blankart/src/r_draw.cpp
NepDisk d7ed03fab9 Remove M_Memcpy
Was quite literally a wrapper that did nothing.
2025-11-08 02:00:30 -05:00

609 lines
17 KiB
C++

// BLANKART
//-----------------------------------------------------------------------------
// Copyright (C) 2024 by Kart Krew.
// Copyright (C) 2020 by Sonic Team Junior.
// Copyright (C) 2000 by DooM Legacy Team.
// Copyright (C) 1996 by id Software, Inc.
//
// 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.cpp
/// \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 "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_color.h" // SRB2kart
#include "i_threads.h"
#include "libdivide.h" // used by NPO2 tilted span functions
#ifdef HWRENDER
#include "hardware/hw_main.h"
#endif
#include <tracy/tracy/Tracy.hpp>
// --------------------------------------------
// assembly or c drawer routines for 8bpp/16bpp
// --------------------------------------------
coldrawfunc_t *colfunc;
coldrawfunc_t *colfuncs[COLDRAWFUNC_MAX];
coldrawfunc_t *colfuncs_bm[COLDRAWFUNC_MAX];
int colfunctype;
spandrawfunc_t *spanfunc;
spandrawfunc_t *spanfuncs[SPANDRAWFUNC_MAX];
spandrawfunc_t *spanfuncs_bm[SPANDRAWFUNC_MAX];
spandrawfunc_t *spanfuncs_npo2[SPANDRAWFUNC_MAX];
spandrawfunc_t *spanfuncs_bm_npo2[SPANDRAWFUNC_MAX];
spandrawfunc_t *spanfuncs_flat[SPANDRAWFUNC_MAX];
drawcolumndata_t g_dc;
drawspandata_t g_ds;
// ==========================================================================
// COMMON DATA FOR 8bpp AND 16bpp
// ==========================================================================
/** \brief view info
*/
INT32 viewwidth, viewheight, viewwindowx, viewwindowy;
UINT8 r8_flatcolor;
// =========================================================================
// COLUMN DRAWING CODE STUFF
// =========================================================================
// -----------------------
// translucency stuff here
// -----------------------
#define NUMTRANSTABLES 11 // how many translucency tables are used
UINT8 *transtables; // translucency tables
UINT8 *blendtables[NUMBLENDMAPS];
UINT8 *palremap, *invremap;
/** \brief R_DrawTransColumn uses this
*/
UINT8 *dc_transmap; // one of the translucency tables
// ----------------------
// translation stuff here
// ----------------------
/** \brief R_DrawTranslatedColumn uses this
*/
UINT8 *dc_translation;
struct r_lightlist_t *dc_lightlist = NULL;
INT32 dc_numlights = 0, dc_maxlights;
// =========================================================================
// SPAN DRAWING CODE STUFF
// =========================================================================
// Vectors for Software's tilted slope drawers
floatv3_t *ds_su, *ds_sv, *ds_sz;
float focallengthf[MAXSPLITSCREENPLAYERS];
float zeroheight;
// =========================================================================
// 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 DASHMODE_TT_CACHE_INDEX (MAXSKINS + 6)
#define HITLAG_TT_CACHE_INDEX (MAXSKINS + 7)
#define INTERMISSION_TT_CACHE_INDEX (MAXSKINS + 8)
#define TT_CACHE_SIZE (MAXSKINS + 9)
static UINT8 **translationtablecache[TT_CACHE_SIZE] = {NULL};
UINT8 skincolor_modified[MAXSKINCOLORS];
static INT32 SkinToCacheIndex(INT32 skinnum)
{
switch (skinnum)
{
case TC_DEFAULT: return DEFAULT_TT_CACHE_INDEX;
case TC_BOSS: return BOSS_TT_CACHE_INDEX;
case TC_METALSONIC: return METALSONIC_TT_CACHE_INDEX;
case TC_ALLWHITE: return ALLWHITE_TT_CACHE_INDEX;
case TC_RAINBOW: return RAINBOW_TT_CACHE_INDEX;
case TC_BLINK: return BLINK_TT_CACHE_INDEX;
case TC_DASHMODE: return DASHMODE_TT_CACHE_INDEX;
default: break;
}
return skinnum;
}
static INT32 CacheIndexToSkin(INT32 ttc)
{
switch (ttc)
{
case DEFAULT_TT_CACHE_INDEX: return TC_DEFAULT;
case BOSS_TT_CACHE_INDEX: return TC_BOSS;
case METALSONIC_TT_CACHE_INDEX: return TC_METALSONIC;
case ALLWHITE_TT_CACHE_INDEX: return TC_ALLWHITE;
case RAINBOW_TT_CACHE_INDEX: return TC_RAINBOW;
case BLINK_TT_CACHE_INDEX: return TC_BLINK;
case DASHMODE_TT_CACHE_INDEX: return TC_DASHMODE;
default: break;
}
return ttc;
}
CV_PossibleValue_t Color_cons_t[MAXSKINCOLORS+1];
#define TRANSTAB_AMTMUL10 (255.0f / 10.0f)
struct GenerateBlendTables_State
{
RGBA_t *masterPalette;
RGBA_t *gammaCorrectedPalette;
};
static void R_GenerateBlendTables_Core(struct GenerateBlendTables_State *state);
static void R_GenerateTranslucencyTable(UINT8 *table, RGBA_t* sourcepal, int style, UINT8 blendamt);
static void R_AllocateBlendTables(void)
{
INT32 i;
for (i = 0; i < NUMBLENDMAPS; i++)
{
if (i == blendtab_modulate)
continue;
blendtables[i] = static_cast<UINT8 *>(Z_MallocAlign((NUMTRANSTABLES + 1) * 0x10000, PU_STATIC, NULL, 16));
}
// Modulation blending only requires a single table
blendtables[blendtab_modulate] = static_cast<UINT8 *>(Z_MallocAlign(0x10000, PU_STATIC, NULL, 16));
}
#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->masterPalette);
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 'TINTTAB'
// NOTE: the TINTTAB resource MUST BE aligned on 64k for the asm
// optimised code (in other words, transtables pointer low word is 0)
transtables = static_cast<UINT8 *>(Z_MallocAlign(NUMTRANSTABLES*0x10000, PU_STATIC, NULL, 16));
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();
}
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->masterPalette = static_cast<RGBA_t *>(memcpy(malloc(palsize), pMasterPalette, 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 = {pMasterPalette, pGammaCorrectedPalette};
R_GenerateBlendTables_Core(&state);
#endif
}
static void R_GenerateBlendTables_Core(struct GenerateBlendTables_State *state)
{
INT32 i;
for (i = 0; i <= 9; i++)
{
const size_t offs = (0x10000 * i);
const UINT8 alpha = (TRANSTAB_AMTMUL10 * ((float)(10-i)));
R_GenerateTranslucencyTable(blendtables[blendtab_add] + offs, state->gammaCorrectedPalette, AST_ADD, alpha);
R_GenerateTranslucencyTable(blendtables[blendtab_subtract] + offs, state->masterPalette, AST_SUBTRACT, alpha); // intentionally uses pMasterPalette
R_GenerateTranslucencyTable(blendtables[blendtab_reversesubtract] + offs, state->gammaCorrectedPalette, AST_REVERSESUBTRACT, alpha);
}
R_GenerateTranslucencyTable(blendtables[blendtab_modulate], state->gammaCorrectedPalette, AST_MODULATE, 0);
}
void R_GenerateTranslucencyTable(UINT8 *table, RGBA_t* sourcepal, int style, UINT8 blendamt)
{
INT16 bg, fg;
RGBA_t backrgba, frontrgba, result;
if (table == NULL)
I_Error("R_GenerateTranslucencyTable: input table was NULL!");
for (bg = 0; bg <= 0xFF; bg++)
{
backrgba = sourcepal[bg];
for (fg = 0; fg <= 0xFF; fg++)
{
frontrgba = sourcepal[fg];
result.rgba = ASTBlendPixel(backrgba, frontrgba, style, blendamt);
table[((fg * 0x100) + bg)] = NearestPaletteColor(result.s.red, result.s.green, result.s.blue, sourcepal);
}
}
}
#define ClipTransLevel(trans) std::clamp<INT32>(trans, 0, NUMTRANSMAPS-2)
UINT8 *R_GetTranslucencyTable(INT32 alphalevel)
{
return transtables + (ClipTransLevel(alphalevel-1) << FF_TRANSSHIFT);
}
UINT8 *R_GetBlendTable(int style, INT32 alphalevel)
{
size_t offs = (ClipTransLevel(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 NULL;
return transtables + (ClipTransLevel(alphalevel) << FF_TRANSSHIFT);
}
void R_InitPaletteRemap(void)
{
palremap = static_cast<UINT8 *>(Z_Malloc(256, PU_STATIC, NULL));
invremap = static_cast<UINT8 *>(Z_Malloc(256, PU_STATIC, NULL));
W_ReadLump(W_GetNumForName("PALREMAP"), palremap);
W_ReadLump(W_GetNumForName("INVREMAP"), invremap);
}
UINT8 R_GetPaletteRemap(UINT8 color)
{
return palremap[color];
}
UINT8 R_InvPaletteRemap(UINT8 color)
{
return invremap[color];
}
/** \brief Retrieves a translation colormap from the cache.
\param skinnum number of skin, TC_DEFAULT or TC_BOSS
\param color translation color
\param flags set GTC_CACHE to use the cache
\return Colormap. If not cached, caller should Z_Free.
*/
UINT8* R_GetTranslationColormap(INT32 skinnum, skincolornum_t color, UINT8 flags)
{
UINT8* ret;
INT32 skintableindex = SkinToCacheIndex(skinnum); // Adjust if we want the default colormap
INT32 i;
if (flags & GTC_CACHE)
{
// Allocate table for skin if necessary
if (!translationtablecache[skintableindex])
translationtablecache[skintableindex] = static_cast<UINT8 **>(Z_Calloc(MAXSKINCOLORS * sizeof(UINT8**), PU_STATIC, NULL));
// Get colormap
ret = translationtablecache[skintableindex][color];
// Rebuild the cache if necessary
if (skincolor_modified[color])
{
for (i = 0; i < (INT32)(sizeof(translationtablecache) / sizeof(translationtablecache[0])); i++)
if (translationtablecache[i] && translationtablecache[i][color])
K_GenerateKartColormap(translationtablecache[i][color], CacheIndexToSkin(i), color);
skincolor_modified[color] = false;
}
}
else ret = NULL;
// Generate the colormap if necessary
if (!ret)
{
ret = static_cast<UINT8 *>(Z_MallocAlign(NUM_PALETTE_ENTRIES, (flags & GTC_CACHE) ? PU_LEVEL : PU_STATIC, NULL, 8));
K_GenerateKartColormap(ret, skinnum, color); //R_GenerateTranslationColormap(ret, skinnum, color); // SRB2kart
// Cache the colormap if desired
if (flags & GTC_CACHE)
translationtablecache[skintableindex][color] = ret;
}
return ret;
}
/** \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
*/
void R_FlushTranslationColormapCache(void)
{
INT32 i;
for (i = 0; i < (INT32)(sizeof(translationtablecache) / sizeof(translationtablecache[0])); i++)
if (translationtablecache[i])
memset(translationtablecache[i], 0, MAXSKINCOLORS * sizeof(UINT8**));
}
UINT16 R_GetColorByName(const char *name)
{
UINT16 color = (UINT16)atoi(name);
if (color > 0 && color < numskincolors)
return color;
for (color = 1; color < numskincolors; color++)
if (!stricmp(skincolors[color].name, name))
return color;
return SKINCOLOR_NONE;
}
UINT16 R_GetSuperColorByName(const char *name)
{
UINT16 i, color = SKINCOLOR_NONE;
char *realname = static_cast<char *>(Z_Malloc(MAXCOLORNAME+1, PU_STATIC, NULL));
snprintf(realname, MAXCOLORNAME+1, "Super %s 1", name);
for (i = 1; i < numskincolors; i++)
if (!stricmp(skincolors[i].name, realname)) {
color = i;
break;
}
Z_Free(realname);
return color;
}
// ==========================================================================
// 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 columncontext_e columncontext = COLUMNCONTEXT_DIRECT;
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.
\return void
*/
void R_InitViewBuffer(void)
{
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 MAIN DRAWERS CODE HERE
// ==========================================================================
#include "r_draw_column.cpp"
#include "r_draw_span.cpp"