irrlicht/source/Irrlicht/SoftwareDriver2_helper.h

// Copyright (C) 2002-2012 Nikolaus Gebhardt / Thomas Alten
// This file is part of the "Irrlicht Engine".
// For conditions of distribution and use, see copyright notice in irrlicht.h

/*
	History:
	- changed behavior for log2 textures ( replaced multiplies by shift )
*/

#ifndef __S_VIDEO_2_SOFTWARE_HELPER_H_INCLUDED__
#define __S_VIDEO_2_SOFTWARE_HELPER_H_INCLUDED__

#include "SoftwareDriver2_compile_config.h"
#include "irrMath.h"
#include "irrMathFastCompat.h"
#include "CSoftwareTexture2.h"
#include "SMaterial.h"


namespace irr
{

// supporting different packed pixel needs many defines...

#ifdef SOFTWARE_DRIVER_2_32BIT
	typedef u32	tVideoSample;

	#define	MASK_A	0xFF000000
	#define	MASK_R	0x00FF0000
	#define	MASK_G	0x0000FF00
	#define	MASK_B	0x000000FF

	#define	SHIFT_A	24
	#define	SHIFT_R	16
	#define	SHIFT_G	8
	#define	SHIFT_B	0

	#define	COLOR_MAX					0xFF
	#define	COLOR_MAX_LOG2				8
	#define	COLOR_BRIGHT_WHITE			0xFFFFFFFF

	#define VIDEO_SAMPLE_GRANULARITY	2

#else
	typedef u16	tVideoSample;

	#define	MASK_A	0x8000
	#define	MASK_R	0x7C00
	#define	MASK_G	0x03E0
	#define	MASK_B	0x001F

	#define	SHIFT_A	15
	#define	SHIFT_R	10
	#define	SHIFT_G	5
	#define	SHIFT_B	0

	#define	COLOR_MAX					0x1F
	#define	COLOR_MAX_LOG2				5
	#define	COLOR_BRIGHT_WHITE			0xFFFF
	#define VIDEO_SAMPLE_GRANULARITY	1

#endif




// ----------------------- Generic ----------------------------------

//! a more useful memset for pixel
// (standard memset only works with 8-bit values)
inline void memset32(void * dest, const u32 value, u32 bytesize)
{
	u32 * d = (u32*) dest;

	u32 i;

	// loops unrolled to reduce the number of increments by factor ~8.
	i = bytesize >> (2 + 3);
	while (i)
	{
		d[0] = value;
		d[1] = value;
		d[2] = value;
		d[3] = value;

		d[4] = value;
		d[5] = value;
		d[6] = value;
		d[7] = value;

		d += 8;
		i -= 1;
	}

	i = (bytesize >> 2 ) & 7;
	while (i)
	{
		d[0] = value;
		d += 1;
		i -= 1;
	}
}

//! a more useful memset for pixel
// (standard memset only works with 8-bit values)
inline void memset16(void * dest, const u16 value, u32 bytesize)
{
	u16 * d = (u16*) dest;

	u32 i;

	// loops unrolled to reduce the number of increments by factor ~8.
	i = bytesize >> (1 + 3);
	while (i)
	{
		d[0] = value;
		d[1] = value;
		d[2] = value;
		d[3] = value;

		d[4] = value;
		d[5] = value;
		d[6] = value;
		d[7] = value;

		d += 8;
		--i;
	}

	i = (bytesize >> 1 ) & 7;
	while (i)
	{
		d[0] = value;
		++d;
		--i;
	}
}

/*
	use biased loop counter
	--> 0 byte copy is forbidden
*/
REALINLINE void memcpy32_small ( void * dest, const void *source, u32 bytesize )
{
	u32 c = bytesize >> 2;

	do
	{
		((u32*) dest ) [ c-1 ] = ((u32*) source) [ c-1 ];
	} while ( --c );

}



// integer log2 of a float ieee 754. TODO: non ieee floating point
static inline s32 s32_log2_f32( f32 f)
{
	u32 x = IR ( f );
	return ((x & 0x7F800000) >> 23) - 127;
}

static inline s32 s32_log2_s32(u32 x)
{
	return s32_log2_f32( (f32) x);
}

static inline s32 s32_abs(s32 x)
{
	s32 b = x >> 31;
	return (x ^ b ) - b;
}


//! conditional set based on mask and arithmetic shift
REALINLINE u32 if_mask_a_else_b ( const u32 mask, const u32 a, const u32 b )
{
	return ( mask & ( a ^ b ) ) ^ b;
}

// ------------------ Video---------------------------------------
/*!
	Pixel = dest * ( 1 - alpha ) + source * alpha
	alpha [0;256]
*/
REALINLINE u32 PixelBlend32 ( const u32 c2, const u32 c1, u32 alpha )
{
	u32 srcRB = c1 & 0x00FF00FF;
	u32 srcXG = c1 & 0x0000FF00;

	u32 dstRB = c2 & 0x00FF00FF;
	u32 dstXG = c2 & 0x0000FF00;


	u32 rb = srcRB - dstRB;
	u32 xg = srcXG - dstXG;

	rb *= alpha;
	xg *= alpha;
	rb >>= 8;
	xg >>= 8;

	rb += dstRB;
	xg += dstXG;

	rb &= 0x00FF00FF;
	xg &= 0x0000FF00;

	return rb | xg;
}

/*!
	Pixel = dest * ( 1 - alpha ) + source * alpha
	alpha [0;32]
*/
inline u16 PixelBlend16 ( const u16 c2, const u32 c1, const u16 alpha )
{
	const u16 srcRB = c1 & 0x7C1F;
	const u16 srcXG = c1 & 0x03E0;

	const u16 dstRB = c2 & 0x7C1F;
	const u16 dstXG = c2 & 0x03E0;

	u32 rb = srcRB - dstRB;
	u32 xg = srcXG - dstXG;

	rb *= alpha;
	xg *= alpha;
	rb >>= 5;
	xg >>= 5;

	rb += dstRB;
	xg += dstXG;

	rb &= 0x7C1F;
	xg &= 0x03E0;

	return (u16)(rb | xg);
}

/*
	Pixel = c0 * (c1/31). c0 Alpha retain
*/
inline u16 PixelMul16 ( const u16 c0, const u16 c1)
{
	return (u16)((( ( (c0 & 0x7C00) * (c1 & 0x7C00) ) & 0x3E000000 ) >> 15 ) |
			(( ( (c0 & 0x03E0) * (c1 & 0x03E0) ) & 0x000F8000 ) >> 10 ) |
			(( ( (c0 & 0x001F) * (c1 & 0x001F) ) & 0x000003E0 ) >> 5 ) |
			(c0 & 0x8000));
}

/*
	Pixel = c0 * (c1/31).
*/
inline u16 PixelMul16_2 ( u16 c0, u16 c1)
{
	return	(u16)(( ( (c0 & 0x7C00) * (c1 & 0x7C00) ) & 0x3E000000 ) >> 15 |
			( ( (c0 & 0x03E0) * (c1 & 0x03E0) ) & 0x000F8000 ) >> 10 |
			( ( (c0 & 0x001F) * (c1 & 0x001F) ) & 0x000003E0 ) >> 5  |
			( c0 & c1 & 0x8000));
}

/*
	Pixel = c0 * (c1/255). c0 Alpha Retain
*/
REALINLINE u32 PixelMul32 ( const u32 c0, const u32 c1)
{
	return	(c0 & 0xFF000000) |
			(( ( (c0 & 0x00FF0000) >> 12 ) * ( (c1 & 0x00FF0000) >> 12 ) ) & 0x00FF0000 ) |
			(( ( (c0 & 0x0000FF00) * (c1 & 0x0000FF00) ) >> 16 ) & 0x0000FF00 ) |
			(( ( (c0 & 0x000000FF) * (c1 & 0x000000FF) ) >> 8  ) & 0x000000FF);
}

/*
	Pixel = c0 * (c1/255).
*/
REALINLINE u32 PixelMul32_2 ( const u32 c0, const u32 c1)
{
	return	(( ( (c0 & 0xFF000000) >> 16 ) * ( (c1 & 0xFF000000) >> 16 ) ) & 0xFF000000 ) |
			(( ( (c0 & 0x00FF0000) >> 12 ) * ( (c1 & 0x00FF0000) >> 12 ) ) & 0x00FF0000 ) |
			(( ( (c0 & 0x0000FF00) * (c1 & 0x0000FF00) ) >> 16 ) & 0x0000FF00 ) |
			(( ( (c0 & 0x000000FF) * (c1 & 0x000000FF) ) >> 8  ) & 0x000000FF);
}

/*
	Pixel = clamp ( c0 + c1, 0, 255 )
*/
REALINLINE u32 PixelAdd32 ( const u32 c2, const u32 c1)
{
	u32 sum = ( c2 & 0x00FFFFFF )  + ( c1 & 0x00FFFFFF );
	u32 low_bits = ( c2 ^ c1 ) & 0x00010101;
	s32 carries  = ( sum - low_bits ) & 0x01010100;
	u32 modulo = sum - carries;
	u32 clamp = carries - ( carries >> 8 );
	return modulo | clamp;
}

#if 0

// 1 - Bit Alpha Blending
inline u16 PixelBlend16 ( const u16 destination, const u16 source )
{
	if((source & 0x8000) == 0x8000)
		return source; // The source is visible, so use it.
	else
		return destination; // The source is transparent, so use the destination.
}

// 1 - Bit Alpha Blending 16Bit SIMD
inline u32 PixelBlend16_simd ( const u32 destination, const u32 source )
{
	switch(source & 0x80008000)
	{
		case 0x80008000: // Both source pixels are visible
			return source;

		case 0x80000000: // Only the first source pixel is visible
			return (source & 0xFFFF0000) | (destination & 0x0000FFFF);

		case 0x00008000: // Only the second source pixel is visible.
			return (destination & 0xFFFF0000) | (source & 0x0000FFFF);

		default: // Neither source pixel is visible.
			return destination;
	}
}
#else

// 1 - Bit Alpha Blending
inline u16 PixelBlend16 ( const u16 c2, const u16 c1 )
{
	u16 mask = ((c1 & 0x8000) >> 15 ) + 0x7fff;
	return (c2 & mask ) | ( c1 & ~mask );
}

// 1 - Bit Alpha Blending 16Bit SIMD
inline u32 PixelBlend16_simd ( const u32 c2, const u32 c1 )
{
	u32 mask = ((c1 & 0x80008000) >> 15 ) + 0x7fff7fff;
	return (c2 & mask ) | ( c1 & ~mask );
}

#endif

/*!
	Pixel = dest * ( 1 - SourceAlpha ) + source * SourceAlpha (OpenGL blending)
*/
inline u32 PixelBlend32 ( const u32 c2, const u32 c1 )
{
	// alpha test
	u32 alpha = c1 & 0xFF000000;

	if ( 0 == alpha )
		return c2;
	if ( 0xFF000000 == alpha )
	{
		return c1;
	}

	alpha >>= 24;

	// add highbit alpha, if ( alpha > 127 ) alpha += 1;
	alpha += ( alpha >> 7);

	u32 srcRB = c1 & 0x00FF00FF;
	u32 srcXG = c1 & 0x0000FF00;

	u32 dstRB = c2 & 0x00FF00FF;
	u32 dstXG = c2 & 0x0000FF00;


	u32 rb = srcRB - dstRB;
	u32 xg = srcXG - dstXG;

	rb *= alpha;
	xg *= alpha;
	rb >>= 8;
	xg >>= 8;

	rb += dstRB;
	xg += dstXG;

	rb &= 0x00FF00FF;
	xg &= 0x0000FF00;

	return (c1 & 0xFF000000) | rb | xg;
}

/*!
	Pixel =>
			color = sourceAlpha > 0 ? source, else dest
			alpha = max(destAlpha, sourceAlpha)
*/
inline u16 PixelCombine16 ( const u16 c2, const u16 c1 )
{
	if ( video::getAlpha(c1) > 0 )
		return c1;
	else
		return c2;
}

/*!
	Pixel =>
			color = dest * ( 1 - SourceAlpha ) + source * SourceAlpha,
			alpha = destAlpha * ( 1 - SourceAlpha ) + sourceAlpha

	where "1" means "full scale" (255)
*/
inline u32 PixelCombine32 ( const u32 c2, const u32 c1 )
{
	// alpha test
	u32 alpha = c1 & 0xFF000000;

	if ( 0 == alpha )
		return c2;
	if ( 0xFF000000 == alpha )
	{
		return c1;
	}

	alpha >>= 24;

	// add highbit alpha, if ( alpha > 127 ) alpha += 1;
	// stretches [0;255] to [0;256] to avoid division by 255. use division 256 == shr 8
	alpha += ( alpha >> 7);

	u32 srcRB = c1 & 0x00FF00FF;
	u32 srcXG = c1 & 0x0000FF00;

	u32 dstRB = c2 & 0x00FF00FF;
	u32 dstXG = c2 & 0x0000FF00;


	u32 rb = srcRB - dstRB;
	u32 xg = srcXG - dstXG;

	rb *= alpha;
	xg *= alpha;
	rb >>= 8;
	xg >>= 8;

	rb += dstRB;
	xg += dstXG;

	rb &= 0x00FF00FF;
	xg &= 0x0000FF00;

	u32 sa = c1 >> 24;
	u32 da = c2 >> 24;
	u32 blendAlpha_fix8 = (sa*256 + da*(256-alpha))>>8;
	return blendAlpha_fix8 << 24 | rb | xg;
}



// ------------------ Fix Point ----------------------------------

typedef s32 tFixPoint;
typedef u32 tFixPointu;

// Fix Point 12
#if 0
	#define FIX_POINT_PRE			12
	#define FIX_POINT_FRACT_MASK	0xFFF
	#define FIX_POINT_SIGNED_MASK	0xFFFFF000
	#define FIX_POINT_UNSIGNED_MASK	0x7FFFF000
	#define FIX_POINT_ONE			0x1000
	#define FIX_POINT_ZERO_DOT_FIVE	0x0800
	#define FIX_POINT_F32_MUL		4096.f
#endif

// Fix Point 10
#if 1
	#define FIX_POINT_PRE			10
	#define FIX_POINT_FRACT_MASK	0x3FF
	#define FIX_POINT_SIGNED_MASK	0xFFFFFC00
	#define FIX_POINT_UNSIGNED_MASK	0x7FFFFE00
	#define FIX_POINT_ONE			0x400
	#define FIX_POINT_ZERO_DOT_FIVE	0x200
	#define FIX_POINT_F32_MUL		1024.f
#endif

// Fix Point 9
#if 0
	#define FIX_POINT_PRE			9
	#define FIX_POINT_FRACT_MASK	0x1FF
	#define FIX_POINT_SIGNED_MASK	0xFFFFFE00
	#define FIX_POINT_UNSIGNED_MASK	0x7FFFFE00
	#define FIX_POINT_ONE			0x200
	#define FIX_POINT_ZERO_DOT_FIVE	0x100
	#define FIX_POINT_F32_MUL		512.f
#endif

// Fix Point 7
#if 0
	#define FIX_POINT_PRE			7
	#define FIX_POINT_FRACT_MASK	0x7F
	#define FIX_POINT_SIGNED_MASK	0xFFFFFF80
	#define FIX_POINT_UNSIGNED_MASK	0x7FFFFF80
	#define FIX_POINT_ONE			0x80
	#define FIX_POINT_ZERO_DOT_FIVE	0x40
	#define FIX_POINT_F32_MUL		128.f
#endif

#define	FIXPOINT_COLOR_MAX		( COLOR_MAX << FIX_POINT_PRE )
#define FIX_POINT_HALF_COLOR ( (tFixPoint) ( ((f32) COLOR_MAX / 2.f * FIX_POINT_F32_MUL ) ) )


/*
	convert signed integer to fixpoint
*/
inline tFixPoint s32_to_fixPoint (const s32 x)
{
	return x << FIX_POINT_PRE;
}

inline tFixPointu u32_to_fixPoint (const u32 x)
{
	return x << FIX_POINT_PRE;
}

inline u32 fixPointu_to_u32 (const tFixPointu x)
{
	return x >> FIX_POINT_PRE;
}


// 1/x * FIX_POINT
#define fix_inverse32(x) (FIX_POINT_F32_MUL / (x))


/*
	convert float to fixpoint
	fast convert (fistp on x86) HAS to be used..
	hints: compileflag /QIfist for msvc7. msvc 8.0 has smth different
	others should use their favourite assembler..
*/
static inline int f_round2(f32 f)
{
	f += (3<<22);
	return IR(f) - 0x4b400000;
}

/*
	convert f32 to Fix Point.
	multiply is needed anyway, so scale mulby
*/
REALINLINE tFixPoint tofix0 (const f32 x, const f32 mulby = FIX_POINT_F32_MUL )
{
	return (tFixPoint) (x * mulby);
}
#define tofix(x,y) (tFixPoint)(x * y)

/*
	Fix Point , Fix Point Multiply
*/
REALINLINE tFixPointu imulFixu(const tFixPointu x, const tFixPointu y)
{
	return (x * y) >> (tFixPointu) FIX_POINT_PRE;
}

/*
	Fix Point , Fix Point Multiply
*/
REALINLINE tFixPoint imulFix(const tFixPoint x, const tFixPoint y)
{
	return ( x * y) >> ( FIX_POINT_PRE );
}

/*
	Fix Point , Fix Point Multiply x * y * 2
*/
REALINLINE tFixPoint imulFix2(const tFixPoint x, const tFixPoint y)
{
	return ( x * y) >> ( FIX_POINT_PRE -1 );
}


/*
	Multiply x * y * 1
*/
REALINLINE tFixPoint imulFix_tex1(const tFixPoint x, const tFixPoint y)
{
	return ( ( (tFixPointu) x >> 2 ) * ( (tFixPointu) y >> 2 ) ) >> (tFixPointu) ( FIX_POINT_PRE + 4 );
}

/*
	Multiply x * y * 2
*/
REALINLINE tFixPoint imulFix_tex2(const tFixPoint x, const tFixPoint y)
{
	return ( ( (tFixPointu) x >> 2 ) * ( (tFixPointu) y >> 2 ) ) >> (tFixPointu) ( FIX_POINT_PRE + 3 );
}

/*
	Multiply x * y * 4
*/
REALINLINE tFixPoint imulFix_tex4(const tFixPoint x, const tFixPoint y)
{
#ifdef SOFTWARE_DRIVER_2_32BIT
	return ( ( (tFixPointu) x >> 2 ) * ( (tFixPointu) y >> 2 ) ) >> (tFixPointu) ( FIX_POINT_PRE + 2 );
#else
	return ( x * y) >> ( FIX_POINT_PRE + ( VIDEO_SAMPLE_GRANULARITY * 3 ) );
#endif
}

/*!
	clamp FixPoint to maxcolor in FixPoint, min(a,31)
*/
REALINLINE tFixPoint clampfix_maxcolor ( const tFixPoint a)
{
	tFixPoint c = (a - FIXPOINT_COLOR_MAX) >> 31;
	return (a & c) | ( FIXPOINT_COLOR_MAX & ~c);
}

/*!
	clamp FixPoint to 0 in FixPoint, max(a,0)
*/
REALINLINE tFixPoint clampfix_mincolor ( const tFixPoint a)
{
	return a - ( a & ( a >> 31 ) );
}

REALINLINE tFixPoint saturateFix ( const tFixPoint a)
{
	return clampfix_mincolor ( clampfix_maxcolor ( a ) );
}


// rount fixpoint to int
inline s32 roundFix ( const tFixPoint x )
{
	return ( x + FIX_POINT_ZERO_DOT_FIVE ) >> FIX_POINT_PRE;
}



// x in [0;1[
inline s32 f32_to_23Bits(const f32 x)
{
	f32 y = x + 1.f;
	return IR(y) & 0x7FFFFF;	// last 23 bits
}

/*!
	return VideoSample from fixpoint
*/
REALINLINE tVideoSample fix_to_color ( const tFixPoint r, const tFixPoint g, const tFixPoint b )
{
	return	( FIXPOINT_COLOR_MAX & FIXPOINT_COLOR_MAX) << ( SHIFT_A - FIX_POINT_PRE ) |
			( r & FIXPOINT_COLOR_MAX) << ( SHIFT_R - FIX_POINT_PRE ) |
			( g & FIXPOINT_COLOR_MAX) >> ( FIX_POINT_PRE - SHIFT_G ) |
			( b & FIXPOINT_COLOR_MAX) >> ( FIX_POINT_PRE - SHIFT_B );
}


/*!
	return VideoSample from fixpoint
*/
REALINLINE tVideoSample fix4_to_color ( const tFixPoint a, const tFixPoint r, const tFixPoint g, const tFixPoint b )
{
	return	( a & (FIX_POINT_FRACT_MASK - 1 )) << ( SHIFT_A - 1 ) |
			( r & FIXPOINT_COLOR_MAX) << ( SHIFT_R - FIX_POINT_PRE ) |
			( g & FIXPOINT_COLOR_MAX) >> ( FIX_POINT_PRE - SHIFT_G ) |
			( b & FIXPOINT_COLOR_MAX) >> ( FIX_POINT_PRE - SHIFT_B );
}

/*!
	return fixpoint from VideoSample granularity COLOR_MAX
*/
inline void color_to_fix ( tFixPoint &r, tFixPoint &g, tFixPoint &b, const tVideoSample t00 )
{
	(tFixPointu&) r =	(t00 & MASK_R) >> ( SHIFT_R - FIX_POINT_PRE );
	(tFixPointu&) g =	(t00 & MASK_G) << ( FIX_POINT_PRE - SHIFT_G );
	(tFixPointu&) b =	(t00 & MASK_B) << ( FIX_POINT_PRE - SHIFT_B );
}

/*!
	return fixpoint from VideoSample granularity COLOR_MAX
*/
inline void color_to_fix ( tFixPoint &a, tFixPoint &r, tFixPoint &g, tFixPoint &b, const tVideoSample t00 )
{
	(tFixPointu&) a =	(t00 & MASK_A) >> ( SHIFT_A - FIX_POINT_PRE );
	(tFixPointu&) r =	(t00 & MASK_R) >> ( SHIFT_R - FIX_POINT_PRE );
	(tFixPointu&) g =	(t00 & MASK_G) << ( FIX_POINT_PRE - SHIFT_G );
	(tFixPointu&) b =	(t00 & MASK_B) << ( FIX_POINT_PRE - SHIFT_B );
}

/*!
	return fixpoint from VideoSample granularity 0..FIX_POINT_ONE
*/
inline void color_to_fix1 ( tFixPoint &r, tFixPoint &g, tFixPoint &b, const tVideoSample t00 )
{
	(tFixPointu&) r =	(t00 & MASK_R) >> ( SHIFT_R + COLOR_MAX_LOG2 - FIX_POINT_PRE );
	(tFixPointu&) g =	(t00 & MASK_G) >> ( SHIFT_G + COLOR_MAX_LOG2 - FIX_POINT_PRE );
	(tFixPointu&) b =	(t00 & MASK_B) << ( FIX_POINT_PRE - COLOR_MAX_LOG2 );
}

/*!
	return fixpoint from VideoSample granularity 0..FIX_POINT_ONE
*/
inline void color_to_fix1 ( tFixPoint &a, tFixPoint &r, tFixPoint &g, tFixPoint &b, const tVideoSample t00 )
{
	(tFixPointu&) a =	(t00 & MASK_A) >> ( SHIFT_A + COLOR_MAX_LOG2 - FIX_POINT_PRE );
	(tFixPointu&) r =	(t00 & MASK_R) >> ( SHIFT_R + COLOR_MAX_LOG2 - FIX_POINT_PRE );
	(tFixPointu&) g =	(t00 & MASK_G) >> ( SHIFT_G + COLOR_MAX_LOG2 - FIX_POINT_PRE );
	(tFixPointu&) b =	(t00 & MASK_B) << ( FIX_POINT_PRE - COLOR_MAX_LOG2 );
}



// ----- FP24 ---- floating point z-buffer

#if 1
typedef f32 fp24;
#else
struct fp24
{
	u32 v;

	fp24() {}

	fp24 ( const f32 f )
	{
		f32 y = f + 1.f;
		v = ((u32&)y) & 0x7FFFFF;	// last 23 bits
	}

	void operator=(const f32 f )
	{
	f32 y = f + 1.f;
		v = ((u32&)y) & 0x7FFFFF;	// last 23 bits
		}

	void operator+=(const fp24 &other )
	{
		v += other.v;
	}

	operator f32 () const
	{
		f32 r = FR ( v );
		return r + 1.f;
	}

};
#endif


// ------------------------ Internal Texture -----------------------------

struct sInternalTexture
{
	u32 textureXMask;
	u32 textureYMask;

	u32 pitchlog2;
	void *data;

	video::CSoftwareTexture2 *Texture;
	s32 lodLevel;
};



// get video sample plain
inline tVideoSample getTexel_plain ( const sInternalTexture * t, const tFixPointu tx, const tFixPointu ty )
{
	u32 ofs;

	ofs = ( ( ty & t->textureYMask ) >> FIX_POINT_PRE ) << t->pitchlog2;
	ofs |= ( tx & t->textureXMask ) >> ( FIX_POINT_PRE - VIDEO_SAMPLE_GRANULARITY );

	// texel
	return *((tVideoSample*)( (u8*) t->data + ofs ));
}

// get video sample to fix
inline void getTexel_fix ( tFixPoint &r, tFixPoint &g, tFixPoint &b,
						const sInternalTexture * t, const tFixPointu tx, const tFixPointu ty
								)
{
	u32 ofs;

	ofs = ( ( ty & t->textureYMask ) >> FIX_POINT_PRE ) << t->pitchlog2;
	ofs |= ( tx & t->textureXMask ) >> ( FIX_POINT_PRE - VIDEO_SAMPLE_GRANULARITY );

	// texel
	tVideoSample t00;
	t00 = *((tVideoSample*)( (u8*) t->data + ofs ));

	r = (t00 & MASK_R) >> ( SHIFT_R - FIX_POINT_PRE);
	g = (t00 & MASK_G) << ( FIX_POINT_PRE - SHIFT_G );
	b = (t00 & MASK_B) << ( FIX_POINT_PRE - SHIFT_B );

}

// get video sample to fixpoint
REALINLINE void getTexel_fix ( tFixPoint &a,
			const sInternalTexture * t, const tFixPointu tx, const tFixPointu ty)
{
	u32 ofs;

	ofs = ( ( ty & t->textureYMask ) >> FIX_POINT_PRE ) << t->pitchlog2;
	ofs |= ( tx & t->textureXMask ) >> ( FIX_POINT_PRE - VIDEO_SAMPLE_GRANULARITY );

	// texel
	tVideoSample t00;
	t00 = *((tVideoSample*)( (u8*) t->data + ofs ));

	a = (t00 & MASK_A) >> ( SHIFT_A - FIX_POINT_PRE);
}


inline void getSample_texture_dither (	tFixPoint &r, tFixPoint &g, tFixPoint &b,
										const sInternalTexture * t, const tFixPointu tx, const tFixPointu ty,
										const u32 x, const u32 y
								)
{
	static const tFixPointu dithermask[] =
	{
		0x00,0x80,0x20,0xa0,
		0xc0,0x40,0xe0,0x60,
		0x30,0xb0,0x10,0x90,
		0xf0,0x70,0xd0,0x50
	};

	const u32 index = (y & 3 ) << 2 | (x & 3);

	const tFixPointu _ntx = (tx + dithermask [ index ] ) & t->textureXMask;
	const tFixPointu _nty = (ty + dithermask [ index ] ) & t->textureYMask;

	u32 ofs;
	ofs = ( ( _nty ) >> FIX_POINT_PRE ) << t->pitchlog2;
	ofs |= ( _ntx ) >> ( FIX_POINT_PRE - VIDEO_SAMPLE_GRANULARITY );

	// texel
	const tVideoSample t00 = *((tVideoSample*)( (u8*) t->data + ofs ));

	(tFixPointu &) r =	(t00 & MASK_R) >> ( SHIFT_R - FIX_POINT_PRE);
	(tFixPointu &) g =	(t00 & MASK_G) << ( FIX_POINT_PRE - SHIFT_G );
	(tFixPointu &) b =	(t00 & MASK_B) << ( FIX_POINT_PRE - SHIFT_B );

}

/*
	load a sample from internal texture at position tx,ty to fixpoint
*/
#ifndef SOFTWARE_DRIVER_2_BILINEAR

// get Sample linear == getSample_fixpoint

inline void getSample_texture ( tFixPoint &r, tFixPoint &g, tFixPoint &b,
						const sInternalTexture * t, const tFixPointu tx, const tFixPointu ty
								)
{
	u32 ofs;

	ofs = ( ( ty & t->textureYMask ) >> FIX_POINT_PRE ) << t->pitchlog2;
	ofs |= ( tx & t->textureXMask ) >> ( FIX_POINT_PRE - VIDEO_SAMPLE_GRANULARITY );

	// texel
	const tVideoSample t00 = *((tVideoSample*)( (u8*) t->data + ofs ));

	(tFixPointu &) r =	(t00 & MASK_R) >> ( SHIFT_R - FIX_POINT_PRE);
	(tFixPointu &) g =	(t00 & MASK_G) << ( FIX_POINT_PRE - SHIFT_G );
	(tFixPointu &) b =	(t00 & MASK_B) << ( FIX_POINT_PRE - SHIFT_B );
}

inline void getSample_texture ( tFixPoint &a, tFixPoint &r, tFixPoint &g, tFixPoint &b,
						const sInternalTexture * t, const tFixPointu tx, const tFixPointu ty
								)
{
	u32 ofs;

	ofs = ( ( ty & t->textureYMask ) >> FIX_POINT_PRE ) << t->pitchlog2;
	ofs |= ( tx & t->textureXMask ) >> ( FIX_POINT_PRE - VIDEO_SAMPLE_GRANULARITY );

	// texel
	const tVideoSample t00 = *((tVideoSample*)( (u8*) t->data + ofs ));

	(tFixPointu &)a =	(t00 & MASK_A) >> ( SHIFT_A - FIX_POINT_PRE);
	(tFixPointu &)r =	(t00 & MASK_R) >> ( SHIFT_R - FIX_POINT_PRE);
	(tFixPointu &)g =	(t00 & MASK_G) << ( FIX_POINT_PRE - SHIFT_G );
	(tFixPointu &)b =	(t00 & MASK_B) << ( FIX_POINT_PRE - SHIFT_B );
}


#else


// get sample linear
REALINLINE void getSample_linear ( tFixPointu &r, tFixPointu &g, tFixPointu &b,
								const sInternalTexture * t, const tFixPointu tx, const tFixPointu ty
								)
{
	u32 ofs;

	ofs = ( ( ty & t->textureYMask ) >> FIX_POINT_PRE ) << t->pitchlog2;
	ofs |= ( tx & t->textureXMask ) >> ( FIX_POINT_PRE - VIDEO_SAMPLE_GRANULARITY );

	// texel
	tVideoSample t00;
	t00 = *((tVideoSample*)( (u8*) t->data + ofs ));

	r =	(t00 & MASK_R) >> SHIFT_R;
	g =	(t00 & MASK_G) >> SHIFT_G;
	b =	(t00 & MASK_B);
}

// get Sample bilinear
REALINLINE void getSample_texture ( tFixPoint &r, tFixPoint &g, tFixPoint &b,
								const sInternalTexture * t, const tFixPointu tx, const tFixPointu ty
								)
{

	tFixPointu r00,g00,b00;
	tFixPointu r01,g01,b01;
	tFixPointu r10,g10,b10;
	tFixPointu r11,g11,b11;

#if 0
	getSample_linear ( r00, g00, b00, t, tx,ty );
	getSample_linear ( r10, g10, b10, t, tx + FIX_POINT_ONE,ty );
	getSample_linear ( r01, g01, b01, t, tx,ty + FIX_POINT_ONE );
	getSample_linear ( r11, g11, b11, t, tx + FIX_POINT_ONE,ty + FIX_POINT_ONE );
#else
	u32 o0, o1,o2,o3;
	tVideoSample t00;

	o0 = ( ( (ty) & t->textureYMask ) >> FIX_POINT_PRE ) << t->pitchlog2;
	o1 = ( ( (ty+FIX_POINT_ONE) & t->textureYMask ) >> FIX_POINT_PRE ) << t->pitchlog2;
	o2 =   ( (tx) & t->textureXMask ) >> ( FIX_POINT_PRE - VIDEO_SAMPLE_GRANULARITY );
	o3 =   ( (tx+FIX_POINT_ONE) & t->textureXMask ) >> ( FIX_POINT_PRE - VIDEO_SAMPLE_GRANULARITY );

	t00 = *((tVideoSample*)( (u8*) t->data + (o0 | o2 ) ));
	r00 =	(t00 & MASK_R) >> SHIFT_R;
	g00 =	(t00 & MASK_G) >> SHIFT_G;
	b00 =	(t00 & MASK_B);

	t00 = *((tVideoSample*)( (u8*) t->data + (o0 | o3 ) ));
	r10 =	(t00 & MASK_R) >> SHIFT_R;
	g10 =	(t00 & MASK_G) >> SHIFT_G;
	b10 =	(t00 & MASK_B);

	t00 = *((tVideoSample*)( (u8*) t->data + (o1 | o2 ) ));
	r01 =	(t00 & MASK_R) >> SHIFT_R;
	g01 =	(t00 & MASK_G) >> SHIFT_G;
	b01 =	(t00 & MASK_B);

	t00 = *((tVideoSample*)( (u8*) t->data + (o1 | o3 ) ));
	r11 =	(t00 & MASK_R) >> SHIFT_R;
	g11 =	(t00 & MASK_G) >> SHIFT_G;
	b11 =	(t00 & MASK_B);

#endif

	const tFixPointu txFract = tx & FIX_POINT_FRACT_MASK;
	const tFixPointu txFractInv = FIX_POINT_ONE - txFract;

	const tFixPointu tyFract = ty & FIX_POINT_FRACT_MASK;
	const tFixPointu tyFractInv = FIX_POINT_ONE - tyFract;

	const tFixPointu w00 = imulFixu ( txFractInv, tyFractInv );
	const tFixPointu w10 = imulFixu ( txFract	, tyFractInv );
	const tFixPointu w01 = imulFixu ( txFractInv, tyFract );
	const tFixPointu w11 = imulFixu ( txFract	, tyFract );

	r =		(r00 * w00 ) +
			(r01 * w01 ) +
			(r10 * w10 ) +
			(r11 * w11 );

	g =		(g00 * w00 ) +
			(g01 * w01 ) +
			(g10 * w10 ) +
			(g11 * w11 );

	b =		(b00 * w00 ) +
			(b01 * w01 ) +
			(b10 * w10 ) +
			(b11 * w11 );

}


// get sample linear
REALINLINE void getSample_linear ( tFixPointu &a, tFixPointu &r, tFixPointu &g, tFixPointu &b,
								const sInternalTexture * t, const tFixPointu tx, const tFixPointu ty
								)
{
	u32 ofs;

	ofs = ( ( ty & t->textureYMask ) >> FIX_POINT_PRE ) << t->pitchlog2;
	ofs |= ( tx & t->textureXMask ) >> ( FIX_POINT_PRE - VIDEO_SAMPLE_GRANULARITY );

	// texel
	tVideoSample t00;
	t00 = *((tVideoSample*)( (u8*) t->data + ofs ));

	a =	(t00 & MASK_A) >> SHIFT_A;
	r =	(t00 & MASK_R) >> SHIFT_R;
	g =	(t00 & MASK_G) >> SHIFT_G;
	b =	(t00 & MASK_B);
}

// get Sample bilinear
REALINLINE void getSample_texture ( tFixPoint &a, tFixPoint &r, tFixPoint &g, tFixPoint &b,
								const sInternalTexture * t, const tFixPointu tx, const tFixPointu ty
								)
{

	tFixPointu a00, r00,g00,b00;
	tFixPointu a01, r01,g01,b01;
	tFixPointu a10, r10,g10,b10;
	tFixPointu a11, r11,g11,b11;

	getSample_linear ( a00, r00, g00, b00, t, tx,ty );
	getSample_linear ( a10, r10, g10, b10, t, tx + FIX_POINT_ONE,ty );
	getSample_linear ( a01, r01, g01, b01, t, tx,ty + FIX_POINT_ONE );
	getSample_linear ( a11, r11, g11, b11, t, tx + FIX_POINT_ONE,ty + FIX_POINT_ONE );

	const tFixPointu txFract = tx & FIX_POINT_FRACT_MASK;
	const tFixPointu txFractInv = FIX_POINT_ONE - txFract;

	const tFixPointu tyFract = ty & FIX_POINT_FRACT_MASK;
	const tFixPointu tyFractInv = FIX_POINT_ONE - tyFract;

	const tFixPointu w00 = imulFixu ( txFractInv, tyFractInv );
	const tFixPointu w10 = imulFixu ( txFract	, tyFractInv );
	const tFixPointu w01 = imulFixu ( txFractInv, tyFract );
	const tFixPointu w11 = imulFixu ( txFract	, tyFract );

	a =		(a00 * w00 ) +
			(a01 * w01 ) +
			(a10 * w10 ) +
			(a11 * w11 );

	r =		(r00 * w00 ) +
			(r01 * w01 ) +
			(r10 * w10 ) +
			(r11 * w11 );

	g =		(g00 * w00 ) +
			(g01 * w01 ) +
			(g10 * w10 ) +
			(g11 * w11 );

	b =		(b00 * w00 ) +
			(b01 * w01 ) +
			(b10 * w10 ) +
			(b11 * w11 );

}


#endif

// some 2D Defines
struct AbsRectangle
{
	s32 x0;
	s32 y0;
	s32 x1;
	s32 y1;
};

//! 2D Intersection test
inline bool intersect ( AbsRectangle &dest, const AbsRectangle& a, const AbsRectangle& b)
{
	dest.x0 = core::s32_max( a.x0, b.x0 );
	dest.y0 = core::s32_max( a.y0, b.y0 );
	dest.x1 = core::s32_min( a.x1, b.x1 );
	dest.y1 = core::s32_min( a.y1, b.y1 );
	return dest.x0 < dest.x1 && dest.y0 < dest.y1;
}

// some 1D defines
struct sIntervall
{
	s32 start;
	s32 end;
};

// returning intersection width
inline s32 intervall_intersect_test( const sIntervall& a, const sIntervall& b)
{
	return core::s32_min( a.end, b.end ) - core::s32_max( a.start, b.start );
}


} // end namespace irr

#endif