/* Copyright 2005 Nicholas Bishop
 * 
 * This file is part of SharpConstruct.
 *
 * SharpConstruct is free software; you can redistribute it and/or modify
 * it under the terms of the GNU General Public License as published by
 * the Free Software Foundation; either version 2 of the License, or
 * (at your option) any later version.
 *
 * SharpConstruct is distributed in the hope that it will be useful,
 * but WITHOUT ANY WARRANTY; without even the implied warranty of
 * MERCHANTABILITY or FITNESS FOR A PARTICULAR PURPOSE.  See the
 * GNU General Public License for more details.
 *
 * You should have received a copy of the GNU General Public License
 * along with SharpConstruct; if not, write to the Free Software
 * Foundation, Inc., 59 Temple Place, Suite 330, Boston, MA  02111-1307  USA */

#include "Optimized.h"
#include <algorithm>
#include <functional>

#include <iostream>

using namespace SharpConstruct::Optimized;

void Point3D::FastDistance( const __m128 in, float& d )
{
	register __m128 t( data_ );
	const register __m128 v( _mm_set_ps1( -0.0 ) );
	t = _mm_sub_ps( t, in );
	t = _mm_andnot_ps( v, t ); // Get absolute value
	t = _mm_add_ps( t, _mm_shuffle_ps( t, t, _MM_SHUFFLE( 2, 1, 0, 3 ) ) ); // Add components
	t = _mm_add_ps( t, _mm_shuffle_ps( t, t, _MM_SHUFFLE( 2, 2, 2, 2 ) ) ); // Final add, answer is in [0]
	_mm_store_ss( &d, t );
}

void Point3D::Midpoint( const Point3D& a, const Point3D& b )
{
	*this = ( a + b ) / 2;
}

void Point3D::CalculatePlaneNormal( const __m128 p1, const __m128 p2,
				    const __m128 p3 )
{
	register __m128 a, b, tmp1, tmp2, shuff1, shuff2;
	
	a = _mm_sub_ps( p1, p2 );
	b = _mm_sub_ps( p3, p1 );
	// Shuffles a into Y,Z,X format
	shuff1 = _mm_shuffle_ps( a, a, _MM_SHUFFLE( 3, 0, 2, 1 ) );
	// Shuffles b into Z,X,Y format
	shuff2 = _mm_shuffle_ps( b, b, _MM_SHUFFLE( 1, 1, 0, 2 ) );
	tmp1 = _mm_mul_ps( shuff1, shuff2 );
	shuff1 = _mm_shuffle_ps( a, a, _MM_SHUFFLE( 1, 1, 0, 2 ) );
	shuff2 = _mm_shuffle_ps( b, b, _MM_SHUFFLE( 3, 0, 2, 1 ) );
	tmp2 = _mm_mul_ps( shuff1, shuff2 );
	a = _mm_sub_ps( tmp1, tmp2 );
	
	// And normalize 'a':
	b = _mm_mul_ps( a, a );

	b = _mm_add_ps( b, _mm_shuffle_ps( b, b, _MM_SHUFFLE( 2, 1, 0, 3 ) ) );
	b = _mm_add_ps( b, _mm_shuffle_ps( b, b, _MM_SHUFFLE( 2, 2, 2, 2 ) ) );

	b = _mm_rsqrt_ss( b );
	b = _mm_shuffle_ps( b, b, _MM_SHUFFLE( 0, 0, 0, 0 ) );
	data_ = _mm_mul_ps( a, b );
}
void Point3D::CalculatePlaneUnormal( const __m128 p1, const __m128 p2,
				     const __m128 p3 )
{
	register __m128 a, b, tmp1, tmp2, shuff1, shuff2;
	
	a = _mm_sub_ps( p1, p2 );
	b = _mm_sub_ps( p3, p1 );
	// Shuffles a into Y,Z,X format
	shuff1 = _mm_shuffle_ps( a, a, _MM_SHUFFLE( 3, 0, 2, 1 ) );
	// Shuffles b into Z,X,Y format
	shuff2 = _mm_shuffle_ps( b, b, _MM_SHUFFLE( 1, 1, 0, 2 ) );
	tmp1 = _mm_mul_ps( shuff1, shuff2 );
	shuff1 = _mm_shuffle_ps( a, a, _MM_SHUFFLE( 1, 1, 0, 2 ) );
	shuff2 = _mm_shuffle_ps( b, b, _MM_SHUFFLE( 3, 0, 2, 1 ) );
	tmp2 = _mm_mul_ps( shuff1, shuff2 );
	data_ = _mm_sub_ps( tmp1, tmp2 );
}

void SharpConstruct::Optimized::Normalize( Point3D& p )
{
	register __m128 tmp( p );
	register __m128 length;
	register __m128 shuff;
	// Square all the components
	tmp = _mm_mul_ps( tmp, tmp );
	// Next six lines are to put X + Y + Z into length
	shuff = _mm_shuffle_ps( tmp, tmp, _MM_SHUFFLE( 1, 2, 3, 0 ) );
	length = shuff;
	shuff = _mm_shuffle_ps( tmp, tmp, _MM_SHUFFLE( 2, 3, 0, 1 ) );
	length = _mm_add_ss( length, shuff );
	shuff = _mm_shuffle_ps( tmp, tmp, _MM_SHUFFLE( 3, 0, 1, 2 ) );
	length = _mm_add_ss( length, shuff );
	// Take the reciprocal square root
	length = _mm_rsqrt_ss( length );
	// Move the rsqrt into all four positions
	length = _mm_shuffle_ps( length, length, _MM_SHUFFLE( 0, 0, 0, 0 ) );
	// Multiply it by data
	p = _mm_mul_ps( *p.RawData(), length );
}

void SharpConstruct::Optimized::Normalize( Point3DVector& v )
{
	const unsigned size( v.size() );
	for( unsigned i = 0; i < size; ++i )
		Normalize( v[ i ] );
}
/*void Point3DVector::NormalizeAll()
{
	for( unsigned i = 0; i < size(); i++ )
	{
		register __m128 sq( data_[ i ] );
		
		// Square each component
		sq = _mm_mul_ps( sq, sq );

		// Horizontal add
		sq = _mm_add_ps( sq, _mm_shuffle_ps( sq, sq, _MM_SHUFFLE( 2, 1, 0, 3 ) ) );
		sq = _mm_add_ps( sq, _mm_shuffle_ps( sq, sq, _MM_SHUFFLE( 2, 2, 2, 2 ) ) );
		
		sq = _mm_rsqrt_ss( sq );
		sq = _mm_shuffle_ps( sq, sq, _MM_SHUFFLE( 0, 0, 0, 0 ) );
		
		data_[ i ] = _mm_mul_ps( data_[ i ], sq );
	}
}*/