// MatchImpl.cpp: implementation of the CMatchImpl class.
//
//////////////////////////////////////////////////////////////////////

#include "stdafx.h"
#include "MatchImpl.h"
#include "SISMath.h"

#ifdef _DEBUG
#define new DEBUG_NEW
#undef THIS_FILE
static char THIS_FILE[] = __FILE__;
#endif

//////////////////////////////////////////////////////////////////////
// Construction/Destruction
//////////////////////////////////////////////////////////////////////
#include <emmintrin.h>
//#include <tmmintrin.h>
//#include <ia32intrin.h>
#include <math.h>

//#define MOSIS_ASM

extern "C"{
	int64 _asm_GetCC_8u(const uchar * vec1, const uchar * vec2, int len, int flag);// 0= aa, 1=au, 2=uu
};



int64	CSISMath::Get_CCU8_A(const uchar * vec1, const uchar * vec2, int len )
{
#if defined(MOSIS_ASM)
	return _asm_GetCC_8u(vec1, vec2, len, 0);
#else
#if defined(MOSIS_DEBUG_ALIGNED_UNALIGNED)
	return Get_CCU8_UA(vec1, vec2, len);
#endif
	int i= 0, s = 0;
	int64 sum = 0;


#ifdef MOSIS_SIMD
	__m128i	mmResult, mmResult2;
	__m128i	mmA, mmB;
	__m128i	mmAA, mmBB;
	__m128i mmZeroData= _mm_setzero_si128();
	mmResult= _mm_setzero_si128();

#ifdef MOSIS_SIMD_INT64_NO
	int64 res64[2];
	int64 sum64= 0;
	for(i= 0; i <= len- 16; i+= 16)
	{
		mmA= _mm_load_si128((__m128i*) (vec1+ i));
		mmB= _mm_load_si128((__m128i*) (vec2+ i));

		// 1. »óÀ§ µ¥ÀÌÅÍ °è»ê.
		// 1.1 »óÀ§ 8°³ÀÇ 8bit µ¥ÀÌÅ͸¦ 8°³ÀÇ 16bit µ¥ÀÌÅÍ·Î º¯È¯ÈÄ.. °öÇÑ´Ù.
		mmAA= _mm_unpackhi_epi8(mmA, mmZeroData);
		mmBB= _mm_unpackhi_epi8(mmB, mmZeroData);

		mmResult2= _mm_madd_epi16(mmAA, mmBB);
		mmAA= _mm_unpackhi_epi32(mmResult2, mmZeroData);
		mmBB= _mm_unpacklo_epi32(mmResult2, mmZeroData);
		mmResult= _mm_add_epi64(mmAA, mmResult);
		mmResult= _mm_add_epi64(mmBB, mmResult);


		mmAA= _mm_unpacklo_epi8(mmA, mmZeroData);
		mmBB= _mm_unpacklo_epi8(mmB, mmZeroData);

		mmResult2= _mm_madd_epi16(mmAA, mmBB);
		mmResult= _mm_add_epi32(mmResult2, mmResult);
	}

	_mm_storeu_si128((__m128i*)res64, mmResult);
	sum64+= res64[0]+ res64[1];
#endif
	UINT32 results[4];

	for(i= 0; i <= len- 16; i+= 16)
	{

		mmA= _mm_load_si128((__m128i*) (vec1+ i));
		mmB= _mm_load_si128((__m128i*) (vec2+ i));

		// 1. »óÀ§ µ¥ÀÌÅÍ °è»ê.
		// 1.1 »óÀ§ 8°³ÀÇ 8bit µ¥ÀÌÅ͸¦ 8°³ÀÇ 16bit µ¥ÀÌÅÍ·Î º¯È¯ÈÄ.. °öÇÑ´Ù.
		mmAA= _mm_unpackhi_epi8(mmA, mmZeroData);
		mmBB= _mm_unpackhi_epi8(mmB, mmZeroData);

		mmResult2= _mm_madd_epi16(mmAA, mmBB);
		mmResult= _mm_add_epi32(mmResult2, mmResult);


		mmAA= _mm_unpacklo_epi8(mmA, mmZeroData);
		mmBB= _mm_unpacklo_epi8(mmB, mmZeroData);

		mmResult2= _mm_madd_epi16(mmAA, mmBB);
		mmResult= _mm_add_epi32(mmResult2, mmResult);
	}

	_mm_storeu_si128((__m128i*)results, mmResult);
	sum+= results[0]+ results[1]+ results[2]+ results[3];

#else
	for( i = 0; i <= len - 4; i += 4 )
	{
		int e = vec1[i] * vec2[i];
		int v = vec1[i + 1] * vec2[i + 1];

		e += v;
		v = vec1[i + 2] * vec2[i + 2];
		e += v;
		v = vec1[i + 3] * vec2[i + 3];
		e += v;
		sum += e;
	}
#endif

	for( ; i < len; i++ )
	{
		s += vec1[i] * vec2[i];
	}

	return sum + s;
#endif
}