MdePkg/Library/BaseMemoryLibOptDxe/Arm/CopyMem.asm

   1 ;------------------------------------------------------------------------------
   2 ;
   3 ; CopyMem() worker for ARM
   4 ;
   5 ; This file started out as C code that did 64 bit moves if the buffer was
   6 ; 32-bit aligned, else it does a byte copy. It also does a byte copy for
   7 ; any trailing bytes. It was updated to do 32-byte copies using stm/ldm.
   8 ;
   9 ; Copyright (c) 2008 - 2010, Apple Inc. All rights reserved.<BR>
  10 ; Copyright (c) 2016, Linaro Ltd. All rights reserved.<BR>
  11 ; This program and the accompanying materials
  12 ; are licensed and made available under the terms and conditions of the BSD License
  13 ; which accompanies this distribution.  The full text of the license may be found at
  14 ; http://opensource.org/licenses/bsd-license.php
  15 ;
  16 ; THE PROGRAM IS DISTRIBUTED UNDER THE BSD LICENSE ON AN "AS IS" BASIS,
  17 ; WITHOUT WARRANTIES OR REPRESENTATIONS OF ANY KIND, EITHER EXPRESS OR IMPLIED.
  18 ;
  19 ;------------------------------------------------------------------------------
  20
  21     EXPORT  InternalMemCopyMem
  22     AREA    SetMem, CODE, READONLY
  23     THUMB
  24
  25 InternalMemCopyMem
  26   stmfd  sp!, {r4-r11, lr}
  27   // Save the input parameters in extra registers (r11 = destination, r14 = source, r12 = length)
  28   mov  r11, r0
  29   mov  r10, r0
  30   mov  r12, r2
  31   mov  r14, r1
  32
  33 memcopy_check_overlapped
  34   cmp  r11, r1
  35   // If (dest < source)
  36   bcc  memcopy_check_optim_default
  37
  38   // If (source + length < dest)
  39   rsb  r3, r1, r11
  40   cmp  r12, r3
  41   bcc  memcopy_check_optim_default
  42   b     memcopy_check_optim_overlap
  43
  44 memcopy_check_optim_default
  45   // Check if we can use an optimized path ((length >= 32) && destination word-aligned && source word-aligned) for the memcopy (optimized path if r0 == 1)
  46   tst  r0, #0xF
  47   movne  r0, #0
  48   bne   memcopy_default
  49   tst  r1, #0xF
  50   movne  r3, #0
  51   moveq  r3, #1
  52   cmp  r2, #31
  53   movls  r0, #0
  54   andhi  r0, r3, #1
  55   b     memcopy_default
  56
  57 memcopy_check_optim_overlap
  58   // r10 = dest_end, r14 = source_end
  59   add  r10, r11, r12
  60   add  r14, r12, r1
  61
  62   // Are we in the optimized case ((length >= 32) && dest_end word-aligned && source_end word-aligned)
  63   cmp  r2, #31
  64   movls  r0, #0
  65   movhi  r0, #1
  66   tst  r10, #0xF
  67   movne  r0, #0
  68   tst  r14, #0xF
  69   movne  r0, #0
  70   b  memcopy_overlapped
  71
  72 memcopy_overlapped_non_optim
  73   // We read 1 byte from the end of the source buffer
  74   sub  r3, r14, #1
  75   sub  r12, r12, #1
  76   ldrb  r3, [r3, #0]
  77   sub  r2, r10, #1
  78   cmp  r12, #0
  79   // We write 1 byte at the end of the dest buffer
  80   sub  r10, r10, #1
  81   sub  r14, r14, #1
  82   strb  r3, [r2, #0]
  83   bne  memcopy_overlapped_non_optim
  84   b   memcopy_end
  85
  86 // r10 = dest_end, r14 = source_end
  87 memcopy_overlapped
  88   // Are we in the optimized case ?
  89   cmp  r0, #0
  90   beq  memcopy_overlapped_non_optim
  91
  92   // Optimized Overlapped - Read 32 bytes
  93   sub  r14, r14, #32
  94   sub  r12, r12, #32
  95   cmp  r12, #31
  96   ldmia  r14, {r2-r9}
  97
  98   // If length is less than 32 then disable optim
  99   movls  r0, #0
 100
 101   cmp  r12, #0
 102
 103   // Optimized Overlapped - Write 32 bytes
 104   sub  r10, r10, #32
 105   stmia  r10, {r2-r9}
 106
 107   // while (length != 0)
 108   bne  memcopy_overlapped
 109   b   memcopy_end
 110
 111 memcopy_default_non_optim
 112   // Byte copy
 113   ldrb  r3, [r14], #1
 114   sub  r12, r12, #1
 115   strb  r3, [r10], #1
 116
 117 memcopy_default
 118   cmp  r12, #0
 119   beq  memcopy_end
 120
 121 // r10 = dest, r14 = source
 122 memcopy_default_loop
 123   cmp  r0, #0
 124   beq  memcopy_default_non_optim
 125
 126   // Optimized memcopy - Read 32 Bytes
 127   sub  r12, r12, #32
 128   cmp  r12, #31
 129   ldmia  r14!, {r2-r9}
 130
 131   // If length is less than 32 then disable optim
 132   movls  r0, #0
 133
 134   cmp  r12, #0
 135
 136   // Optimized memcopy - Write 32 Bytes
 137   stmia  r10!, {r2-r9}
 138
 139   // while (length != 0)
 140   bne  memcopy_default_loop
 141
 142 memcopy_end
 143   mov  r0, r11
 144   ldmfd  sp!, {r4-r11, pc}
 145
 146   END
 147