1;------------------------------------------------------------------------------
2;
3; Copyright (c) 2006, Intel Corporation. All rights reserved.<BR>
4; This program and the accompanying materials
5; are licensed and made available under the terms and conditions of the BSD License
6; which accompanies this distribution.  The full text of the license may be found at
7; http://opensource.org/licenses/bsd-license.php.
8;
9; THE PROGRAM IS DISTRIBUTED UNDER THE BSD LICENSE ON AN "AS IS" BASIS,
10; WITHOUT WARRANTIES OR REPRESENTATIONS OF ANY KIND, EITHER EXPRESS OR IMPLIED.
11;
12; Module Name:
13;
14;   CopyMem.asm
15;
16; Abstract:
17;
18;   CopyMem function
19;
20; Notes:
21;
22;------------------------------------------------------------------------------
23
24    .686
25    .model  flat,C
26    .xmm
27    .code
28
29;------------------------------------------------------------------------------
30;  VOID *
31;  InternalMemCopyMem (
32;    IN VOID   *Destination,
33;    IN VOID   *Source,
34;    IN UINTN  Count
35;    );
36;------------------------------------------------------------------------------
37InternalMemCopyMem  PROC    USES    esi edi
38    mov     esi, [esp + 16]             ; esi <- Source
39    mov     edi, [esp + 12]             ; edi <- Destination
40    mov     edx, [esp + 20]             ; edx <- Count
41    lea     eax, [esi + edx - 1]        ; eax <- End of Source
42    cmp     esi, edi
43    jae     @F
44    cmp     eax, edi                    ; Overlapped?
45    jae     @CopyBackward               ; Copy backward if overlapped
46@@:
47    xor     ecx, ecx
48    sub     ecx, edi
49    and     ecx, 15                     ; ecx + edi aligns on 16-byte boundary
50    jz      @F
51    cmp     ecx, edx
52    cmova   ecx, edx
53    sub     edx, ecx                    ; edx <- remaining bytes to copy
54    rep     movsb
55@@:
56    mov     ecx, edx
57    and     edx, 15
58    shr     ecx, 4                      ; ecx <- # of DQwords to copy
59    jz      @CopyBytes
60    add     esp, -16
61    movdqu  [esp], xmm0                 ; save xmm0
62@@:
63    movdqu  xmm0, [esi]                 ; esi may not be 16-bytes aligned
64    movntdq [edi], xmm0                 ; edi should be 16-bytes aligned
65    add     esi, 16
66    add     edi, 16
67    loop    @B
68    mfence
69    movdqu  xmm0, [esp]                 ; restore xmm0
70    add     esp, 16                     ; stack cleanup
71    jmp     @CopyBytes
72@CopyBackward:
73    mov     esi, eax                    ; esi <- Last byte in Source
74    lea     edi, [edi + edx - 1]        ; edi <- Last byte in Destination
75    std
76@CopyBytes:
77    mov     ecx, edx
78    rep     movsb
79    cld
80    mov     eax, [esp + 12]             ; eax <- Destination as return value
81    ret
82InternalMemCopyMem  ENDP
83
84    END
85