Update the copyright notice format

[mirror_edk2.git] / EdkCompatibilityPkg / Foundation / Library / EfiCommonLib / Ia32 / EfiCopyMemSSE2.c

/*++\r
\r
Copyright (c) 2006, Intel Corporation. All rights reserved.<BR>\r
This program and the accompanying materials                          \r
are licensed and made available under the terms and conditions of the BSD License         \r
which accompanies this distribution.  The full text of the license may be found at        \r
http://opensource.org/licenses/bsd-license.php                                            \r
                                                                                          \r
THE PROGRAM IS DISTRIBUTED UNDER THE BSD LICENSE ON AN "AS IS" BASIS,                     \r
WITHOUT WARRANTIES OR REPRESENTATIONS OF ANY KIND, EITHER EXPRESS OR IMPLIED.             \r
\r
Module Name:\r
\r
  EfiCopyMemSSE2.c\r
\r
Abstract:\r
\r
  This is the code that supports IA32-optimized CopyMem service\r
\r
--*/\r
\r
#include "Tiano.h"\r
\r
VOID\r
EfiCommonLibCopyMem (\r
  IN VOID   *Destination,\r
  IN VOID   *Source,\r
  IN UINTN  Count\r
  )\r
/*++\r
\r
Routine Description:\r
\r
  Copy Length bytes from Source to Destination.\r
\r
Arguments:\r
\r
  Destination - Target of copy\r
\r
  Source      - Place to copy from\r
\r
  Length      - Number of bytes to copy\r
\r
Returns:\r
\r
  None\r
\r
--*/\r
{\r
  __asm {\r
  mov   ecx, Count\r
  mov   esi, Source\r
  mov   edi, Destination\r
\r
  ; First off, make sure we have no overlap. That is to say,\r
  ;   if (Source == Destination)           => do nothing\r
  ;   if (Source + Count <= Destination)   => regular copy\r
  ;   if (Destination + Count <= Source)   => regular copy\r
  ;   otherwise, do a reverse copy\r
  mov   eax, esi\r
  add   eax, ecx                      ; Source + Count\r
  cmp   eax, edi\r
  jle   _StartByteCopy\r
\r
  mov   eax, edi\r
  add   eax, ecx                      ; Dest + Count\r
  cmp   eax, esi\r
  jle   _StartByteCopy\r
\r
  cmp   esi, edi\r
  je    _CopyMemDone         \r
  jl    _CopyOverlapped               ; too bad -- overlaps\r
\r
  ; Pick up misaligned start bytes to get destination pointer 4-byte aligned\r
_StartByteCopy:\r
  cmp   ecx, 0\r
  je    _CopyMemDone                ; Count == 0, all done\r
  mov   edx, edi\r
  and   dl, 3                       ; check lower 2 bits of address\r
  test  dl, dl          \r
  je    SHORT _CopyBlocks           ; already aligned?\r
\r
  ; Copy a byte\r
  mov   al, BYTE PTR [esi]          ; get byte from Source\r
  mov   BYTE PTR [edi], al          ; write byte to Destination\r
  dec    ecx\r
  inc   edi\r
  inc   esi\r
  jmp   _StartByteCopy               ; back to top of loop\r
\r
_CopyBlocks:\r
  ; Compute how many 64-byte blocks we can clear \r
  mov   eax, ecx                    ; get Count in eax\r
  shr   eax, 6                      ; convert to 64-byte count\r
  shl   eax, 6                      ; convert back to bytes\r
  sub   ecx, eax                    ; subtract from the original count\r
  shr   eax, 6                      ; and this is how many 64-byte blocks\r
\r
  ; If no 64-byte blocks, then skip \r
  cmp   eax, 0\r
  je    _CopyRemainingDWords\r
\r
\r
copyxmm:\r
  \r
  movdqu xmm0, OWORD PTR ds:[esi]\r
  movdqu QWORD PTR ds:[edi], xmm0\r
  movdqu xmm1, OWORD PTR ds:[esi+16]\r
  movdqu QWORD PTR ds:[edi+16], xmm1\r
  movdqu xmm2, OWORD PTR ds:[esi+32]\r
  movdqu QWORD PTR ds:[edi+32], xmm2\r
  movdqu xmm3, OWORD PTR ds:[esi+48]\r
  movdqu QWORD PTR ds:[edi+48], xmm3\r
  \r
  add   edi, 64\r
  add   esi, 64\r
  dec   eax\r
  jnz   copyxmm\r
  \r
\r
  ; Copy as many DWORDS as possible\r
_CopyRemainingDWords:\r
  cmp   ecx, 4\r
  jb    _CopyRemainingBytes\r
\r
  mov   eax, DWORD PTR [esi]        ; get data from Source\r
  mov   DWORD PTR [edi], eax        ; write byte to Destination\r
  sub   ecx, 4                      ; decrement Count\r
  add   esi, 4                      ; advance Source pointer\r
  add   edi, 4                      ; advance Destination pointer\r
  jmp   _CopyRemainingDWords        ; back to top\r
\r
_CopyRemainingBytes:\r
  cmp   ecx, 0\r
  je    _CopyMemDone\r
  mov   al, BYTE PTR [esi]          ; get byte from Source\r
  mov   BYTE PTR [edi], al          ; write byte to Destination\r
  dec    ecx\r
  inc    esi\r
  inc   edi                      ; advance Destination pointer\r
  jmp   SHORT _CopyRemainingBytes   ; back to top of loop\r
\r
  ;\r
  ; We do this block if the source and destination buffers overlap. To\r
  ; handle it, copy starting at the end of the source buffer and work\r
  ; your way back. Since this is the atypical case, this code has not\r
  ; been optimized, and thus simply copies bytes.\r
  ;\r
_CopyOverlapped:\r
   \r
  ; Move the source and destination pointers to the end of the range\r
  add   esi, ecx                      ; Source + Count\r
  dec    esi\r
  add   edi, ecx                      ; Dest + Count\r
  dec    edi\r
\r
_CopyOverlappedLoop:\r
  cmp   ecx, 0\r
  je    _CopyMemDone\r
  mov   al, BYTE PTR [esi]          ; get byte from Source\r
  mov   BYTE PTR [edi], al          ; write byte to Destination\r
  dec    ecx\r
  dec    esi\r
  dec   edi\r
  jmp   _CopyOverlappedLoop         ; back to top of loop\r
\r
_CopyMemDone:\r
  }\r
}\r