPageBoth+kPort32to64)