snes_ntsc_impl.h

linux下的任天堂模拟器代码。供大家参考。

/* Common implementation of NTSC filters */

#include <assert.h>
#include <math.h>

/* Copyright (C) 2006 Shay Green. This module is free software; you
can redistribute it and/or modify it under the terms of the GNU Lesser
General Public License as published by the Free Software Foundation; either
version 2.1 of the License, or (at your option) any later version. This
module 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 Lesser General Public License for more
details. You should have received a copy of the GNU Lesser General Public
License along with this module; if not, write to the Free Software Foundation,
Inc., 51 Franklin Street, Fifth Floor, Boston, MA 02110-1301 USA */

#define DISABLE_CORRECTION 0

#undef PI
#define PI 3.14159265358979323846f

#ifndef LUMA_CUTOFF
	#define LUMA_CUTOFF 0.20
#endif
#ifndef gamma_size
	#define gamma_size 1
#endif
#ifndef rgb_bits
	#define rgb_bits 8
#endif
#ifndef artifacts_max
	#define artifacts_max (artifacts_mid * 1.5f)
#endif
#ifndef fringing_max
	#define fringing_max (fringing_mid * 2)
#endif

#define ext_decoder_hue     (std_decoder_hue + 15)
#define rgb_unit            (1 << rgb_bits)
#define rgb_offset          (rgb_unit * 2 + 0.5f)

enum { burst_size  = snes_ntsc_entry_size / burst_count };
enum { kernel_half = 16 };
enum { kernel_size = kernel_half * 2 + 1 };

typedef struct init_t
{
	float to_rgb [burst_count * 6];
	float to_float [gamma_size];
	float contrast;
	float brightness;
	float artifacts;
	float fringing;
	float kernel [rescale_out * kernel_size * 2];
} init_t;

#define ROTATE_IQ( i, q, sin_b, cos_b ) {\
	float t;\
	t = i * cos_b - q * sin_b;\
	q = i * sin_b + q * cos_b;\
	i = t;\
}

static void init_filters( init_t* impl, snes_ntsc_setup_t const* setup )
{
#if rescale_out > 1
	float kernels [kernel_size * 2];
#else
	float* const kernels = impl->kernel;
#endif

	/* generate luma (y) filter using sinc kernel */
	{
		/* sinc with rolloff (dsf) */
		float const rolloff = 1 + (float) setup->sharpness * (float) 0.032;
		float const maxh = 32;
		float const pow_a_n = (float) pow( rolloff, maxh );
		float sum;
		int i;
		/* quadratic mapping to reduce negative (blurring) range */
		float to_angle = (float) setup->resolution + 1;
		to_angle = PI / maxh * (float) LUMA_CUTOFF * (to_angle * to_angle + 1);

		kernels [kernel_size * 3 / 2] = maxh; /* default center value */
		for ( i = 0; i < kernel_half * 2 + 1; i++ )
		{
			int x = i - kernel_half;
			float angle = x * to_angle;
			/* instability occurs at center point with rolloff very close to 1.0 */
			if ( x || pow_a_n > (float) 1.056 || pow_a_n < (float) 0.981 )
			{
				float rolloff_cos_a = rolloff * (float) cos( angle );
				float num = 1 - rolloff_cos_a -
						pow_a_n * (float) cos( maxh * angle ) +
						pow_a_n * rolloff * (float) cos( (maxh - 1) * angle );
				float den = 1 - rolloff_cos_a - rolloff_cos_a + rolloff * rolloff;
				float dsf = num / den;
				kernels [kernel_size * 3 / 2 - kernel_half + i] = dsf - (float) 0.5;
			}
		}

		/* apply blackman window and find sum */
		sum = 0;
		for ( i = 0; i < kernel_half * 2 + 1; i++ )
		{
			float x = PI * 2 / (kernel_half * 2) * i;
			float blackman = 0.42f - 0.5f * (float) cos( x ) + 0.08f * (float) cos( x * 2 );
			sum += (kernels [kernel_size * 3 / 2 - kernel_half + i] *= blackman);
		}

		/* normalize kernel */
		sum = 1.0f / sum;
		for ( i = 0; i < kernel_half * 2 + 1; i++ )
		{
			int x = kernel_size * 3 / 2 - kernel_half + i;
			kernels [x] *= sum;
			assert( kernels [x] == kernels [x] ); /* catch numerical instability */
		}
	}

	/* generate chroma (iq) filter using gaussian kernel */
	{
		float const cutoff_factor = -0.03125f;
		float cutoff = (float) setup->bleed;
		int i;

		if ( cutoff < 0 )
		{
			/* keep extreme value accessible only near upper end of scale (1.0) */
			cutoff *= cutoff;
			cutoff *= cutoff;
			cutoff *= cutoff;
			cutoff *= -30.0f / 0.65f;
		}
		cutoff = cutoff_factor - 0.65f * cutoff_factor * cutoff;

		for ( i = -kernel_half; i <= kernel_half; i++ )
			kernels [kernel_size / 2 + i] = (float) exp( i * i * cutoff );

		/* normalize even and odd phases separately */
		for ( i = 0; i < 2; i++ )
		{
			float sum = 0;
			int x;
			for ( x = i; x < kernel_size; x += 2 )
				sum += kernels [x];

			sum = 1.0f / sum;
			for ( x = i; x < kernel_size; x += 2 )
			{
				kernels [x] *= sum;
				assert( kernels [x] == kernels [x] ); /* catch numerical instability */
			}
		}
	}

	/*
	printf( "luma:\n" );
	for ( i = kernel_size; i < kernel_size * 2; i++ )
		printf( "%f\n", kernels [i] );
	printf( "chroma:\n" );
	for ( i = 0; i < kernel_size; i++ )
		printf( "%f\n", kernels [i] );
	*/

	/* generate linear rescale kernels */
	#if rescale_out > 1
	{
		float weight = 1.0f;
		float* out = impl->kernel;
		int n = rescale_out;
		do
		{
			float remain = 0;
			int i;
			weight -= 1.0f / rescale_in;
			for ( i = 0; i < kernel_size * 2; i++ )
			{
				float cur = kernels [i];
				float m = cur * weight;
				*out++ = m + remain;
				remain = cur - m;
			}
		}
		while ( --n );
	}
	#endif
}

static float const default_decoder [6] =
	{ 0.956f, 0.621f, -0.272f, -0.647f, -1.105f, 1.702f };

static void init( init_t* impl, snes_ntsc_setup_t const* setup )
{
	impl->brightness = (float) setup->brightness * (0.5f * rgb_unit) + rgb_offset;
	impl->contrast   = (float) setup->contrast   * (0.5f * rgb_unit) + rgb_unit;
	#ifdef default_palette_contrast
		if ( !setup->palette )
			impl->contrast *= default_palette_contrast;
	#endif

	impl->artifacts = (float) setup->artifacts;
	if ( impl->artifacts > 0 )
		impl->artifacts *= artifacts_max - artifacts_mid;
	impl->artifacts = impl->artifacts * artifacts_mid + artifacts_mid;

	impl->fringing = (float) setup->fringing;
	if ( impl->fringing > 0 )
		impl->fringing *= fringing_max - fringing_mid;
	impl->fringing = impl->fringing * fringing_mid + fringing_mid;

	init_filters( impl, setup );

	/* generate gamma table */
	if ( gamma_size > 1 )
	{
		float const to_float = 1.0f / (gamma_size - (gamma_size > 1));
		float const gamma = 1.1333f - (float) setup->gamma * 0.5f;
		/* match common PC's 2.2 gamma to TV's 2.65 gamma */
		int i;
		for ( i = 0; i < gamma_size; i++ )
			impl->to_float [i] =
					(float) pow( i * to_float, gamma ) * impl->contrast + impl->brightness;
	}

	/* setup decoder matricies */
	{
		float hue = (float) setup->hue * PI + PI / 180 * ext_decoder_hue;
		float sat = (float) setup->saturation + 1;
		float const* decoder = setup->decoder_matrix;
		if ( !decoder )
		{
			decoder = default_decoder;
			hue += PI / 180 * (std_decoder_hue - ext_decoder_hue);
		}

		{
			float s = (float) sin( hue ) * sat;
			float c = (float) cos( hue ) * sat;
			float* out = impl->to_rgb;
			int n;

			n = burst_count;
			do
			{
				float const* in = decoder;
				int n = 3;
				do
				{
					float i = *in++;
					float q = *in++;
					*out++ = i * c - q * s;
					*out++ = i * s + q * c;
				}
				while ( --n );
				if ( burst_count <= 1 )
					break;
				ROTATE_IQ( s, c, 0.866025f, -0.5f ); /* +120 degrees */
			}
			while ( --n );
		}
	}
}

/* kernel generation */

#define RGB_TO_YIQ( r, g, b, y, i ) (\
	(y = (r) * 0.299f + (g) * 0.587f + (b) * 0.114f),\
	(i = (r) * 0.596f - (g) * 0.275f - (b) * 0.321f),\
	((r) * 0.212f - (g) * 0.523f + (b) * 0.311f)\
)

#define YIQ_TO_RGB( y, i, q, to_rgb, type, r, g ) (\
	r = (type) (y + to_rgb [0] * i + to_rgb [1] * q),\
	g = (type) (y + to_rgb [2] * i + to_rgb [3] * q),\
	(type) (y + to_rgb [4] * i + to_rgb [5] * q)\
)

#define PACK_RGB( r, g, b ) ((r) << 21 | (g) << 11 | (b) << 1)

enum { rgb_kernel_size = burst_size / alignment_count };
enum { rgb_bias = rgb_unit * 2 * snes_ntsc_rgb_builder };

typedef struct pixel_info_t
{
	int offset;
	float negate;
	float kernel [4];
} pixel_info_t;

#if rescale_in > 1
	#define PIXEL_OFFSET_( ntsc, scaled ) \
		(kernel_size / 2 + ntsc + (scaled != 0) + (rescale_out - scaled) % rescale_out + \
				(kernel_size * 2 * scaled))

	#define PIXEL_OFFSET( ntsc, scaled ) \
		PIXEL_OFFSET_( ((ntsc) - (scaled) / rescale_out * rescale_in),\
				(((scaled) + rescale_out * 10) % rescale_out) ),\
		(1.0f - (((ntsc) + 100) & 2))
#else
	#define PIXEL_OFFSET( ntsc, scaled ) \
		(kernel_size / 2 + (ntsc) - (scaled)),\
		(1.0f - (((ntsc) + 100) & 2))
#endif

extern pixel_info_t const snes_ntsc_pixels [alignment_count];

/* Generate pixel at all burst phases and column alignments */
static void gen_kernel( init_t* impl, float y, float i, float q, snes_ntsc_rgb_t* out )
{
	/* generate for each scanline burst phase */
	float const* to_rgb = impl->to_rgb;
	int burst_remain = burst_count;
	y -= rgb_offset;
	do
	{
		/* Encode yiq into *two* composite signals (to allow control over artifacting).
		Convolve these with kernels which: filter respective components, apply
		sharpening, and rescale horizontally. Convert resulting yiq to rgb and pack
		into integer. Based on algorithm by NewRisingSun. */
		pixel_info_t const* pixel = snes_ntsc_pixels;
		int alignment_remain = alignment_count;
		do
		{
			/* negate is -1 when composite starts at odd multiple of 2 */
			float const yy = y * impl->fringing * pixel->negate;
			float const ic0 = (i + yy) * pixel->kernel [0];
			float const qc1 = (q + yy) * pixel->kernel [1];
			float const ic2 = (i - yy) * pixel->kernel [2];
			float const qc3 = (q - yy) * pixel->kernel [3];

			float const factor = impl->artifacts * pixel->negate;
			float const ii = i * factor;
			float const yc0 = (y + ii) * pixel->kernel [0];
			float const yc2 = (y - ii) * pixel->kernel [2];

			float const qq = q * factor;
			float const yc1 = (y + qq) * pixel->kernel [1];
			float const yc3 = (y - qq) * pixel->kernel [3];

			float const* k = &impl->kernel [pixel->offset];
			int n;
			++pixel;
			for ( n = rgb_kernel_size; n; --n )
			{
				float i = k[0]*ic0 + k[2]*ic2;
				float q = k[1]*qc1 + k[3]*qc3;
				float y = k[kernel_size+0]*yc0 + k[kernel_size+1]*yc1 +
				          k[kernel_size+2]*yc2 + k[kernel_size+3]*yc3 + rgb_offset;
				if ( rescale_out <= 1 )
					k--;
				else if ( k < &impl->kernel [kernel_size * 2 * (rescale_out - 1)] )
					k += kernel_size * 2 - 1;
				else
					k -= kernel_size * 2 * (rescale_out - 1) + 2;
				{
					int r, g, b = YIQ_TO_RGB( y, i, q, to_rgb, int, r, g );
					*out++ = PACK_RGB( r, g, b ) - rgb_bias;
				}
			}
		}
		while ( alignment_count > 1 && --alignment_remain );

		if ( burst_count <= 1 )
			break;

		to_rgb += 6;

		ROTATE_IQ( i, q, -0.866025f, -0.5f ); /* -120 degrees */
	}
	while ( --burst_remain );
}

static void correct_errors( snes_ntsc_rgb_t color, snes_ntsc_rgb_t* out );

#if DISABLE_CORRECTION
	#define CORRECT_ERROR( a ) { out [i] += rgb_bias; }
	#define DISTRIBUTE_ERROR( a, b, c ) { out [i] += rgb_bias; }
#else
	#define CORRECT_ERROR( a ) { out [a] += error; }
	#define DISTRIBUTE_ERROR( a, b, c ) {\
		snes_ntsc_rgb_t fourth = (error + 2 * snes_ntsc_rgb_builder) >> 2;\
		fourth &= (rgb_bias >> 1) - snes_ntsc_rgb_builder;\
		fourth -= rgb_bias >> 2;\
		out [a] += fourth;\
		out [b] += fourth;\
		out [c] += fourth;\
		out [i] += error - (fourth * 3);\
	}
#endif

#define RGB_PALETTE_OUT( rgb, out_ )\
{\
	unsigned char* out = (out_);\
	snes_ntsc_rgb_t clamped = (rgb);\
	SNES_NTSC_CLAMP_( clamped, (8 - rgb_bits) );\
	out [0] = (unsigned char) (clamped >> 21);\
	out [1] = (unsigned char) (clamped >> 11);\
	out [2] = (unsigned char) (clamped >>  1);\
}

/* blitter related */

#ifndef restrict
	#if defined (__GNUC__)
		#define restrict __restrict__
	#elif defined (_MSC_VER) && _MSC_VER > 1300
		#define restrict __restrict
	#else
		/* no support for restricted pointers */
		#define restrict
	#endif
#endif

#include <limits.h>

#if SNES_NTSC_OUT_DEPTH <= 16
	#if USHRT_MAX == 0xFFFF
		typedef unsigned short snes_ntsc_out_t;
	#else
		#error "Need 16-bit int type"
	#endif

#else
	#if UINT_MAX == 0xFFFFFFFF
		typedef unsigned int  snes_ntsc_out_t;
	#elif ULONG_MAX == 0xFFFFFFFF
		typedef unsigned long snes_ntsc_out_t;
	#else
		#error "Need 32-bit int type"
	#endif

#endif