// SONIC ROBO BLAST 2
//-----------------------------------------------------------------------------
// Copyright (C) 1993-1996 by id Software, Inc.
// Copyright (C) 1998-2000 by DooM Legacy Team.
// Copyright (C) 1999-2018 by Sonic Team Junior.
//
// This program is free software distributed under the
// terms of the GNU General Public License, version 2.
// See the 'LICENSE' file for more details.
//-----------------------------------------------------------------------------
/// \file  r_draw.c
/// \brief span / column drawer functions, for 8bpp and 16bpp
///        All drawing to the view buffer is accomplished in this file.
///        The other refresh files only know about ccordinates,
///        not the architecture of the frame buffer.
///        The frame buffer is a linear one, and we need only the base address.

#include <algorithm>

#include "doomdef.h"
#include "doomstat.h"
#include "r_local.h"
#include "r_things.h"
#include "st_stuff.h" // need ST_HEIGHT
#include "i_video.h"
#include "v_video.h"
#include "m_misc.h"
#include "w_wad.h"
#include "z_zone.h"
#include "console.h" // Until buffering gets finished
#include "k_kart.h" // SRB2kart
#include "i_threads.h"

#ifdef HWRENDER
#include "hardware/hw_main.h"
#endif

// ==========================================================================
//                     COMMON DATA FOR 8bpp AND 16bpp
// ==========================================================================

/**	\brief view info
*/
INT32 viewwidth = 0, viewheight = 0, viewwindowx = 0, viewwindowy = 0;

// =========================================================================
//                      COLUMN DRAWING CODE STUFF
// =========================================================================

drawcolumndata_t g_dc = {};

// -----------------------
// translucency stuff here
// -----------------------
#define NUMTRANSTABLES 9 // how many translucency tables are used

UINT8 *transtables = NULL; // translucency tables
UINT8 *blendtables[NUMBLENDMAPS] = {};

// --------------------------------------------
// c drawer routines
// --------------------------------------------

coldrawfunc_t *colfunc = NULL;
coldrawfunc_t *colfuncs[COLDRAWFUNC_MAX] = {};
int colfunctype = 0;

// =========================================================================
//                      SPAN DRAWING CODE STUFF
// =========================================================================

drawspandata_t g_ds = {};

// Vectors for Software's tilted slope drawers
floatv3_t *ds_su = NULL, *ds_sv = NULL, *ds_sz = NULL;

float focallengthf = 0.0f;

// For, uh, tilted lighting, duh.
//static INT32 *tiltlighting;

// --------------------------------------------
// c drawer routines
// --------------------------------------------

spandrawfunc_t *spanfunc = NULL;
spandrawfunc_t *spanfuncs[SPANDRAWFUNC_MAX] = {};

// ==========================================================================
//                        OLD DOOM FUZZY EFFECT
// ==========================================================================

// =========================================================================
//                   TRANSLATION COLORMAP CODE
// =========================================================================

#define DEFAULT_TT_CACHE_INDEX MAXSKINS
#define BOSS_TT_CACHE_INDEX (MAXSKINS + 1)
#define METALSONIC_TT_CACHE_INDEX (MAXSKINS + 2)
#define ALLWHITE_TT_CACHE_INDEX (MAXSKINS + 3)
#define RAINBOW_TT_CACHE_INDEX (MAXSKINS + 4)
#define BLINK_TT_CACHE_INDEX (MAXSKINS + 5)
#define TT_CACHE_SIZE (MAXSKINS + 6)
#define SKIN_RAMP_LENGTH 16
#define DEFAULT_STARTTRANSCOLOR 160
#define NUM_PALETTE_ENTRIES 256

static UINT8 **translationtablecache[TT_CACHE_SIZE] = {};
static UINT8 **localtranslationtablecache[MAXLOCALSKINS] = {};

CV_PossibleValue_t Color_cons_t[MAXSKINCOLORS+1] = {};

struct GenerateBlendTables_State
{
	RGBA_t *LocalPalette;
	RGBA_t *gammaCorrectedPalette;
};

static void R_GenerateBlendTables_Core(struct GenerateBlendTables_State *state);
static void R_AllocateBlendTables(void);

#ifdef HAVE_THREADS
static void R_GenerateBlendTables_Thread(void *userdata)
{
	struct GenerateBlendTables_State *state = static_cast<struct GenerateBlendTables_State *>(userdata);

	R_GenerateBlendTables_Core(state);

	free(state->LocalPalette);
	free(state->gammaCorrectedPalette);
	free(state);
}
#endif

/** \brief Initializes the translucency tables used by the Software renderer.
*/
void R_InitTranslucencyTables(void)
{
	// Load here the transparency lookup tables 'TRANSx0'
	// NOTE: the TRANSx0 resources MUST BE aligned on 64k for the asm
	// optimised code (in other words, transtables pointer low word is 0)
	transtables = static_cast<UINT8*>(Z_Malloc(NUMTRANSTABLES*0x10000, PU_STATIC, NULL));

	W_ReadLump(W_GetNumForName("TRANS10"), transtables);
	W_ReadLump(W_GetNumForName("TRANS20"), transtables+0x10000);
	W_ReadLump(W_GetNumForName("TRANS30"), transtables+0x20000);
	W_ReadLump(W_GetNumForName("TRANS40"), transtables+0x30000);
	W_ReadLump(W_GetNumForName("TRANS50"), transtables+0x40000);
	W_ReadLump(W_GetNumForName("TRANS60"), transtables+0x50000);
	W_ReadLump(W_GetNumForName("TRANS70"), transtables+0x60000);
	W_ReadLump(W_GetNumForName("TRANS80"), transtables+0x70000);
	W_ReadLump(W_GetNumForName("TRANS90"), transtables+0x80000);

	R_AllocateBlendTables();
	R_GenerateBlendTables();
}

static colorlookup_t transtab_lut;
static colorlookup_t transtab_lut_corrected;

static void BlendTab_Translucent(UINT8 *table, int style, UINT8 blendamt)
{
	INT16 bg, fg;
	RGBA_t backrgba, frontrgba, result;

	if (table == NULL)
		I_Error("BlendTab_Translucent: input table was NULL!");

	for (bg = 0; bg < 0x100; bg++)
	{
		backrgba = pGammaCorrectedPalette[bg];

		for (fg = 0; fg < 0x100; fg++)
		{
			frontrgba = pGammaCorrectedPalette[fg];

			result.rgba = ASTBlendPixel(backrgba, frontrgba, style, 0xFF);
			result.rgba = ASTBlendPixel(result, frontrgba, AST_TRANSLUCENT, blendamt);

			table[((bg * 0x100) + fg)] = GetColorLUT(&transtab_lut_corrected, result.s.red, result.s.green, result.s.blue); // pGammaCorrectedPalette
		}
	}
}

static void BlendTab_Subtractive(UINT8 *table, int style, UINT8 blendamt)
{
	INT16 bg, fg;
	RGBA_t backrgba, frontrgba, result;

	if (table == NULL)
		I_Error("BlendTab_Subtractive: input table was NULL!");

	if (blendamt == 0xFF)
	{
		memset(table, GetColorLUT(&transtab_lut, 0, 0, 0), 0x10000);
		return;
	}

	for (bg = 0; bg < 0x100; bg++)
	{
		backrgba = pLocalPalette[bg]; // intentionally uses pLocalPalette

		for (fg = 0; fg < 0x100; fg++)
		{
			frontrgba = pLocalPalette[fg];

			result.rgba    = ASTBlendPixel(backrgba, frontrgba, style, 0xFF);
			result.s.red   = std::max(0, result.s.red - blendamt);
			result.s.green = std::max(0, result.s.green - blendamt);
			result.s.blue  = std::max(0, result.s.blue - blendamt);

			table[((bg * 0x100) + fg)] = GetColorLUT(&transtab_lut, result.s.red, result.s.green, result.s.blue);
		}
	}
}

static void BlendTab_Modulative(UINT8 *table)
{
	INT16 bg, fg;
	RGBA_t backrgba, frontrgba, result;

	if (table == NULL)
		I_Error("BlendTab_Modulative: input table was NULL!");

	for (bg = 0; bg < 0x100; bg++)
	{
		backrgba = pGammaCorrectedPalette[bg];

		for (fg = 0; fg < 0x100; fg++)
		{
			frontrgba = pGammaCorrectedPalette[fg];

			result.rgba = ASTBlendPixel(backrgba, frontrgba, AST_MODULATE, 0);
			table[((bg * 0x100) + fg)] = GetColorLUT(&transtab_lut_corrected, result.s.red, result.s.green, result.s.blue);
		}
	}
}

static INT32 BlendTab_Count[NUMBLENDMAPS] =
{
	NUMTRANSTABLES+1, // blendtab_add
	NUMTRANSTABLES+1, // blendtab_subtract
	NUMTRANSTABLES+1, // blendtab_reversesubtract
	1                 // blendtab_modulate
};

static INT32 BlendTab_FromStyle[] =
{
	0,                        // AST_COPY
	0,                        // AST_TRANSLUCENT
	blendtab_add,             // AST_ADD
	blendtab_subtract,        // AST_SUBTRACT
	blendtab_reversesubtract, // AST_REVERSESUBTRACT
	blendtab_modulate,        // AST_MODULATE
	0                         // AST_OVERLAY
};

static void BlendTab_GenerateMaps(INT32 tab, INT32 style, void (*genfunc)(UINT8 *, int, UINT8))
{
	INT32 i = 0, num = BlendTab_Count[tab];
	const float amtmul = (256.0f / (float)(NUMTRANSTABLES + 1));
	for (; i < num; i++)
	{
		const size_t offs = (0x10000 * i);
		const UINT16 alpha = std::fmin(amtmul * i, 0xFF);
		genfunc(blendtables[tab] + offs, style, alpha);
	}
}

static void R_AllocateBlendTables(void)
{
	INT32 i;

	for (i = 0; i < NUMBLENDMAPS; i++)
	{
		blendtables[i] = static_cast<UINT8*>(Z_Malloc(BlendTab_Count[i] * 0x10000, PU_STATIC, NULL));
	}
}

void R_GenerateBlendTables(void)
{
#ifdef HAVE_THREADS
	// Allocate copies for the worker thread since the originals can be freed in the main thread.
	struct GenerateBlendTables_State *state = static_cast<struct GenerateBlendTables_State *>(malloc(sizeof *state));
	size_t palsize = 256 * sizeof(RGBA_t);

	state->LocalPalette = static_cast<RGBA_t *>(memcpy(malloc(palsize), pLocalPalette, palsize));
	state->gammaCorrectedPalette = static_cast<RGBA_t *>(memcpy(malloc(palsize), pGammaCorrectedPalette, palsize));

	I_spawn_thread("blend-tables",
			R_GenerateBlendTables_Thread, state);
#else
	struct GenerateBlendTables_State state = {pLocalPalette, pGammaCorrectedPalette};
	R_GenerateBlendTables_Core(&state);
#endif
}

static void R_GenerateBlendTables_Core(struct GenerateBlendTables_State *state)
{
	InitColorLUT(&transtab_lut, state->LocalPalette, false);
	InitColorLUT(&transtab_lut_corrected, state->gammaCorrectedPalette, false);

	// Additive
	BlendTab_GenerateMaps(blendtab_add, AST_ADD, BlendTab_Translucent);

	// Subtractive
#if 1
	BlendTab_GenerateMaps(blendtab_subtract, AST_SUBTRACT, BlendTab_Subtractive);
#else
	BlendTab_GenerateMaps(blendtab_subtract, AST_SUBTRACT, BlendTab_Translucent);
#endif

	// Reverse subtractive
	BlendTab_GenerateMaps(blendtab_reversesubtract, AST_REVERSESUBTRACT, BlendTab_Translucent);

	// Modulative blending only requires a single table
	BlendTab_Modulative(blendtables[blendtab_modulate]);
}

#define ClipBlendLevel(style, trans) std::max(std::min((trans), BlendTab_Count[BlendTab_FromStyle[style]]-1), 0)
#define ClipTransLevel(trans) std::max(std::min((trans), NUMTRANSMAPS-2), 0)

UINT8 *R_GetTranslucencyTable(INT32 alphalevel)
{
	return transtables + (ClipTransLevel(alphalevel-1) << FF_TRANSSHIFT);
}

UINT8 *R_GetBlendTable(int style, INT32 alphalevel)
{
	size_t offs;

	if (style <= AST_COPY || style >= AST_OVERLAY)
		return transtables + (ClipTransLevel(alphalevel) << FF_TRANSSHIFT);

	offs = (ClipBlendLevel(style, alphalevel) << FF_TRANSSHIFT);

	// Lactozilla: Returns the equivalent to AST_TRANSLUCENT
	// if no alpha style matches any of the blend tables.
	switch (style)
	{
		case AST_ADD:
			return blendtables[blendtab_add] + offs;
		case AST_SUBTRACT:
			return blendtables[blendtab_subtract] + offs;
		case AST_REVERSESUBTRACT:
			return blendtables[blendtab_reversesubtract] + offs;
		case AST_MODULATE:
			return blendtables[blendtab_modulate];
		default:
			break;
	}

	// Return a normal translucency table
	if (--alphalevel >= 0)
		return transtables + (ClipTransLevel(alphalevel) << FF_TRANSSHIFT);

	return NULL;
}

boolean R_BlendLevelVisible(INT32 blendmode, INT32 alphalevel)
{
	if (blendmode <= AST_COPY || blendmode == AST_SUBTRACT ||
		blendmode == AST_MODULATE || blendmode >= AST_OVERLAY)
		return true;

	return (alphalevel < BlendTab_Count[BlendTab_FromStyle[blendmode]]);
}

/**	\brief	Retrieves a translation colormap from the cache.

	\param	skinnum		skin number, or a translation mode
	\param	color		translation color
	\param	flags		set GTC_CACHE to use the cache

	\return	Colormap. If not cached, caller should Z_Free.
*/
static UINT8* RGetTranslationColormap(INT32 skinnum, skincolors_t color, UINT8 flags, boolean local)
{
	UINT8 ***tt;
	UINT8* ret;
	INT32 skintableindex;

	if (local)
	{
		tt = localtranslationtablecache;
		skintableindex = skinnum;
	}
	else
	{
		tt = translationtablecache;

		// Adjust if we want the default colormap
		switch (skinnum)
		{
			case TC_DEFAULT:    skintableindex = DEFAULT_TT_CACHE_INDEX; break;
			case TC_BOSS:       skintableindex = BOSS_TT_CACHE_INDEX; break;
			case TC_METALSONIC: skintableindex = METALSONIC_TT_CACHE_INDEX; break;
			case TC_ALLWHITE:   skintableindex = ALLWHITE_TT_CACHE_INDEX; break;
			case TC_RAINBOW:    skintableindex = RAINBOW_TT_CACHE_INDEX; break;
			case TC_BLINK:      skintableindex = BLINK_TT_CACHE_INDEX; break;
			default:            skintableindex = skinnum; break;
		}
	}

	if (flags & GTC_CACHE)
	{
		// Allocate table for skin if necessary
		if (!tt[skintableindex])
			tt[skintableindex] = static_cast<UINT8**>(Z_Calloc(MAXTRANSLATIONS * sizeof(UINT8**), PU_STATIC, NULL));

		// Get colormap
		ret = tt[skintableindex][color];
	}
	else
		ret = NULL;

	// Generate the colormap if necessary
	if (!ret)
	{
		ret = static_cast<UINT8*>(Z_Malloc(NUM_PALETTE_ENTRIES, PU_STATIC, NULL));
		K_GenerateKartColormap(ret, skinnum, color, local); //R_GenerateTranslationColormap(ret, skinnum, color);		// SRB2kart

		// Cache the colormap if desired
		if (flags & GTC_CACHE)
			tt[skintableindex][color] = ret;
	}

	return ret;
}

UINT8* R_GetTranslationColormap(INT32 skinnum, skincolors_t color, UINT8 flags)
{
	return RGetTranslationColormap(skinnum, color, flags, false);
}

UINT8* R_GetLocalTranslationColormap(skin_t *skin, skin_t *localskin, skincolors_t color, UINT8 flags, boolean local)
{
	return RGetTranslationColormap(K_GetMobjLocalSkinNum(skin, localskin, local), color, flags, (localskin && local));
}

patch_t* R_GetSkinFaceRank(player_t* ply)
{
	if (ply->localskin)
	{
		if (ply->skinlocal)
			return localfacerankprefix[ply->localskin - 1];

		return facerankprefix[ply->localskin - 1];
	}

	return facerankprefix[ply->skin];
}

patch_t* R_GetSkinFaceWant(player_t* ply)
{
	if (ply->localskin)
	{
		if (ply->skinlocal)
			return localfacewantprefix[ply->localskin - 1];

		return facewantprefix[ply->localskin - 1];
	}

	return facewantprefix[ply->skin];
}

patch_t* R_GetSkinFaceMini(player_t* ply)
{
	if (ply->localskin)
	{
		if (ply->skinlocal)
			return localfacemmapprefix[ply->localskin - 1];

		return facemmapprefix[ply->localskin - 1];
	}

	return facemmapprefix[ply->skin];
}

/**	\brief	Flushes cache of translation colormaps.

	Flushes cache of translation colormaps, but doesn't actually free the
	colormaps themselves. These are freed when PU_LEVEL blocks are purged,
	at or before which point, this function should be called.

	\return	void
*/

/* There seems to be no reason to free cached colormaps tho? Not like they can change :ChonkBunckle: */

/*
void R_FlushTranslationColormapCache(void)
{
	INT32 i;

	for (i = 0; i < (INT32)(sizeof(translationtablecache) / sizeof(translationtablecache[0])); i++)
		if (translationtablecache[i])
			memset(translationtablecache[i], 0, MAXTRANSLATIONS * sizeof(UINT8**));

	for (i = 0; i < (INT32)(sizeof(localtranslationtablecache) / sizeof(localtranslationtablecache[0])); i++)
		if (localtranslationtablecache[i])
			memset(localtranslationtablecache[i], 0, MAXTRANSLATIONS * sizeof(UINT8**));
}
*/

// ==========================================================================
//               COMMON DRAWER FOR 8 AND 16 BIT COLOR MODES
// ==========================================================================

// in a perfect world, all routines would be compatible for either mode,
// and optimised enough
//
// in reality, the few routines that can work for either mode, are
// put here

enum ColumnFlushType
{
	FLUSH_NONE				= 0x0000,
	FLUSH_OPAQUE			= 0x0001,
	FLUSH_TRANS				= 0x0002,
	FLUSH_COLORMAP			= 0x0004,
	FLUSH_COLORMAP_TRANS	= 0x0008,
};

typedef struct drawcolumndata_temp_s
{
	INT32    x;
	INT32    yl[8], yh[8];

	// e6y: resolution limitation is removed
	UINT8 *buf;

	INT32 startx;

	ColumnFlushType type;

	INT32 commontop, commonbot;

	UINT8 *transmap;
	// SoM 7-28-04: Fix the fuzz problem.
	UINT8 *translation;
} drawcolumndata_temp_t;

drawcolumndata_temp_t temp_dc = {};

//
// Error functions that will abort if R_FlushColumns tries to flush
// columns without a column type.
//
FUNCNORETURN static ATTRNORETURN void R_FlushWholeError(void)
{
	I_Error("R_FlushWholeColumns called without being initialized.\n");
}

FUNCNORETURN static ATTRNORETURN void R_FlushHTError(void)
{
	I_Error("R_FlushHTColumns called without being initialized.\n");
}

FUNCNORETURN static ATTRNORETURN void R_QuadFlushError(void)
{
	I_Error("R_FlushQuadColumn called without being initialized.\n");
}

static void (*R_FlushWholeColumns)(void) = R_FlushWholeError;
static void (*R_FlushHTColumns)(void)    = R_FlushHTError;
static void (*R_FlushQuadColumn)(void)   = R_QuadFlushError;

static void R_FlushColumns(void)
{
	if (temp_dc.x != 8 || temp_dc.commontop >= temp_dc.commonbot)
		R_FlushWholeColumns();
	else
	{
		R_FlushHTColumns();
		R_FlushQuadColumn();
	}

	temp_dc.x = 0;
}

//
// R_ResetColumnBuffer
//
// haleyjd 09/13/04: new function to call from main rendering loop
// which gets rid of the unnecessary reset of various variables during
// column drawing.
//
void R_ResetColumnBuffer(void)
{
	// haleyjd 10/06/05: this must not be done if x == 0!
	if (temp_dc.x)
	{
		R_FlushColumns();
	}

	temp_dc.type = FLUSH_NONE;
	R_FlushWholeColumns = R_FlushWholeError;
	R_FlushHTColumns    = R_FlushHTError;
	R_FlushQuadColumn   = R_QuadFlushError;
}

/**	\brief	The R_InitViewBuffer function

	Creates lookup tables for getting the framebuffer address
	of a pixel to draw.

	\param	width	witdh of buffer
	\param	height	hieght of buffer

	\return	void
*/

static void R_AllocViewMemory(void)
{
	negonearray       = static_cast<INT16*>(Z_Realloc(negonearray, sizeof(*negonearray) * viewwidth, PU_STATIC, NULL));
	screenheightarray = static_cast<INT16*>(Z_Realloc(screenheightarray, sizeof(*screenheightarray) * viewwidth, PU_STATIC, NULL));

	floorclip         = static_cast<INT16*>(Z_Realloc(floorclip, sizeof(*floorclip) * viewwidth, PU_STATIC, NULL));
	ceilingclip       = static_cast<INT16*>(Z_Realloc(ceilingclip, sizeof(*ceilingclip) * viewwidth, PU_STATIC, NULL));

	frontscale        = static_cast<fixed_t*>(Z_Realloc(frontscale, sizeof(*frontscale) * viewwidth, PU_STATIC, NULL));

	xtoviewangle      = static_cast<angle_t*>(Z_Realloc(xtoviewangle, sizeof(*xtoviewangle) * (viewwidth + 1), PU_STATIC, NULL));

	R_AllocSegMemory();
	R_AllocClipSegMemory();
	R_AllocPlaneMemory();
#ifdef FLOORSPLATS
	R_AllocFloorSpriteTables();
#endif
	R_AllocVisSpriteMemory();
}

void R_InitViewBuffer(void)
{
	R_AllocViewMemory();

	viewwindowx = 0;
	viewwindowy = 0;

	INT32 bufsize = (vid.width * 8) * sizeof(*temp_dc.buf);

	if (temp_dc.buf)
	{
#if defined(__SSE__)
		aligned_free(temp_dc.buf);
#else
		Z_Free(temp_dc.buf);
#endif
	}

	memset(&temp_dc, 0, sizeof(temp_dc));

#if defined(__SSE__)
	while (bufsize & 15)
		bufsize++;
	temp_dc.buf = static_cast<UINT8*>(aligned_alloc(16, bufsize));
#else
	temp_dc.buf = static_cast<UINT8*>(Z_Malloc(bufsize, PU_STATIC, NULL));
#endif

	memset(temp_dc.buf, 0, bufsize);
}

/**	\brief	The R_VideoErase function

	Copy a screen buffer.

	\param	ofs	offest from buffer
	\param	count	bytes to erase

	\return	void


*/
void R_VideoErase(size_t ofs, INT32 count)
{
	// LFB copy.
	// This might not be a good idea if memcpy
	//  is not optimal, e.g. byte by byte on
	//  a 32bit CPU, as GNU GCC/Linux libc did
	//  at one point.
	memcpy(vid.screens[0] + ofs, vid.screens[1] + ofs, count);
}

// ==========================================================================
//                   INCLUDE DRAWING CODE HERE
// ==========================================================================

#include "r_draw_column.cpp"
#include "r_draw_span.cpp"