thirtyflightsofloving/renderer/r_upscale.c

/*
Copyright (C) 2013 Andrey Nazarov

This program 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.

This program 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 this program; if not, write to the Free Software Foundation, Inc.,
51 Franklin Street, Fifth Floor, Boston, MA 02110-1301 USA.
*/

/*
HQ2x filter
authors: byuu and blargg
license: public domain

note: this is a clean reimplementation of the original HQ2x filter, which was
written by Maxim Stepin (MaxSt). it is not 100% identical, but very similar.
*/

//#include "shared/shared.h"
//#include "common/cvar.h"
//#include "refresh/images.h"
#include "r_local.h"

#define q_noinline

#define uint8_t		byte
#define int32_t		int
#define uint32_t	unsigned int
#define uint64_t	uint64

#define UINT64_C(val)	val##ui64
#define FIX(x)			(int)((x) * (1 << 16))

typedef union {
	uint32_t u32;
	uint8_t u8[4];
} diffColor_t;

static const uint8_t hqTable[256] = {
	1, 1, 2,  4, 1, 1, 2,  4, 3,  5,  7,  8, 3,  5, 13, 15,
	1, 1, 2, 10, 1, 1, 2, 10, 3,  5,  8,  8, 3,  5,  6,  8,
	1, 1, 2,  4, 1, 1, 2,  4, 3,  5, 12, 14, 3,  5,  9, 16,
	1, 1, 2, 10, 1, 1, 2, 10, 3,  5,  9,  8, 3,  5,  6, 16,
	1, 1, 2,  4, 1, 1, 2,  4, 3, 11,  8,  8, 3, 11,  9,  8,
	1, 1, 2,  4, 1, 1, 2,  4, 3,  5,  9,  8, 3,  5,  9,  8,
	1, 1, 2,  4, 1, 1, 2,  4, 3, 11,  6,  8, 3, 11,  6, 16,
	1, 1, 2,  4, 1, 1, 2, 10, 3,  5,  9,  8, 3, 11,  6, 16,
	1, 1, 2,  4, 1, 1, 2,  4, 3,  5,  7,  8, 3,  5, 13, 15,
	1, 1, 2,  4, 1, 1, 2,  4, 3,  5,  9,  8, 3,  5,  9,  8,
	1, 1, 2,  4, 1, 1, 2,  4, 3,  5, 12, 14, 3,  5,  9, 16,
	1, 1, 2,  4, 1, 1, 2,  4, 3,  5,  9, 14, 3,  5,  6, 16,
	1, 1, 2,  4, 1, 1, 2,  4, 3,  5,  9,  8, 3,  5,  9, 15,
	1, 1, 2,  4, 1, 1, 2,  4, 3,  5,  9,  8, 3,  5,  6,  8,
	1, 1, 2,  4, 1, 1, 2,  4, 3,  5,  9,  8, 3,  5,  6, 16,
	1, 1, 2,  4, 1, 1, 2,  4, 3,  5,  6,  8, 3,  5,  6, 16,
};

static uint8_t  rotTable[256];
static int32_t  yccTable[8][256];
static int32_t  maxY, maxCb, maxCr;

static q_noinline int diff (uint32_t A_u32, uint32_t B_u32)
{
	diffColor_t A, B;
	int32_t a, b;

	A.u32 = A_u32;
	B.u32 = B_u32;

	if (A.u8[3] == 0 && B.u8[3] == 0)
		return 0;

	if (A.u8[3] == 0 || B.u8[3] == 0)
		return 1;

	a = yccTable[0][A.u8[0]] + yccTable[1][A.u8[1]] + yccTable[2][A.u8[2]];
	b = yccTable[0][B.u8[0]] + yccTable[1][B.u8[1]] + yccTable[2][B.u8[2]];
	if (abs(a - b) > maxY)
		return 1;

	a = yccTable[3][A.u8[0]] + yccTable[4][A.u8[1]] + yccTable[5][A.u8[2]];
	b = yccTable[3][B.u8[0]] + yccTable[4][B.u8[1]] + yccTable[5][B.u8[2]];
	if (abs(a - b) > maxCb)
		return 1;

	a = yccTable[5][A.u8[0]] + yccTable[6][A.u8[1]] + yccTable[7][A.u8[2]];
	b = yccTable[5][B.u8[0]] + yccTable[6][B.u8[1]] + yccTable[7][B.u8[2]];
	if (abs(a - b) > maxCr)
		return 1;

	return 0;
}

static /*inline*/ int same (uint32_t A, uint32_t B)
{
	return !diff(A, B);
}

static /*inline*/ uint64_t grow (uint64_t n)
{
	n |= n << 24;
	n &= UINT64_C(0xff00ff00ff00ff);
	return n;
}

static /*inline*/ uint32_t pack (uint64_t n)
{
	n &= UINT64_C(0xff00ff00ff00ff);
	n |= n >> 24;
	return n;
}

static q_noinline uint32_t blend_1_1 (uint32_t A, uint32_t B)
{
	return pack((grow(A) + grow(B)) >> 1);
}

static q_noinline uint32_t blend_3_1 (uint32_t A, uint32_t B)
{
	return pack((grow(A) * 3 + grow(B)) >> 2);
}

static q_noinline uint32_t blend_7_1 (uint32_t A, uint32_t B)
{
	return pack((grow(A) * 7 + grow(B)) >> 3);
}

static q_noinline uint32_t blend_5_3 (uint32_t A, uint32_t B)
{
	return pack((grow(A) * 5 + grow(B) * 3) >> 3);
}

static q_noinline uint32_t blend_2_1_1 (uint32_t A, uint32_t B, uint32_t C)
{
	return pack((grow(A) * 2 + grow(B) + grow(C)) >> 2);
}

static q_noinline uint32_t blend_5_2_1 (uint32_t A, uint32_t B, uint32_t C)
{
	return pack((grow(A) * 5 + grow(B) * 2 + grow(C)) >> 3);
}

static q_noinline uint32_t blend_6_1_1 (uint32_t A, uint32_t B, uint32_t C)
{
	return pack((grow(A) * 6 + grow(B) + grow(C)) >> 3);
}

static q_noinline uint32_t blend_2_3_3 (uint32_t A, uint32_t B, uint32_t C)
{
	return pack((grow(A) * 2 + (grow(B) + grow(C)) * 3) >> 3);
}

static q_noinline uint32_t blend_14_1_1 (uint32_t A, uint32_t B, uint32_t C)
{
	return pack((grow(A) * 14 + grow(B) + grow(C)) >> 4);
}

static q_noinline uint32_t R_Upscale2x_blend (int rule, uint32_t E, uint32_t A, uint32_t B, uint32_t D, uint32_t F, uint32_t H)
{
	switch (rule)
	{
	case 1:
		return blend_2_1_1(E, D, B);
	case 2:
		return blend_2_1_1(E, A, D);
	case 3:
		return blend_2_1_1(E, A, B);
	case 4:
		return blend_3_1(E, D);
	case 5:
		return blend_3_1(E, B);
	case 6:
		return blend_3_1(E, A);
	case 7:
		return same(B, D) ? blend_2_1_1(E, D, B) : blend_3_1(E, A);
	case 8:
		return same(B, D) ? blend_2_1_1(E, D, B) : E;
	case 9:
		return same(B, D) ? blend_6_1_1(E, D, B) : blend_3_1(E, A);
	case 10:
		return same(B, F) ? blend_5_2_1(E, B, D) : blend_3_1(E, D);
	case 11:
		return same(D, H) ? blend_5_2_1(E, D, B) : blend_3_1(E, B);
	case 12:
	case 13:
		return same(B, D) ? blend_2_3_3(E, D, B) : blend_3_1(E, A);
	case 14:
	case 15:
		return same(B, D) ? blend_2_3_3(E, D, B) : E;
	case 16:
		return same(B, D) ? blend_14_1_1(E, D, B) : E;
	default:
	//	Com_Error(ERR_FATAL, "%s: bad rule %d", __func__, rule);
		Com_Error(ERR_FATAL, "HQ2x_blend: bad rule %d", rule);
		return 0;
	}
}

static q_noinline void R_Upscale4x_blend (int rule, uint32_t *p00, uint32_t *p01, uint32_t *p10, uint32_t *p11,
								   uint32_t E, uint32_t A, uint32_t B, uint32_t D, uint32_t F, uint32_t H)
{
	switch (rule)
	{
	case 1:
		*p00 = blend_2_1_1(E, B, D);
		*p01 = blend_5_2_1(E, B, D);
		*p10 = blend_5_2_1(E, D, B);
		*p11 = blend_6_1_1(E, D, B);
		break;
	case 2:
		*p00 = blend_5_3(E, A);
		*p01 = blend_3_1(E, A);
		*p10 = blend_5_2_1(E, D, A);
		*p11 = blend_7_1(E, A);
		break;
	case 3:
		*p00 = blend_5_3(E, A);
		*p01 = blend_5_2_1(E, B, A);
		*p10 = blend_3_1(E, A);
		*p11 = blend_7_1(E, A);
		break;
	case 4:
		*p00 = blend_5_3(E, D);
		*p01 = blend_7_1(E, D);
		*p10 = blend_5_3(E, D);
		*p11 = blend_7_1(E, D);
		break;
	case 5:
		*p00 = blend_5_3(E, B);
		*p01 = blend_5_3(E, B);
		*p10 = blend_7_1(E, B);
		*p11 = blend_7_1(E, B);
		break;
	case 6:
		*p00 = blend_5_3(E, A);
		*p01 = blend_3_1(E, A);
		*p10 = blend_3_1(E, A);
		*p11 = blend_7_1(E, A);
		break;
	case 7:
		if (same(B, D)) {
			*p00 = blend_1_1(B, D);
			*p01 = blend_1_1(B, E);
			*p10 = blend_1_1(D, E);
			*p11 = E;
		} else {
			*p00 = blend_5_3(E, A);
			*p01 = blend_3_1(E, A);
			*p10 = blend_3_1(E, A);
			*p11 = blend_7_1(E, A);
		}
		break;
	case 8:
		if (same(B, D)) {
			*p00 = blend_1_1(B, D);
			*p01 = blend_1_1(B, E);
			*p10 = blend_1_1(D, E);
		}
		else {
			*p00 = E;
			*p01 = E;
			*p10 = E;
		}
		*p11 = E;
		break;
	case 9:
		if (same(B, D)) {
			*p00 = blend_2_1_1(E, B, D);
			*p01 = blend_3_1(E, B);
			*p10 = blend_3_1(E, D);
			*p11 = E;
		}
		else {
			*p00 = blend_5_3(E, A);
			*p01 = blend_3_1(E, A);
			*p10 = blend_3_1(E, A);
			*p11 = blend_7_1(E, A);
		}
		break;
	case 10:
		if (same(B, F)) {
			*p00 = blend_3_1(E, B);
			*p01 = blend_3_1(B, E);
		}
		else {
			*p00 = blend_5_3(E, D);
			*p01 = blend_7_1(E, D);
		}
		*p10 = blend_5_3(E, D);
		*p11 = blend_7_1(E, D);
		break;
	case 11:
		if (same(D, H)) {
			*p00 = blend_3_1(E, D);
			*p10 = blend_3_1(D, E);
		}
		else {
			*p00 = blend_5_3(E, B);
			*p10 = blend_7_1(E, B);
		}
		*p01 = blend_5_3(E, B);
		*p11 = blend_7_1(E, B);
		break;
	case 12:
		if (same(B, D)) {
			*p00 = blend_1_1(B, D);
			*p01 = blend_2_1_1(B, E, D);
			*p10 = blend_5_3(D, B);
			*p11 = blend_6_1_1(E, D, B);
		}
		else {
			*p00 = blend_5_3(E, A);
			*p01 = blend_3_1(E, A);
			*p10 = blend_3_1(E, A);
			*p11 = blend_7_1(E, A);
		}
		break;
	case 13:
		if (same(B, D)) {
			*p00 = blend_1_1(B, D);
			*p01 = blend_5_3(B, D);
			*p10 = blend_2_1_1(D, E, B);
			*p11 = blend_6_1_1(E, D, B);
		}
		else {
			*p00 = blend_5_3(E, A);
			*p01 = blend_3_1(E, A);
			*p10 = blend_3_1(E, A);
			*p11 = blend_7_1(E, A);
		}
		break;
	case 14:
		if (same(B, D)) {
			*p00 = blend_1_1(B, D);
			*p01 = blend_2_1_1(B, E, D);
			*p10 = blend_5_3(D, B);
			*p11 = blend_6_1_1(E, D, B);
		}
		else {
			*p00 = E;
			*p01 = E;
			*p10 = E;
			*p11 = E;
		}
		break;
	case 15:
		if (same(B, D)) {
			*p00 = blend_1_1(B, D);
			*p01 = blend_5_3(B, D);
			*p10 = blend_2_1_1(D, E, B);
			*p11 = blend_6_1_1(E, D, B);
		}
		else {
			*p00 = E;
			*p01 = E;
			*p10 = E;
			*p11 = E;
		}
		break;
	case 16:
		if (same(B, D))
			*p00 = blend_2_1_1(E, B, D);
		else
			*p00 = E;
		*p01 = E;
		*p10 = E;
		*p11 = E;
		break;
	default:
	//	Com_Error(ERR_FATAL, "%s: bad rule %d", __func__, rule);
		Com_Error(ERR_FATAL, "HQ4x_blend: bad rule %d",rule);
		break;
	}
}


/*
===============
R_Upscale2x_Render
===============
*/
//void R_Upscale2x_Render (uint32_t *output, const uint32_t *input, int width, int height)
void R_Upscale2x_Render (unsigned int *output, const unsigned int *input, int width, int height)
{
	int				x, y;
	int				prevline, nextline;
	const uint32_t	*in;
	uint32_t		*out0, *out1;
	size_t			ofs0, ofs1;

	for (y = 0; y < height; y++)
	{
		in = input + y * width;
	//	uint32_t *out0 = output + (y * 2 + 0) * width * 2;
	//	uint32_t *out1 = output + (y * 2 + 1) * width * 2;
		ofs0 = (y * 2 + 0) * width * 2;
		ofs1 = (y * 2 + 1) * width * 2;
		out0 = output + ofs0;
		out1 = output + ofs1;

		prevline = (y == 0 ? 0 : width);
		nextline = (y == height - 1 ? 0 : width);

		for (x = 0; x < width; x++)
		{
			int prev = (x == 0 ? 0 : 1);
			int next = (x == width - 1 ? 0 : 1);

			uint32_t A = *(in - prevline - prev);
			uint32_t B = *(in - prevline);
			uint32_t C = *(in - prevline + next);
			uint32_t D = *(in - prev);
			uint32_t E = *(in);
			uint32_t F = *(in + next);
			uint32_t G = *(in + nextline - prev);
			uint32_t H = *(in + nextline);
			uint32_t I = *(in + nextline + next);

		//	int pattern;
			byte pattern;
			pattern  = diff(E, A) << 0;
			pattern |= diff(E, B) << 1;
			pattern |= diff(E, C) << 2;
			pattern |= diff(E, D) << 3;
			pattern |= diff(E, F) << 4;
			pattern |= diff(E, G) << 5;
			pattern |= diff(E, H) << 6;
			pattern |= diff(E, I) << 7;

			*(out0 + 0) = R_Upscale2x_blend(hqTable[pattern], E, A, B, D, F, H);
			pattern = rotTable[pattern];
			*(out0 + 1) = R_Upscale2x_blend(hqTable[pattern], E, C, F, B, H, D);
			pattern = rotTable[pattern];
			*(out1 + 1) = R_Upscale2x_blend(hqTable[pattern], E, I, H, F, D, B);
			pattern = rotTable[pattern];
			*(out1 + 0) = R_Upscale2x_blend(hqTable[pattern], E, G, D, H, B, F);

			in++;
			out0 += 2;
			out1 += 2;
		}
	}
}


/*
===============
R_Upscale4x_Render
===============
*/
//void R_Upscale4x_Render (unsigned int *output, const unsigned int *input, int width, int height)
void R_Upscale4x_Render (unsigned int *output, const unsigned int *input, int width, int height)
{
	int				x, y;
	int				prevline, nextline;
	const uint32_t	*in;
	uint32_t		*out0, *out1, *out2, *out3;
	size_t			ofs0, ofs1, ofs2, ofs3;

	for (y = 0; y < height; y++)
	{
		in = input + y * width;
	//	uint32_t *out0 = output + (y * 4 + 0) * width * 4;
	//	uint32_t *out1 = output + (y * 4 + 1) * width * 4;
	//	uint32_t *out2 = output + (y * 4 + 2) * width * 4;
	//	uint32_t *out3 = output + (y * 4 + 3) * width * 4;
		ofs0 = (y * 4 + 0) * width * 4;
		ofs1 = (y * 4 + 1) * width * 4;
		ofs2 = (y * 4 + 2) * width * 4;
		ofs3 = (y * 4 + 3) * width * 4;
		out0 = output + ofs0;
		out1 = output + ofs1;
		out2 = output + ofs2;
		out3 = output + ofs3;

		prevline = (y == 0 ? 0 : width);
		nextline = (y == height - 1 ? 0 : width);

		for (x = 0; x < width; x++)
		{
			int prev = (x == 0 ? 0 : 1);
			int next = (x == width - 1 ? 0 : 1);

			uint32_t A = *(in - prevline - prev);
			uint32_t B = *(in - prevline);
			uint32_t C = *(in - prevline + next);
			uint32_t D = *(in - prev);
			uint32_t E = *(in);
			uint32_t F = *(in + next);
			uint32_t G = *(in + nextline - prev);
			uint32_t H = *(in + nextline);
			uint32_t I = *(in + nextline + next);

		//	int pattern;
			byte pattern;
			pattern  = diff(E, A) << 0;
			pattern |= diff(E, B) << 1;
			pattern |= diff(E, C) << 2;
			pattern |= diff(E, D) << 3;
			pattern |= diff(E, F) << 4;
			pattern |= diff(E, G) << 5;
			pattern |= diff(E, H) << 6;
			pattern |= diff(E, I) << 7;

			R_Upscale4x_blend(hqTable[pattern], out0 + 0, out0 + 1, out1 + 0, out1 + 1, E, A, B, D, F, H);
			pattern = rotTable[pattern];
			R_Upscale4x_blend(hqTable[pattern], out0 + 3, out1 + 3, out0 + 2, out1 + 2, E, C, F, B, H, D);
			pattern = rotTable[pattern];
			R_Upscale4x_blend(hqTable[pattern], out3 + 3, out3 + 2, out2 + 3, out2 + 2, E, I, H, F, D, B);
			pattern = rotTable[pattern];
			R_Upscale4x_blend(hqTable[pattern], out3 + 0, out2 + 0, out3 + 1, out2 + 1, E, G, D, H, B, F);

			in++;
			out0 += 4;
			out1 += 4;
			out2 += 4;
			out3 += 4;
		}
	}
}


/*
===============
R_Upscale_Init
===============
*/
void R_Upscale_Init (void)
{
	int		n;

	cvar_t *r_upscale_y  = Cvar_Get("r_upscale_y", "48", 0);	// was CVAR_FILES
	cvar_t *r_upscale_cb = Cvar_Get("r_upscale_cb", "7", 0);	// was CVAR_FILES
	cvar_t *r_upscale_cr = Cvar_Get("r_upscale_cr", "6", 0);	// was CVAR_FILES

	maxY  = FIX(Cvar_ClampValue(r_upscale_y,  0, 256));
	maxCb = FIX(Cvar_ClampValue(r_upscale_cb, 0, 256));
	maxCr = FIX(Cvar_ClampValue(r_upscale_cr, 0, 256));

	for (n = 0; n < 256; n++) {
		rotTable[n] = ((n >> 2) & 0x11) | ((n << 2) & 0x88)
			| ((n & 0x01) << 5) | ((n & 0x08) << 3)
			| ((n & 0x10) >> 3) | ((n & 0x80) >> 5);
	}

	// libjpeg YCbCr coefficients
	for (n = 0; n < 256; n++) {
		yccTable[0][n] =  FIX(0.29900f) * n;
		yccTable[1][n] =  FIX(0.58700f) * n;
		yccTable[2][n] =  FIX(0.11400f) * n;
		yccTable[3][n] = -FIX(0.16874f) * n;
		yccTable[4][n] = -FIX(0.33126f) * n;
		yccTable[5][n] =  FIX(0.50000f) * n;
		yccTable[6][n] = -FIX(0.41869f) * n;
		yccTable[7][n] = -FIX(0.08131f) * n;
	}
}