BaseMemoryLibSse2: Take advantage of write combining buffers

[mirror_edk2.git] / MdePkg / Library / BaseMemoryLibSse2 / X64 / SetMem16.nasm

;------------------------------------------------------------------------------\r
;\r
; Copyright (c) 2006, Intel Corporation. All rights reserved.<BR>\r
; SPDX-License-Identifier: BSD-2-Clause-Patent\r
;\r
; Module Name:\r
;\r
;   SetMem16.nasm\r
;\r
; Abstract:\r
;\r
;   SetMem16 function\r
;\r
; Notes:\r
;\r
;------------------------------------------------------------------------------\r
\r
    DEFAULT REL\r
    SECTION .text\r
\r
;------------------------------------------------------------------------------\r
;  VOID *\r
;  InternalMemSetMem16 (\r
;    IN VOID   *Buffer,\r
;    IN UINTN  Count,\r
;    IN UINT16 Value\r
;    )\r
;------------------------------------------------------------------------------\r
global ASM_PFX(InternalMemSetMem16)\r
ASM_PFX(InternalMemSetMem16):\r
    push    rdi\r
    mov     rdi, rcx\r
    mov     r9, rdi\r
    xor     rcx, rcx\r
    sub     rcx, rdi\r
    and     rcx, 63\r
    mov     rax, r8\r
    jz      .0\r
    shr     rcx, 1\r
    cmp     rcx, rdx\r
    cmova   rcx, rdx\r
    sub     rdx, rcx\r
    rep     stosw\r
.0:\r
    mov     rcx, rdx\r
    and     edx, 31\r
    shr     rcx, 5\r
    jz      @SetWords\r
    movd    xmm0, eax\r
    pshuflw xmm0, xmm0, 0\r
    movlhps xmm0, xmm0\r
.1:\r
    movntdq [rdi], xmm0\r
    movntdq [rdi + 16], xmm0\r
    movntdq [rdi + 32], xmm0\r
    movntdq [rdi + 48], xmm0\r
    add     rdi, 64\r
    loop    .1\r
    mfence\r
@SetWords:\r
    mov     ecx, edx\r
    rep     stosw\r
    mov     rax, r9\r
    pop     rdi\r
    ret\r
\r