// Copyright (c) 2007-2009 Nokia Corporation and/or its subsidiary(-ies).
// All rights reserved.
// This component and the accompanying materials are made available
// under the terms of the License "Eclipse Public License v1.0"
// which accompanies this distribution, and is available
// at the URL "http://www.eclipse.org/legal/epl-v10.html".
//
// Initial Contributors:
// Nokia Corporation - initial contribution.
//
// Contributors:
//
// Description:
// e32\common\x86\cmem.cia
// 
//

#include "common.h"


#if defined(__MEM_MACHINE_CODED__)
extern "C" {


// See header file e32cmn.h for the in-source documentation.
EXPORT_C __NAKED__ TAny* memmove(TAny* /*aTrg*/, const TAny* /*aSrc*/, unsigned int /*aLength*/)
	{
	asm("push ebx");
	asm("push esi");
	asm("push edi");
	asm("mov edi,[esp+16]");// destination address into edi
	asm("mov esi,[esp+20]");// source address into esi
	asm("mov ecx,[esp+24]");// byte count into ecx
	asm("pushfd");

	asm("mov eax,edi");		// return value
	asm("cld");				// assume forward copy initially
	asm("test ecx,ecx");	//
	asm("jz short memcopy0");// if length=0, nothing to do
	asm("xor edx,edx");		//
	asm("cmp edi,esi");		// compare source and dest addresses
	asm("je short memcopy0");// if equal, nothing to do
	asm("jc short memcopy1");// if dest<source, must go forwards
	asm("std");				// else go backwards
	asm("add esi,ecx");		// and start at end of block
	asm("add edi,ecx");		//
	asm("inc edx");			// edx=1 if backwards, 0 if forwards
	asm("memcopy1:");
	asm("cmp ecx,16");		// if length<16 don't bother with alignment check
	asm("jc short memcopy2");//
	asm("mov ebx,edi");		// ebx = destination address
	asm("and ebx,3");		// ebx bottom 2 bits = alignment of destination wrt hardware bus
	asm("jz short memcopy3");// if aligned, proceed with block move
	asm("or edx,edx");		// check direction of move
	asm("jnz short memcopy4");// if backwards, ebx = number of byte moves to align destination
	asm("neg ebx");			// else number of byte moves = 4-ebx
	asm("add ebx,4");		//
	asm("memcopy4:");
	asm("sub ecx,ebx");		// subtract number of bytes from length
	asm("xchg ecx,ebx");	// temporarily put length in ebx
	asm("sub edi,edx");		// adjust if backwards move
	asm("sub esi,edx");		//
	asm("rep movsb");		// move bytes to align destination
	asm("add edi,edx");		// adjust if backwards move
	asm("add esi,edx");		//
	asm("mov ecx,ebx");		// length back into ecx
	asm("memcopy3:");
	asm("mov ebx,ecx");		// save length in ebx
	asm("shl edx,2");		// adjustment 4 for backwards move
	asm("shr ecx,2");		// number of dwords to move into ecx
	asm("sub edi,edx");		// adjust if backwards move
	asm("sub esi,edx");		//
	asm("rep movsd");		// perform DWORD block move
	asm("add edi,edx");		// adjust if backwards move
	asm("add esi,edx");		//
	asm("mov ecx,ebx");		// length back into ecx
	asm("and ecx,3");		// number of remaining bytes to move
	asm("jz short memcopy0");// if zero, we are finished
	asm("shr edx,2");		// adjustment 1 for backwards move
	asm("memcopy2:");		// *** come here for small move
	asm("sub edi,edx");		// adjust if backwards move
	asm("sub esi,edx");		//
	asm("rep movsb");			// move remaining bytes

	asm("memcopy0:");
	asm("popfd");
	asm("pop edi");
	asm("pop esi");
	asm("pop ebx");
	asm("ret");				// finished - return value in EAX
	}


// See header file e32cmn.h for the in-source documentation.
EXPORT_C __NAKED__ TAny* memcpy(TAny* /*aTrg*/, const TAny* /*aSrc*/, unsigned int /*aLength*/)
	{
	asm("jmp %a0": :"i"(&memmove));
	}




// See header file e32cmn.h for the in-source documentation.
EXPORT_C __NAKED__ TAny* wordmove(TAny* /*aTrg*/, const TAny* /*aSrc*/, unsigned int /*aLength*/)
	{
	asm("push esi");
	asm("push edi");
	asm("mov edi,[esp+12]");// destination address into edi
	asm("mov esi,[esp+16]");// source address into esi
	asm("mov ecx,[esp+20]");// byte count into ecx
	asm("pushfd");

	asm("mov eax,edi");		// return value
	asm("cld");				//
	asm("test ecx,ecx");	//
	asm("jz short memmove0");// if length=0, nothing to do
	asm("cmp edi,esi");		// compare source and dest addresses
	asm("je short memmove0");// if equal, nothing to do
	asm("jc short memmove1");// if dest<source, must go forwards
	asm("std");				// else go backwards
	asm("lea esi,[esi+ecx-4]"); // and start at end of block - 4
	asm("lea edi,[edi+ecx-4]");	//
	asm("memmove1:");
	asm("shr ecx,2");		// ecx now contains number of dwords to move
	asm("rep movsd");		// do dword block move

	asm("memmove0:");
	asm("popfd");
	asm("pop edi");
	asm("pop esi");
	asm("ret");				// finished - return value in EAX
	}



// See header file e32cmn.h for the in-source documentation.
EXPORT_C __NAKED__ TAny* memset(TAny* /*aTrg*/, TInt /*aValue*/, unsigned int /*aLength*/)
	{
	asm("push edi");
	asm("mov edi,[esp+8]"); // destination address into edi
	asm("mov al,[esp+12]");	// fill value into al
	asm("mov ecx,[esp+16]");// number of bytes to fill into ecx
	asm("push edi");		// save destination for return value
	asm("pushfd");
	asm("cld");				// go forwards through array
	asm("test ecx,ecx");	//
	asm("jz short memfill0");// if length zero, nothing to do
	asm("cmp ecx,8");		// if array very small, just do byte fills
	asm("jb short memfill1");

	asm("mov ah,al");		// repeat al in all bytes of eax
	asm("movzx edx,ax");	//
	asm("shl eax,16");		//
	asm("or eax,edx");		//
	asm("mov edx,ecx");		// length into edx
	asm("mov ecx,4");		// number of byte fills to align = 4-[edi mod 4]
	asm("sub ecx,edi");		//
	asm("and ecx,3");		//
	asm("jz short memfill2");// if already aligned, proceed to dword fill
	asm("sub edx,ecx");		// subtract alignment bytes from length
	asm("rep stosb");		// do byte fills to align
	asm("memfill2:");
	asm("mov ecx,edx");		// length remaining into ecx
	asm("shr ecx,2");		// number of dwords to fill into ecx
	asm("rep stosd");		// perform dword fill
	asm("mov ecx,edx");		// calculate number of leftover bytes
	asm("and ecx,3");		// in ecx
	asm("jz short memfill0");// if none left, exit
	asm("memfill1:");
	asm("rep stosb");		// do byte fills to make up correct length

	asm("memfill0:");
	asm("popfd");
	asm("pop eax");			// return value
	asm("pop edi");
	asm("ret");
	}




// See header file e32cmn.h for the in-source documentation.
EXPORT_C __NAKED__ TAny* memclr(TAny* /*aTrg*/, unsigned int /*aLength*/)
	{
	asm("push edi");
	asm("mov edi,[esp+8]");	// destination address into edi
	asm("mov ecx,[esp+12]");// number of bytes to fill into ecx
	asm("xor eax, eax");	// fill value
	asm("push edi");		// save destination for return value
	asm("pushfd");
	asm("cld");				// go forwards through array
	asm("test ecx,ecx");	//
	asm("jz short memclr0");// if length zero, nothing to do
	asm("cmp ecx,8");		// if array very small, just do byte fills
	asm("jb short memclr1");

	asm("mov edx,ecx");		// length into edx
	asm("mov ecx,4");		// number of byte fills to align = 4-[edi mod 4]
	asm("sub ecx,edi");		//
	asm("and ecx,3");		//
	asm("jz short memclr2");// if already aligned, proceed to dword fill
	asm("sub edx,ecx");		// subtract alignment bytes from length
	asm("rep stosb");		// do byte fills to align
	asm("memclr2:");
	asm("mov ecx,edx");		// length remaining into ecx
	asm("shr ecx,2");		// number of dwords to fill into ecx
	asm("rep stosd");		// perform dword fill
	asm("mov ecx,edx");		// calculate number of leftover bytes
	asm("and ecx,3");		// in ecx
	asm("jz short memclr0");// if none left, exit
	asm("memclr1:");
	asm("rep stosb");		// do byte fills to make up correct length

	asm("memclr0:");
	asm("popfd");
	asm("pop eax");			// return value
	asm("pop edi");
	asm("ret");
	}



//EXPORT_C __NAKED__ TInt Mem::Compare(const TUint8* /*aLeft*/, TInt /*aLeftL*/, const TUint8* /*aRight*/, TInt /*aRightL*/)
//	{
//	asm("jmp _memcompare");
//	}

// See header file e32cmn.h for the in-source documentation.
EXPORT_C __NAKED__ TInt memcompare(const TUint8* /*aLeft*/, TInt /*aLeftL*/, const TUint8* /*aRight*/, TInt /*aRightL*/)
	{
	asm("push esi");
	asm("push edi");
	asm("mov esi,[esp+12]");// left buffer address into esi
	asm("mov ecx,[esp+16]");// left buffer length into ecx
	asm("mov edi,[esp+20]");// right buffer address into edi
	asm("mov eax,[esp+24]");// right buffer length into eax
	asm("pushfd");

	asm("mov edx,ecx");		// edx = left len
	asm("cld");				// go forwards through arrays
	asm("sub edx,eax");		// edx = [left len - right len]
	asm("jbe short memcmp1");//
	asm("mov ecx,eax");		// ecx = min[ left len, right len ]
	asm("memcmp1:");
	asm("mov eax,edx");		// save length difference for later use
	asm("test ecx,ecx");	//
	asm("jz short memcmp0");// if length=0, nothing to compare
	asm("cmp ecx,8");		// for small buffers don't bother with alignment check
	asm("jc short memcmp2");//
	asm("mov edx,ecx");		// comparison length into edx
	asm("mov ecx,4");		// number of byte compares to align left ptr = 4-[esi mod 4]
	asm("sub ecx,esi");		//
	asm("and ecx,3");		// number of byte compares to align in ecx
	asm("jz short memcmp3");// if already aligned, proceed with dword compare
	asm("sub edx,ecx");		// subtract number of byte compares from length
	asm("repe cmpsb");		// do byte compares to align
	asm("jne short memcmp4");// if a difference was found, exit
	asm("memcmp3:");
	asm("mov ecx,edx");		// length back into ecx
	asm("shr ecx,2");		// number of dwords to compare
	asm("repe cmpsd");		// perform DWORD block comparison
	asm("je short memcmp5");// if no difference found, proceed with leftover bytes
	asm("mov eax,[esi-4]");	// differing left dword into eax [first char in al]
	asm("mov edx,[edi-4]");	// differing right dword into edx [first char in dl]
	asm("cmp al,dl");		// compare 1st chars
	asm("jne short memcmp7");// if difference found, exit
	asm("cmp ah,dh");		// compare 2nd chars
	asm("jne short memcmp6");// if difference found, exit
	asm("shr eax,16");		// 3rd and 4th chars into al, ah respectively
	asm("shr edx,16");		// 3rd and 4th chars into dl, dh respectively
	asm("cmp al,dl");		// compare 3rd chars
	asm("jne short memcmp7");// if difference found, exit, else 4th chars must differ
	asm("memcmp6:");
	asm("mov al,ah");		// differing bytes into al and dl
	asm("mov dl,dh");		//
	asm("memcmp7:");
	asm("movzx eax,al");	// zero extend al into eax
	asm("movzx edx,dl");	// zero extend dl into edx
	asm("jmp short memcmp8");// and exit with eax = difference of bytes
	asm("memcmp5:");
	asm("mov ecx,edx");		// length back into ecx
	asm("and ecx,3");		// number of remaining bytes to compare
	asm("jz short memcmp0");// if zero, exit with eax=length difference
	asm("memcmp2:");		// *** come here for small buffers
	asm("repe cmpsb");		// compare remaining bytes
	asm("je short memcmp0");// if no difference found, exit with eax=length difference
	asm("memcmp4:");		// *** come here if a byte has been compared and differs
	asm("movzx eax,byte ptr [esi-1]");
	asm("movzx edx,byte ptr [edi-1]");
	asm("memcmp8:");
	asm("sub eax,edx");		// return value is difference of bytes
	
	asm("memcmp0:");		// *** come here if common initial segment identical
	asm("popfd");
	asm("pop edi");
	asm("pop esi");
	asm("ret");
	}
}
#endif