MdePkg/BaseMemoryLibOptDxe: add accelerated ARM routines

[mirror_edk2.git] / MdePkg / Library / BaseMemoryLibOptDxe / Arm / CopyMem.S

#------------------------------------------------------------------------------\r
#\r
# CopyMem() worker for ARM\r
#\r
# This file started out as C code that did 64 bit moves if the buffer was\r
# 32-bit aligned, else it does a byte copy. It also does a byte copy for\r
# any trailing bytes. It was updated to do 32-byte copies using stm/ldm.\r
#\r
# Copyright (c) 2008 - 2010, Apple Inc. All rights reserved.<BR>\r
# Copyright (c) 2016, Linaro Ltd. 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
#------------------------------------------------------------------------------\r
\r
    .text\r
    .thumb\r
    .syntax unified\r
\r
/**\r
  Copy Length bytes from Source to Destination. Overlap is OK.\r
\r
  This implementation\r
\r
  @param  Destination Target of copy\r
  @param  Source      Place to copy from\r
  @param  Length      Number of bytes to copy\r
\r
  @return Destination\r
\r
\r
VOID *\r
EFIAPI\r
InternalMemCopyMem (\r
  OUT     VOID                      *DestinationBuffer,\r
  IN      CONST VOID                *SourceBuffer,\r
  IN      UINTN                     Length\r
  )\r
**/\r
ASM_GLOBAL ASM_PFX(InternalMemCopyMem)\r
ASM_PFX(InternalMemCopyMem):\r
    push    {r4-r11, lr}\r
    // Save the input parameters in extra registers (r11 = destination, r14 = source, r12 = length)\r
    mov     r11, r0\r
    mov     r10, r0\r
    mov     r12, r2\r
    mov     r14, r1\r
\r
    cmp     r11, r1\r
    // If (dest < source)\r
    bcc     memcopy_check_optim_default\r
\r
    // If (source + length < dest)\r
    rsb     r3, r1, r11\r
    cmp     r12, r3\r
    bcc     memcopy_check_optim_default\r
    b       memcopy_check_optim_overlap\r
\r
memcopy_check_optim_default:\r
    // Check if we can use an optimized path ((length >= 32) && destination word-aligned && source word-aligned) for the memcopy (optimized path if r0 == 1)\r
    tst     r0, #0xF\r
    it      ne\r
    movne   r0, #0\r
    bne     memcopy_default\r
    tst     r1, #0xF\r
    ite     ne\r
    movne   r3, #0\r
    moveq   r3, #1\r
    cmp     r2, #31\r
    ite     ls\r
    movls   r0, #0\r
    andhi   r0, r3, #1\r
    b       memcopy_default\r
\r
memcopy_check_optim_overlap:\r
    // r10 = dest_end, r14 = source_end\r
    add     r10, r11, r12\r
    add     r14, r12, r1\r
\r
    // Are we in the optimized case ((length >= 32) && dest_end word-aligned && source_end word-aligned)\r
    cmp     r2, #31\r
    ite     ls\r
    movls   r0, #0\r
    movhi   r0, #1\r
    tst     r10, #0xF\r
    it      ne\r
    movne   r0, #0\r
    tst     r14, #0xF\r
    it      ne\r
    movne   r0, #0\r
    b       memcopy_overlapped\r
\r
memcopy_overlapped_non_optim:\r
    // We read 1 byte from the end of the source buffer\r
    sub     r3, r14, #1\r
    sub     r12, r12, #1\r
    ldrb    r3, [r3, #0]\r
    sub     r2, r10, #1\r
    cmp     r12, #0\r
    // We write 1 byte at the end of the dest buffer\r
    sub     r10, r10, #1\r
    sub     r14, r14, #1\r
    strb    r3, [r2, #0]\r
    bne     memcopy_overlapped_non_optim\r
    b       memcopy_end\r
\r
// r10 = dest_end, r14 = source_end\r
memcopy_overlapped:\r
    // Are we in the optimized case ?\r
    cmp     r0, #0\r
    beq     memcopy_overlapped_non_optim\r
\r
    // Optimized Overlapped - Read 32 bytes\r
    sub     r14, r14, #32\r
    sub     r12, r12, #32\r
    cmp     r12, #31\r
    ldmia   r14, {r2-r9}\r
\r
    // If length is less than 32 then disable optim\r
    it      ls\r
    movls   r0, #0\r
\r
    cmp     r12, #0\r
\r
    // Optimized Overlapped - Write 32 bytes\r
    sub     r10, r10, #32\r
    stmia   r10, {r2-r9}\r
\r
    // while (length != 0)\r
    bne     memcopy_overlapped\r
    b       memcopy_end\r
\r
memcopy_default_non_optim:\r
    // Byte copy\r
    ldrb    r3, [r14], #1\r
    sub     r12, r12, #1\r
    strb    r3, [r10], #1\r
\r
memcopy_default:\r
    cmp     r12, #0\r
    beq     memcopy_end\r
\r
// r10 = dest, r14 = source\r
memcopy_default_loop:\r
    cmp     r0, #0\r
    beq     memcopy_default_non_optim\r
\r
    // Optimized memcopy - Read 32 Bytes\r
    sub     r12, r12, #32\r
    cmp     r12, #31\r
    ldmia   r14!, {r2-r9}\r
\r
    // If length is less than 32 then disable optim\r
    it      ls\r
    movls   r0, #0\r
\r
    cmp     r12, #0\r
\r
    // Optimized memcopy - Write 32 Bytes\r
    stmia   r10!, {r2-r9}\r
\r
    // while (length != 0)\r
    bne     memcopy_default_loop\r
\r
memcopy_end:\r
    mov     r0, r11\r
    pop     {r4-r11, pc}\r