ShuriZma
/
project64
mirror of https://github.com/project64/project64.git
1
0
Fork 0
Code Issues Packages Projects Releases Wiki Activity

GlideHQ: Add

This commit is contained in:
zilmar 2013-04-17 20:30:38 +10:00
parent 9e9fdc25bc
commit e928a5995d
42 changed files with 22953 additions and 0 deletions
Show Stats Download Patch File Download Diff File Expand all files Collapse all files

27
Source/GlideHQ/Ext_TxFilter.cpp Normal file
View File

@ -0,0 +1,27 @@
/*
* Texture Filtering
* Version: 1.0
*
* Copyright (C) 2007 Hiroshi Morii All Rights Reserved.
* Email koolsmoky(at)users.sourceforge.net
* Web http://www.3dfxzone.it/koolsmoky
*
* this 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, or (at your option)
* any later version.
*
* this 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 GNU Make; see the file COPYING. If not, write to
* the Free Software Foundation, 675 Mass Ave, Cambridge, MA 02139, USA.
*/
#include <memory.h>
#include <stdlib.h>
#include "Ext_TxFilter.h"

212
Source/GlideHQ/Ext_TxFilter.h Normal file
View File

@ -0,0 +1,212 @@
/*
* Texture Filtering
* Version: 1.0
*
* Copyright (C) 2007 Hiroshi Morii All Rights Reserved.
* Email koolsmoky(at)users.sourceforge.net
* Web http://www.3dfxzone.it/koolsmoky
*
* this 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, or (at your option)
* any later version.
*
* this 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 GNU Make; see the file COPYING. If not, write to
* the Free Software Foundation, 675 Mass Ave, Cambridge, MA 02139, USA.
*/
#ifndef __EXT_TXFILTER_H__
#define __EXT_TXFILTER_H__
#ifdef WIN32
#include <windows.h>
#define TXHMODULE HMODULE
#define DLOPEN(a) LoadLibraryW(a)
#define DLCLOSE(a) FreeLibrary(a)
#define DLSYM(a, b) GetProcAddress(a, b)
#define GETCWD(a, b) GetCurrentDirectoryW(a, b)
#define CHDIR(a) SetCurrentDirectoryW(a)
#else
#include <iostream>
#include <dlfcn.h>
#define MAX_PATH 4095
#define TXHMODULE void*
#define DLOPEN(a) dlopen(a, RTLD_LAZY|RTLD_GLOBAL)
#define DLCLOSE(a) dlclose(a)
#define DLSYM(a, b) dlsym(a, b)
#define GETCWD(a, b) getcwd(b, a)
#define CHDIR(a) chdir(a)
#endif
#ifdef WIN32
typedef __int64 int64;
typedef unsigned __int64 uint64;
typedef unsigned char boolean;
#else
typedef long long int64;
typedef unsigned long long uint64;
typedef unsigned char boolean;
#endif
#define NO_OPTIONS 0x00000000
#define FILTER_MASK 0x000000ff
#define NO_FILTER 0x00000000
#define SMOOTH_FILTER_MASK 0x0000000f
#define NO_SMOOTH_FILTER 0x00000000
#define SMOOTH_FILTER_1 0x00000001
#define SMOOTH_FILTER_2 0x00000002
#define SMOOTH_FILTER_3 0x00000003
#define SMOOTH_FILTER_4 0x00000004
#define SHARP_FILTER_MASK 0x000000f0
#define NO_SHARP_FILTER 0x00000000
#define SHARP_FILTER_1 0x00000010
#define SHARP_FILTER_2 0x00000020
#define ENHANCEMENT_MASK 0x00000f00
#define NO_ENHANCEMENT 0x00000000
#define X2_ENHANCEMENT 0x00000100
#define X2SAI_ENHANCEMENT 0x00000200
#define HQ2X_ENHANCEMENT 0x00000300
#define LQ2X_ENHANCEMENT 0x00000400
#define HQ4X_ENHANCEMENT 0x00000500
#define HQ2XS_ENHANCEMENT 0x00000600
#define LQ2XS_ENHANCEMENT 0x00000700
#define COMPRESSION_MASK 0x0000f000
#define NO_COMPRESSION 0x00000000
#define FXT1_COMPRESSION 0x00001000
#define NCC_COMPRESSION 0x00002000
#define S3TC_COMPRESSION 0x00003000
#define HIRESTEXTURES_MASK 0x000f0000
#define NO_HIRESTEXTURES 0x00000000
#define GHQ_HIRESTEXTURES 0x00010000
#define RICE_HIRESTEXTURES 0x00020000
#define JABO_HIRESTEXTURES 0x00030000
#define COMPRESS_TEX 0x00100000
#define COMPRESS_HIRESTEX 0x00200000
#define GZ_TEXCACHE 0x00400000
#define GZ_HIRESTEXCACHE 0x00800000
#define DUMP_TEXCACHE 0x01000000
#define DUMP_HIRESTEXCACHE 0x02000000
#define TILE_HIRESTEX 0x04000000
#define UNDEFINED_0 0x08000000
#define FORCE16BPP_HIRESTEX 0x10000000
#define FORCE16BPP_TEX 0x20000000
#define LET_TEXARTISTS_FLY 0x40000000 /* a little freedom for texture artists */
#define DUMP_TEX 0x80000000
#ifndef __GLIDE_H__ /* GLIDE3 */
/* from 3Dfx Interactive Inc. glide.h */
#define GR_TEXFMT_ALPHA_8 0x2
#define GR_TEXFMT_INTENSITY_8 0x3
#define GR_TEXFMT_ALPHA_INTENSITY_44 0x4
#define GR_TEXFMT_P_8 0x5
#define GR_TEXFMT_RGB_565 0xa
#define GR_TEXFMT_ARGB_1555 0xb
#define GR_TEXFMT_ARGB_4444 0xc
#define GR_TEXFMT_ALPHA_INTENSITY_88 0xd
/* from 3Dfx Interactive Inc. g3ext.h */
#define GR_TEXFMT_ARGB_CMP_FXT1 0x11
#define GR_TEXFMT_ARGB_8888 0x12
#define GR_TEXFMT_ARGB_CMP_DXT1 0x16
#define GR_TEXFMT_ARGB_CMP_DXT3 0x18
#define GR_TEXFMT_ARGB_CMP_DXT5 0x1A
#endif /* GLIDE3 */
struct GHQTexInfo {
unsigned char *data;
int width;
int height;
unsigned short format;
int smallLodLog2;
int largeLodLog2;
int aspectRatioLog2;
int tiles;
int untiled_width;
int untiled_height;
unsigned char is_hires_tex;
};
/* Callback to display hires texture info.
* Gonetz <gonetz(at)ngs.ru>
*
* void DispInfo(const char *format, ...)
* {
* va_list args;
* char buf[INFO_BUF];
*
* va_start(args, format);
* vsprintf(buf, format, args);
* va_end(args);
*
* printf(buf);
* }
*/
#define INFO_BUF 4095
typedef void (*dispInfoFuncExt)(const wchar_t *format, ...);
#ifndef TXFILTER_DLL
boolean ext_ghq_init(int maxwidth, /* maximum texture width supported by hardware */
int maxheight,/* maximum texture height supported by hardware */
int maxbpp, /* maximum texture bpp supported by hardware */
int options, /* options */
int cachesize,/* cache textures to system memory */
wchar_t *path, /* plugin directory. must be smaller than MAX_PATH */
wchar_t *ident, /* name of ROM. must be no longer than 64 in character. */
dispInfoFuncExt callback /* callback function to display info */
);
void ext_ghq_shutdown(void);
boolean ext_ghq_txfilter(unsigned char *src, /* input texture */
int srcwidth, /* width of input texture */
int srcheight, /* height of input texture */
unsigned short srcformat, /* format of input texture */
uint64 g64crc, /* glide64 crc */
GHQTexInfo *info /* output */
);
boolean ext_ghq_hirestex(uint64 g64crc, /* glide64 crc */
uint64 r_crc64, /* checksum hi:palette low:texture */
unsigned short *palette, /* palette for CI textures */
GHQTexInfo *info /* output */
);
uint64 ext_ghq_checksum(unsigned char *src, /* input texture */
int width, /* width of texture */
int height, /* height of texture */
int size, /* type of texture pixel */
int rowStride, /* row stride in bytes */
unsigned char *palette /* palette */
);
boolean ext_ghq_dmptx(unsigned char *src, /* input texture (must be in 3Dfx Glide format) */
int width, /* width of texture */
int height, /* height of texture */
int rowStridePixel, /* row stride of input texture in pixels */
unsigned short gfmt, /* glide format of input texture */
unsigned short n64fmt,/* N64 format hi:format low:size */
uint64 r_crc64 /* checksum hi:palette low:texture */
);
boolean ext_ghq_reloadhirestex();
#endif /* TXFILTER_DLL */
#endif /* __EXT_TXFILTER_H__ */

295
Source/GlideHQ/GlideHQ.vcproj Normal file
View File

@ -0,0 +1,295 @@
<?xml version="1.0" encoding="shift_jis"?>
<VisualStudioProject
ProjectType="Visual C++"
Version="9.00"
Name="GlideHQ"
ProjectGUID="{2DADDAA5-0F57-46ED-A974-747908DDC7F3}"
RootNamespace="GlideHQ"
TargetFrameworkVersion="0"
>
<Platforms>
<Platform
Name="Win32"
/>
</Platforms>
<ToolFiles>
</ToolFiles>
<Configurations>
<Configuration
Name="Debug|Win32"
ConfigurationType="4"
InheritedPropertySheets="$(SolutionDir)PropertySheets/Win32.Debug.vsprops"
>
<Tool
Name="VCPreBuildEventTool"
/>
<Tool
Name="VCCustomBuildTool"
/>
<Tool
Name="VCXMLDataGeneratorTool"
/>
<Tool
Name="VCWebServiceProxyGeneratorTool"
/>
<Tool
Name="VCMIDLTool"
/>
<Tool
Name="VCCLCompilerTool"
AdditionalIncludeDirectories="&quot;$(Root)Source/3rd Party/wx/src&quot;;inc"
UsePrecompiledHeader="0"
/>
<Tool
Name="VCManagedResourceCompilerTool"
/>
<Tool
Name="VCResourceCompilerTool"
/>
<Tool
Name="VCPreLinkEventTool"
/>
<Tool
Name="VCLibrarianTool"
/>
<Tool
Name="VCALinkTool"
/>
<Tool
Name="VCXDCMakeTool"
/>
<Tool
Name="VCBscMakeTool"
/>
<Tool
Name="VCFxCopTool"
/>
<Tool
Name="VCPostBuildEventTool"
/>
</Configuration>
<Configuration
Name="Release|Win32"
ConfigurationType="4"
InheritedPropertySheets="$(SolutionDir)PropertySheets/Win32.Release.vsprops"
>
<Tool
Name="VCPreBuildEventTool"
/>
<Tool
Name="VCCustomBuildTool"
/>
<Tool
Name="VCXMLDataGeneratorTool"
/>
<Tool
Name="VCWebServiceProxyGeneratorTool"
/>
<Tool
Name="VCMIDLTool"
/>
<Tool
Name="VCCLCompilerTool"
AdditionalIncludeDirectories="inc"
UsePrecompiledHeader="0"
/>
<Tool
Name="VCManagedResourceCompilerTool"
/>
<Tool
Name="VCResourceCompilerTool"
/>
<Tool
Name="VCPreLinkEventTool"
/>
<Tool
Name="VCLibrarianTool"
/>
<Tool
Name="VCALinkTool"
/>
<Tool
Name="VCXDCMakeTool"
/>
<Tool
Name="VCBscMakeTool"
/>
<Tool
Name="VCFxCopTool"
/>
<Tool
Name="VCPostBuildEventTool"
/>
</Configuration>
</Configurations>
<References>
</References>
<Files>
<Filter
Name="Source Files"
Filter="cpp;c;cc;cxx;def;odl;idl;hpj;bat;asm;asmx"
UniqueIdentifier="{4FC737F1-C7A5-4376-A066-2A32D752A2FF}"
>
<File
RelativePath="Ext_TxFilter.cpp"
>
</File>
<File
RelativePath="TextureFilters.cpp"
>
</File>
<File
RelativePath="TextureFilters_2xsai.cpp"
>
</File>
<File
RelativePath="TextureFilters_hq2x.cpp"
>
</File>
<File
RelativePath="TextureFilters_hq4x.cpp"
>
</File>
<File
RelativePath="TxCache.cpp"
>
</File>
<File
RelativePath="TxDbg.cpp"
>
</File>
<File
RelativePath="TxFilter.cpp"
>
</File>
<File
RelativePath="TxFilterExport.cpp"
>
</File>
<File
RelativePath="TxHiResCache.cpp"
>
</File>
<File
RelativePath="TxImage.cpp"
>
</File>
<File
RelativePath="TxQuantize.cpp"
>
</File>
<File
RelativePath="TxReSample.cpp"
>
</File>
<File
RelativePath="TxTexCache.cpp"
>
</File>
<File
RelativePath="TxUtil.cpp"
>
</File>
</Filter>
<Filter
Name="Header Files"
Filter="h;hpp;hxx;hm;inl;inc;xsd"
UniqueIdentifier="{93995380-89BD-4b04-88EB-625FBE52EBFB}"
>
<File
RelativePath="Ext_TxFilter.h"
>
</File>
<File
RelativePath="Internal.h"
>
</File>
<File
RelativePath="TextureFilters.h"
>
</File>
<File
RelativePath="TextureFilters_2xsai.h"
>
</File>
<File
RelativePath="TextureFilters_hq2x.h"
>
</File>
<File
RelativePath="TextureFilters_hq4x.h"
>
</File>
<File
RelativePath="TextureFilters_lq2x.h"
>
</File>
<File
RelativePath="TxCache.h"
>
</File>
<File
RelativePath="TxDbg.h"
>
</File>
<File
RelativePath="TxFilter.h"
>
</File>
<File
RelativePath="TxHiResCache.h"
>
</File>
<File
RelativePath="TxImage.h"
>
</File>
<File
RelativePath="TxQuantize.h"
>
</File>
<File
RelativePath="TxTexCache.h"
>
</File>
<File
RelativePath="TxUtil.h"
>
</File>
</Filter>
<Filter
Name="tc-1.1+"
>
<File
RelativePath=".\tc-1.1+\dxtn.c"
>
</File>
<File
RelativePath=".\tc-1.1+\fxt1.c"
>
</File>
<File
RelativePath=".\tc-1.1+\fxt1.h"
>
</File>
<File
RelativePath=".\tc-1.1+\internal.h"
>
</File>
<File
RelativePath=".\tc-1.1+\texstore.c"
>
</File>
<File
RelativePath=".\tc-1.1+\types.h"
>
</File>
<File
RelativePath=".\tc-1.1+\wrapper.c"
>
</File>
</Filter>
</Files>
<Globals>
</Globals>
</VisualStudioProject>

94
Source/GlideHQ/README.txt Normal file
View File

@ -0,0 +1,94 @@
/*
* GlideHQ (Texture enhancer library for Glide64)
* Version: 1.5
*
* Copyright (C) 2007 Hiroshi Morii aka KoolSmoky All Rights Reserved.
* Email koolsmoky(at)users.sourceforge.net
* Web http://www.3dfxzone.it/koolsmoky
*
* this 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, or (at your option)
* any later version.
*
* this 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 GNU Make; see the file COPYING. If not, write to
* the Free Software Foundation, 675 Mass Ave, Cambridge, MA 02139, USA.
*/
About:
This is a realtime texture enhancer library with hi-resolution texture
pack support for Glide64 (http://glide64.emuxhaven.net). Traditional and
non-traditional techniques have been used to achieve speed and high image
quality even on a 9 year old 3Dfx Voodoo2.
Although the 3Dfx Glide3x texture format naming conventions are used, the
library can be expanded for generic use.
Supported:
OS: 32bit Linux and MS Windows
Enhancers: Hq4x, Hq2x, Hq2xS, Lq2x, Lq2xS, Super2xSai, x2
Filters: Smooth (1,2,3,4), Sharp (1,2)
Compressors: FXT1, S3TC
Input formats: GR_TEXFMT_ALPHA_8,
GR_TEXFMT_RGB_565,
GR_TEXFMT_ARGB_1555,
GR_TEXFMT_ARGB_4444,
GR_TEXFMT_ARGB_8888,
GR_TEXFMT_ALPHA_INTENSITY_44,
GR_TEXFMT_ALPHA_INTENSITY_88
Output formats: Same as input unless compression or hires packs are used.
Hires texture packs: Rice format (Jabo and GlideHQ format coming later)
Acknowledgments:
I hope you enjoy GlideHQ (texture enhancer library for Glide64). Greatest
thanks to Gonetz for making this happen in his busy time. We've rushed
everything to share the eye-candy with all of you N64 emulation fans. I
would also like to thank a great friend of mine, Daniel Borca for providing
the texture compression code, Maxim Stepin (hq2x 4x), and Derek Liauw Kie Fa
(2xSaI) for the filtering engines, Rice for his N64 graphics plugin source
code, and Mudlord for the hq2xS lq2xS code. GlideHQ also uses the boost C++
libraries, zlib general purpose compression library, and the Portable Network
Graphics library. Thanks to all the developers for making them available. And
special thanks to the Glide64 beta testing crew. Without their feedbacks
this library would not have seen daylight. Thank you all.
The source code for GlideHQ is released in hopes that it will be improved.
I know the coding is not on par after so much late night caffeine boosts.
If you have suggestions or modifications, please feel free to post them on
the Glide64 forum at emuxhaven.
Porting the library to other platforms should not be so hard. The coding is
done with cross platform compatibility in mind and will build with GCC and
GNU make. Currently supported are 32bit Linux and MS Windows.
If you are looking for driver updates for your 3Dfx Interactive Inc. gfx
card, grab them from the forums at http://www.3dfxzone.it/enboard/
Unbelievable as it seems, drivers are still being updated after 6 years
from 3Dfx's demise.
I know N64 rules, anyone up for PSX? :))
-KoolSmoky
References:
[1] R.W. Floyd & L. Steinberg, An adaptive algorithm for spatial grey scale,
Proceedings of the Society of Information Display 17, pp75-77, 1976
[2] Ken Turkowski, Filters for Common Resampling Tasks, Apple Computer 1990
http://www.worldserver.com/turk/computergraphics/ResamplingFilters.pdf
[3] Don P. Mitchell and Arun N. Netravali, Reconstruction Filters in Computer
Graphics, SIGGRAPH '88, Proceedings of the 15th annual conference on
Computer graphics and interactive techniques, pp221-228, 1988
[4] J. F. Kaiser and W. A. Reed, Data smoothing using low-pass digital
filters, Rev. Sci. instrum. 48 (11), pp1447-1457, 1977
[5] Maxim Stepin, hq4x Magnification Filter, http://www.hiend3d.com/hq4x.html
[6] Derek Liauw Kie Fa, 2xSaI, http://elektron.its.tudelft.nl/~dalikifa
[7] Dirk Stevens, Eagle engine http://www.retrofx.com/rfxtech.html
[8] 3DFX_texture_compression_FXT1 and EXT_texture_compression_s3tc extension
specs from the OpenGL Extension Registry. http://oss.sgi.com/projects/
ogl-sample/registry/

715
Source/GlideHQ/TextureFilters.cpp Normal file
View File

@ -0,0 +1,715 @@
/*
Copyright (C) 2003 Rice1964
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., 59 Temple Place - Suite 330, Boston, MA 02111-1307, USA.
*/
/* Copyright (C) 2007 Hiroshi Morii <koolsmoky(at)users.sourceforge.net>
* Modified for the Texture Filtering library
*/
#include <string.h>
#include "TextureFilters.h"
/************************************************************************/
/* 2X filters */
/************************************************************************/
#define DWORD_MAKE(r, g, b, a) ((uint32) (((a) << 24) | ((r) << 16) | ((g) << 8) | (b)))
#define WORD_MAKE(r, g, b, a) ((uint16) (((a) << 12) | ((r) << 8) | ((g) << 4) | (b)))
// Basic 2x R8G8B8A8 filter with interpolation
void Texture2x_32(uint8 *srcPtr, uint32 srcPitch, uint8 *dstPtr, uint32 dstPitch, int width, int height)
{
uint32 *pDst1, *pDst2;
uint32 *pSrc, *pSrc2;
uint32 nWidth = width;
uint32 nHeight = height;
uint32 b1;
uint32 g1;
uint32 r1;
uint32 a1;
uint32 b2;
uint32 g2;
uint32 r2;
uint32 a2;
uint32 b3;
uint32 g3;
uint32 r3;
uint32 a3;
uint32 b4;
uint32 g4;
uint32 r4;
uint32 a4;
uint32 xSrc;
uint32 ySrc;
for (ySrc = 0; ySrc < nHeight; ySrc++)
{
pSrc = (uint32*)(((uint8*)srcPtr)+ySrc*srcPitch);
pSrc2 = (uint32*)(((uint8*)srcPtr)+(ySrc+1)*srcPitch);
pDst1 = (uint32*)(((uint8*)dstPtr)+(ySrc*2)*dstPitch);
pDst2 = (uint32*)(((uint8*)dstPtr)+(ySrc*2+1)*dstPitch);
for (xSrc = 0; xSrc < nWidth; xSrc++)
{
b1 = (pSrc[xSrc]>>0)&0xFF;
g1 = (pSrc[xSrc]>>8)&0xFF;
r1 = (pSrc[xSrc]>>16)&0xFF;
a1 = (pSrc[xSrc]>>24)&0xFF;
if( xSrc<nWidth-1 )
{
b2 = (pSrc[xSrc+1]>>0)&0xFF;
g2 = (pSrc[xSrc+1]>>8)&0xFF;
r2 = (pSrc[xSrc+1]>>16)&0xFF;
a2 = (pSrc[xSrc+1]>>24)&0xFF;
}
if( ySrc<nHeight-1 )
{
b3 = (pSrc2[xSrc]>>0)&0xFF;
g3 = (pSrc2[xSrc]>>8)&0xFF;
r3 = (pSrc2[xSrc]>>16)&0xFF;
a3 = (pSrc2[xSrc]>>24)&0xFF;
if( xSrc<nWidth-1 )
{
b4 = (pSrc2[xSrc+1]>>0)&0xFF;
g4 = (pSrc2[xSrc+1]>>8)&0xFF;
r4 = (pSrc2[xSrc+1]>>16)&0xFF;
a4 = (pSrc2[xSrc+1]>>24)&0xFF;
}
}
// Pixel 1
pDst1[xSrc*2] = pSrc[xSrc];
// Pixel 2
if( xSrc<nWidth-1 )
{
pDst1[xSrc*2+1] = DWORD_MAKE((r1+r2)/2, (g1+g2)/2, (b1+b2)/2, (a1+a2)/2);
}
else
pDst1[xSrc*2+1] = pSrc[xSrc];
// Pixel 3
if( ySrc<nHeight-1 )
{
pDst2[xSrc*2] = DWORD_MAKE((r1+r3)/2, (g1+g3)/2, (b1+b3)/2, (a1+a3)/2);
}
else
pDst2[xSrc*2] = pSrc[xSrc];
// Pixel 4
if( xSrc<nWidth-1 )
{
if( ySrc<nHeight-1 )
{
pDst2[xSrc*2+1] = DWORD_MAKE((r1+r2+r3+r4)/4, (g1+g2+g3+g4)/4, (b1+b2+b3+b4)/4, (a1+a2+a3+a4)/4);
}
else
{
pDst2[xSrc*2+1] = DWORD_MAKE((r1+r2)/2, (g1+g2)/2, (b1+b2)/2, (a1+a2)/2);
}
}
else
{
if( ySrc<nHeight-1 )
{
pDst2[xSrc*2+1] = DWORD_MAKE((r1+r3)/2, (g1+g3)/2, (b1+b3)/2, (a1+a3)/2);
}
else
pDst2[xSrc*2+1] = pSrc[xSrc];
}
}
}
}
#if !_16BPP_HACK
// Basic 2x R4G4B4A4 filter with interpolation
void Texture2x_16(uint8 *srcPtr, uint32 srcPitch, uint8 *dstPtr, uint32 dstPitch, int width, int height)
{
uint16 *pDst1, *pDst2;
uint16 *pSrc, *pSrc2;
uint32 nWidth = width;
uint32 nHeight = height;
uint16 b1;
uint16 g1;
uint16 r1;
uint16 a1;
uint16 b2;
uint16 g2;
uint16 r2;
uint16 a2;
uint16 b3;
uint16 g3;
uint16 r3;
uint16 a3;
uint16 b4;
uint16 g4;
uint16 r4;
uint16 a4;
uint16 xSrc;
uint16 ySrc;
for (ySrc = 0; ySrc < nHeight; ySrc++)
{
pSrc = (uint16*)(((uint8*)srcPtr)+ySrc*srcPitch);
pSrc2 = (uint16*)(((uint8*)srcPtr)+(ySrc+1)*srcPitch);
pDst1 = (uint16*)(((uint8*)dstPtr)+(ySrc*2)*dstPitch);
pDst2 = (uint16*)(((uint8*)dstPtr)+(ySrc*2+1)*dstPitch);
for (xSrc = 0; xSrc < nWidth; xSrc++)
{
b1 = (pSrc[xSrc]>> 0)&0xF;
g1 = (pSrc[xSrc]>> 4)&0xF;
r1 = (pSrc[xSrc]>> 8)&0xF;
a1 = (pSrc[xSrc]>>12)&0xF;
if( xSrc<nWidth-1 )
{
b2 = (pSrc[xSrc+1]>> 0)&0xF;
g2 = (pSrc[xSrc+1]>> 4)&0xF;
r2 = (pSrc[xSrc+1]>> 8)&0xF;
a2 = (pSrc[xSrc+1]>>12)&0xF;
}
if( ySrc<nHeight-1 )
{
b3 = (pSrc2[xSrc]>> 0)&0xF;
g3 = (pSrc2[xSrc]>> 4)&0xF;
r3 = (pSrc2[xSrc]>> 8)&0xF;
a3 = (pSrc2[xSrc]>>12)&0xF;
if( xSrc<nWidth-1 )
{
b4 = (pSrc2[xSrc+1]>> 0)&0xF;
g4 = (pSrc2[xSrc+1]>> 4)&0xF;
r4 = (pSrc2[xSrc+1]>> 8)&0xF;
a4 = (pSrc2[xSrc+1]>>12)&0xF;
}
}
// Pixel 1
pDst1[xSrc*2] = pSrc[xSrc];
// Pixel 2
if( xSrc<nWidth-1 )
{
pDst1[xSrc*2+1] = WORD_MAKE((r1+r2)/2, (g1+g2)/2, (b1+b2)/2, (a1+a2)/2);
}
else
pDst1[xSrc*2+1] = pSrc[xSrc];
// Pixel 3
if( ySrc<nHeight-1 )
{
pDst2[xSrc*2] = WORD_MAKE((r1+r3)/2, (g1+g3)/2, (b1+b3)/2, (a1+a3)/2);
}
else
pDst2[xSrc*2] = pSrc[xSrc];
// Pixel 4
if( xSrc<nWidth-1 )
{
if( ySrc<nHeight-1 )
{
pDst2[xSrc*2+1] = WORD_MAKE((r1+r2+r3+r4)/4, (g1+g2+g3+g4)/4, (b1+b2+b3+b4)/4, (a1+a2+a3+a4)/4);
}
else
{
pDst2[xSrc*2+1] = WORD_MAKE((r1+r2)/2, (g1+g2)/2, (b1+b2)/2, (a1+a2)/2);
}
}
else
{
if( ySrc<nHeight-1 )
{
pDst2[xSrc*2+1] = WORD_MAKE((r1+r3)/2, (g1+g3)/2, (b1+b3)/2, (a1+a3)/2);
}
else
pDst2[xSrc*2+1] = pSrc[xSrc];
}
}
}
}
#endif /* !_16BPP_HACK */
/*
* Sharp filters
* Hiroshi Morii <koolsmoky@users.sourceforge.net>
*/
void SharpFilter_8888(uint32 *src, uint32 srcwidth, uint32 srcheight, uint32 *dest, uint32 filter)
{
// NOTE: for now we get away with copying the boundaries
// filter the boundaries if we face problems
uint32 mul1, mul2, mul3, shift4;
uint32 x,y,z;
uint32 *_src1, *_src2, *_src3, *_dest;
uint32 val[4];
uint32 t1,t2,t3,t4,t5,t6,t7,t8,t9;
switch( filter )
{
case SHARP_FILTER_2:
mul1=1;
mul2=8;
mul3=12;
shift4=2;
break;
case SHARP_FILTER_1:
default:
mul1=1;
mul2=8;
mul3=16;
shift4=3;
break;
}
// setup rows
_src1 = src;
_src2 = _src1 + srcwidth;
_src3 = _src2 + srcwidth;
_dest = dest;
// copy the first row
memcpy(_dest, _src1, (srcwidth << 2));
_dest += srcwidth;
// filter 2nd row to 1 row before the last
for (y = 1; y < srcheight-1; y++) {
// copy the first pixel
_dest[0] = *_src2;
// filter 2nd pixel to 1 pixel before last
for (x = 1; x < srcwidth-1; x++) {
for (z=0; z<4; z++) {
t1 = *((uint8*)(_src1+x-1)+z);
t2 = *((uint8*)(_src1+x )+z);
t3 = *((uint8*)(_src1+x+1)+z);
t4 = *((uint8*)(_src2+x-1)+z);
t5 = *((uint8*)(_src2+x )+z);
t6 = *((uint8*)(_src2+x+1)+z);
t7 = *((uint8*)(_src3+x-1)+z);
t8 = *((uint8*)(_src3+x )+z);
t9 = *((uint8*)(_src3+x+1)+z);
if( (t5*mul2) > (t1+t3+t7+t9+t2+t4+t6+t8)*mul1 ) {
val[z]= ((t5*mul3) - (t1+t3+t7+t9+t2+t4+t6+t8)*mul1)>>shift4;
if (val[z] > 0xFF) val[z] = 0xFF;
} else {
val[z] = t5;
}
}
_dest[x] = val[0]|(val[1]<<8)|(val[2]<<16)|(val[3]<<24);
}
// copy the ending pixel
_dest[srcwidth-1] = *(_src3 - 1);
// next row
_src1 += srcwidth;
_src2 += srcwidth;
_src3 += srcwidth;
_dest += srcwidth;
}
// copy the last row
memcpy(_dest, _src2, (srcwidth << 2));
}
#if !_16BPP_HACK
void SharpFilter_4444(uint16 *src, uint32 srcwidth, uint32 srcheight, uint16 *dest, uint32 filter)
{
// NOTE: for now we get away with copying the boundaries
// filter the boundaries if we face problems
uint16 mul1, mul2, mul3, shift4;
uint32 x,y,z;
uint16 *_src1, *_src2, *_src3, *_dest;
uint16 val[4];
uint16 t1,t2,t3,t4,t5,t6,t7,t8,t9;
switch( filter ) {
case SHARP_FILTER_2:
mul1=1;
mul2=8;
mul3=12;
shift4=2;
break;
case SHARP_FILTER_1:
default:
mul1=1;
mul2=8;
mul3=16;
shift4=3;
break;
}
// setup rows
_src1 = src;
_src2 = _src1 + srcwidth;
_src3 = _src2 + srcwidth;
_dest = dest;
// copy the first row
memcpy(_dest, _src1, (srcwidth << 1));
_dest += srcwidth;
// filter 2nd row to 1 row before the last
for( y = 1; y < srcheight - 1; y++) {
// copy the first pixel
_dest[0] = *_src2;
// filter 2nd pixel to 1 pixel before last
for( x = 1; x < srcwidth - 1; x++) {
for( z = 0; z < 4; z++ ) {
/* Hiroshi Morii <koolsmoky@users.sourceforge.net>
* Read the entire 16bit pixel and then extract the A,R,G,B components.
*/
uint32 shift = z << 2;
t1 = ((*((uint16*)(_src1+x-1))) >> shift) & 0xF;
t2 = ((*((uint16*)(_src1+x ))) >> shift) & 0xF;
t3 = ((*((uint16*)(_src1+x+1))) >> shift) & 0xF;
t4 = ((*((uint16*)(_src2+x-1))) >> shift) & 0xF;
t5 = ((*((uint16*)(_src2+x ))) >> shift) & 0xF;
t6 = ((*((uint16*)(_src2+x+1))) >> shift) & 0xF;
t7 = ((*((uint16*)(_src3+x-1))) >> shift) & 0xF;
t8 = ((*((uint16*)(_src3+x ))) >> shift) & 0xF;
t9 = ((*((uint16*)(_src3+x+1))) >> shift) & 0xF;
if( (t5*mul2) > (t1+t3+t7+t9+t2+t4+t6+t8)*mul1 ) {
val[z] = ((t5*mul3) - (t1+t3+t7+t9+t2+t4+t6+t8)*mul1)>>shift4;
if (val[z] > 0xF) val[z] = 0xF;
} else {
val[z] = t5;
}
}
_dest[x] = val[0]|(val[1]<<4)|(val[2]<<8)|(val[3]<<12);
}
// copy the ending pixel
_dest[srcwidth-1] = *(_src3 - 1);
// next row
_src1 += srcwidth;
_src2 += srcwidth;
_src3 += srcwidth;
_dest += srcwidth;
}
// copy the last row
memcpy(_dest, _src2, (srcwidth << 1));
}
#endif /* !_16BPP_HACK */
/*
* Smooth filters
* Hiroshi Morii <koolsmoky@users.sourceforge.net>
*/
void SmoothFilter_8888(uint32 *src, uint32 srcwidth, uint32 srcheight, uint32 *dest, uint32 filter)
{
// NOTE: for now we get away with copying the boundaries
// filter the boundaries if we face problems
uint32 mul1, mul2, mul3, shift4;
uint32 x,y,z;
uint32 *_src1, *_src2, *_src3, *_dest;
uint32 val[4];
uint32 t1,t2,t3,t4,t5,t6,t7,t8,t9;
switch( filter ) {
case SMOOTH_FILTER_4:
mul1=1;
mul2=2;
mul3=4;
shift4=4;
break;
case SMOOTH_FILTER_3:
mul1=1;
mul2=1;
mul3=8;
shift4=4;
break;
case SMOOTH_FILTER_2:
mul1=1;
mul2=1;
mul3=2;
shift4=2;
break;
case SMOOTH_FILTER_1:
default:
mul1=1;
mul2=1;
mul3=6;
shift4=3;
break;
}
switch (filter) {
case SMOOTH_FILTER_3:
case SMOOTH_FILTER_4:
// setup rows
_src1 = src;
_src2 = _src1 + srcwidth;
_src3 = _src2 + srcwidth;
_dest = dest;
// copy the first row
memcpy(_dest, _src1, (srcwidth << 2));
_dest += srcwidth;
// filter 2nd row to 1 row before the last
for (y = 1; y < srcheight - 1; y++){
// copy the first pixel
_dest[0] = _src2[0];
// filter 2nd pixel to 1 pixel before last
for (x = 1; x < srcwidth - 1; x++) {
for (z = 0; z < 4; z++ ) {
t1 = *((uint8*)(_src1+x-1)+z);
t2 = *((uint8*)(_src1+x )+z);
t3 = *((uint8*)(_src1+x+1)+z);
t4 = *((uint8*)(_src2+x-1)+z);
t5 = *((uint8*)(_src2+x )+z);
t6 = *((uint8*)(_src2+x+1)+z);
t7 = *((uint8*)(_src3+x-1)+z);
t8 = *((uint8*)(_src3+x )+z);
t9 = *((uint8*)(_src3+x+1)+z);
/* the component value must not overflow 0xFF */
val[z] = ((t1+t3+t7+t9)*mul1+((t2+t4+t6+t8)*mul2)+(t5*mul3))>>shift4;
if (val[z] > 0xFF) val[z] = 0xFF;
}
_dest[x] = val[0]|(val[1]<<8)|(val[2]<<16)|(val[3]<<24);
}
// copy the ending pixel
_dest[srcwidth-1] = *(_src3 - 1);
// next row
_src1 += srcwidth;
_src2 += srcwidth;
_src3 += srcwidth;
_dest += srcwidth;
}
// copy the last row
memcpy(_dest, _src2, (srcwidth << 2));
break;
case SMOOTH_FILTER_1:
case SMOOTH_FILTER_2:
default:
// setup rows
_src1 = src;
_src2 = _src1 + srcwidth;
_src3 = _src2 + srcwidth;
_dest = dest;
// copy the first row
memcpy(_dest, _src1, (srcwidth << 2));
_dest += srcwidth;
// filter 2nd row to 1 row before the last
for (y = 1; y < srcheight - 1; y++) {
// filter 1st pixel to the last
if (y & 1) {
for( x = 0; x < srcwidth; x++) {
for( z = 0; z < 4; z++ ) {
t2 = *((uint8*)(_src1+x )+z);
t5 = *((uint8*)(_src2+x )+z);
t8 = *((uint8*)(_src3+x )+z);
/* the component value must not overflow 0xFF */
val[z] = ((t2+t8)*mul2+(t5*mul3))>>shift4;
if (val[z] > 0xFF) val[z] = 0xFF;
}
_dest[x] = val[0]|(val[1]<<8)|(val[2]<<16)|(val[3]<<24);
}
} else {
memcpy(_dest, _src2, (srcwidth << 2));
}
// next row
_src1 += srcwidth;
_src2 += srcwidth;
_src3 += srcwidth;
_dest += srcwidth;
}
// copy the last row
memcpy(_dest, _src2, (srcwidth << 2));
break;
}
}
#if !_16BPP_HACK
void SmoothFilter_4444(uint16 *src, uint32 srcwidth, uint32 srcheight, uint16 *dest, uint32 filter)
{
// NOTE: for now we get away with copying the boundaries
// filter the boundaries if we face problems
uint16 mul1, mul2, mul3, shift4;
uint32 x,y,z;
uint16 *_src1, *_src2, *_src3, *_dest;
uint16 val[4];
uint16 t1,t2,t3,t4,t5,t6,t7,t8,t9;
switch( filter ) {
case SMOOTH_FILTER_4:
mul1=1;
mul2=2;
mul3=4;
shift4=4;
break;
case SMOOTH_FILTER_3:
mul1=1;
mul2=1;
mul3=8;
shift4=4;
break;
case SMOOTH_FILTER_2:
mul1=1;
mul2=1;
mul3=2;
shift4=2;
break;
case SMOOTH_FILTER_1:
default:
mul1=1;
mul2=1;
mul3=6;
shift4=3;
break;
}
switch (filter) {
case SMOOTH_FILTER_3:
case SMOOTH_FILTER_4:
// setup rows
_src1 = src;
_src2 = _src1 + srcwidth;
_src3 = _src2 + srcwidth;
_dest = dest;
// copy the first row
memcpy(_dest, _src1, (srcwidth << 1));
_dest += srcwidth;
// filter 2nd row to 1 row before the last
for (y = 1; y < srcheight - 1; y++) {
// copy the first pixel
_dest[0] = *_src2;
// filter 2nd pixel to 1 pixel before last
for (x = 1; x < srcwidth - 1; x++) {
for (z = 0; z < 4; z++ ) {
/* Read the entire 16bit pixel and then extract the A,R,G,B components. */
uint32 shift = z << 2;
t1 = ((*(uint16*)(_src1+x-1)) >> shift) & 0xF;
t2 = ((*(uint16*)(_src1+x )) >> shift) & 0xF;
t3 = ((*(uint16*)(_src1+x+1)) >> shift) & 0xF;
t4 = ((*(uint16*)(_src2+x-1)) >> shift) & 0xF;
t5 = ((*(uint16*)(_src2+x )) >> shift) & 0xF;
t6 = ((*(uint16*)(_src2+x+1)) >> shift) & 0xF;
t7 = ((*(uint16*)(_src3+x-1)) >> shift) & 0xF;
t8 = ((*(uint16*)(_src3+x )) >> shift) & 0xF;
t9 = ((*(uint16*)(_src3+x+1)) >> shift) & 0xF;
/* the component value must not overflow 0xF */
val[z] = ((t1+t3+t7+t9)*mul1+((t2+t4+t6+t8)*mul2)+(t5*mul3))>>shift4;
if (val[z] > 0xF) val[z] = 0xF;
}
_dest[x] = val[0]|(val[1]<<4)|(val[2]<<8)|(val[3]<<12);
}
// copy the ending pixel
_dest[srcwidth-1] = *(_src3 - 1);
// next row
_src1 += srcwidth;
_src2 += srcwidth;
_src3 += srcwidth;
_dest += srcwidth;
}
// copy the last row
memcpy(_dest, _src2, (srcwidth << 1));
break;
case SMOOTH_FILTER_1:
case SMOOTH_FILTER_2:
default:
// setup rows
_src1 = src;
_src2 = _src1 + srcwidth;
_src3 = _src2 + srcwidth;
_dest = dest;
// copy the first row
memcpy(_dest, _src1, (srcwidth << 1));
_dest += srcwidth;
// filter 2nd row to 1 row before the last
for( y = 1; y < srcheight - 1; y++) {
if (y & 1) {
for( x = 0; x < srcwidth; x++) {
for( z = 0; z < 4; z++ ) {
/* Read the entire 16bit pixel and then extract the A,R,G,B components. */
uint32 shift = z << 2;
t2 = ((*(uint16*)(_src1+x)) >> shift) & 0xF;
t5 = ((*(uint16*)(_src2+x)) >> shift) & 0xF;
t8 = ((*(uint16*)(_src3+x)) >> shift) & 0xF;
/* the component value must not overflow 0xF */
val[z] = ((t2+t8)*mul2+(t5*mul3))>>shift4;
if (val[z] > 0xF) val[z] = 0xF;
}
_dest[x] = val[0]|(val[1]<<4)|(val[2]<<8)|(val[3]<<12);
}
} else {
memcpy(_dest, _src2, (srcwidth << 1));
}
// next row
_src1 += srcwidth;
_src2 += srcwidth;
_src3 += srcwidth;
_dest += srcwidth;
}
// copy the last row
memcpy(_dest, _src2, (srcwidth << 1));
break;
}
}
#endif /* !_16BPP_HACK */
void filter_8888(uint32 *src, uint32 srcwidth, uint32 srcheight, uint32 *dest, uint32 filter) {
switch (filter & ENHANCEMENT_MASK) {
case HQ4X_ENHANCEMENT:
hq4x_8888((uint8*)src, (uint8*)dest, srcwidth, srcheight, srcwidth, (srcwidth << 4));
return;
case HQ2X_ENHANCEMENT:
hq2x_32((uint8*)src, (srcwidth << 2), (uint8*)dest, (srcwidth << 3), srcwidth, srcheight);
return;
case HQ2XS_ENHANCEMENT:
hq2xS_32((uint8*)src, (srcwidth << 2), (uint8*)dest, (srcwidth << 3), srcwidth, srcheight);
return;
case LQ2X_ENHANCEMENT:
lq2x_32((uint8*)src, (srcwidth << 2), (uint8*)dest, (srcwidth << 3), srcwidth, srcheight);
return;
case LQ2XS_ENHANCEMENT:
lq2xS_32((uint8*)src, (srcwidth << 2), (uint8*)dest, (srcwidth << 3), srcwidth, srcheight);
return;
case X2SAI_ENHANCEMENT:
Super2xSaI_8888((uint32*)src, (uint32*)dest, srcwidth, srcheight, srcwidth);
return;
case X2_ENHANCEMENT:
Texture2x_32((uint8*)src, (srcwidth << 2), (uint8*)dest, (srcwidth << 3), srcwidth, srcheight);
return;
}
switch (filter & (SMOOTH_FILTER_MASK|SHARP_FILTER_MASK)) {
case SMOOTH_FILTER_1:
case SMOOTH_FILTER_2:
case SMOOTH_FILTER_3:
case SMOOTH_FILTER_4:
SmoothFilter_8888((uint32*)src, srcwidth, srcheight, (uint32*)dest, (filter & SMOOTH_FILTER_MASK));
return;
case SHARP_FILTER_1:
case SHARP_FILTER_2:
SharpFilter_8888((uint32*)src, srcwidth, srcheight, (uint32*)dest, (filter & SHARP_FILTER_MASK));
return;
}
}

81
Source/GlideHQ/TextureFilters.h Normal file
View File

@ -0,0 +1,81 @@
/*
* Texture Filtering
* Version: 1.0
*
* Copyright (C) 2007 Hiroshi Morii All Rights Reserved.
* Email koolsmoky(at)users.sourceforge.net
* Web http://www.3dfxzone.it/koolsmoky
*
* this 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, or (at your option)
* any later version.
*
* this 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 GNU Make; see the file COPYING. If not, write to
* the Free Software Foundation, 675 Mass Ave, Cambridge, MA 02139, USA.
*/
#ifndef __TEXTUREFILTERS_H__
#define __TEXTUREFILTERS_H__
/* 16bpp filters are somewhat buggy and output image is not clean.
* Since there's not much time, we'll just convert them to ARGB8888
* and use 32bpp filters until fixed.
* (1:enable hack, 0:disable hack) */
#define _16BPP_HACK 1
#include "TxInternal.h"
/* enhancers */
void hq4x_8888(unsigned char * pIn, unsigned char * pOut, int Xres, int Yres, int SrcPPL, int BpL);
void hq2x_32(uint8 *srcPtr, uint32 srcPitch, uint8 *dstPtr, uint32 dstPitch, int width, int height);
void hq2xS_32(uint8 *srcPtr, uint32 srcPitch, uint8 *dstPtr, uint32 dstPitch, int width, int height);
void lq2x_32(uint8 *srcPtr, uint32 srcPitch, uint8 *dstPtr, uint32 dstPitch, int width, int height);
void lq2xS_32(uint8 *srcPtr, uint32 srcPitch, uint8 *dstPtr, uint32 dstPitch, int width, int height);
void Super2xSaI_8888(uint32 *srcPtr, uint32 *destPtr, uint32 width, uint32 height, uint32 pitch);
void Texture2x_32(uint8 *srcPtr, uint32 srcPitch, uint8 *dstPtr, uint32 dstPitch, int width, int height);
/* filters */
void SharpFilter_8888(uint32 *src, uint32 srcwidth, uint32 srcheight, uint32 *dest, uint32 filter);
void SmoothFilter_8888(uint32 *src, uint32 srcwidth, uint32 srcheight, uint32 *dest, uint32 filter);
/* helper */
void filter_8888(uint32 *src, uint32 srcwidth, uint32 srcheight, uint32 *dest, uint32 filter);
#if !_16BPP_HACK
void hq4x_init(void);
void hq4x_4444(unsigned char * pIn, unsigned char * pOut, int Xres, int Yres, int SrcPPL, int BpL);
void hq4x_1555(unsigned char * pIn, unsigned char * pOut, int Xres, int Yres, int SrcPPL, int BpL);
void hq4x_565 (unsigned char * pIn, unsigned char * pOut, int Xres, int Yres, int SrcPPL, int BpL);
void hq2x_16(uint8 *srcPtr, uint32 srcPitch, uint8 *dstPtr, uint32 dstPitch, int width, int height);
void hq2xS_16(uint8 *srcPtr, uint32 srcPitch, uint8 *dstPtr, uint32 dstPitch, int width, int height);
void lq2x_16(uint8 *srcPtr, uint32 srcPitch, uint8 *dstPtr, uint32 dstPitch, int width, int height);
void lq2xS_16(uint8 *srcPtr, uint32 srcPitch, uint8 *dstPtr, uint32 dstPitch, int width, int height);
void Super2xSaI_4444(uint16 *srcPtr, uint16 *destPtr, uint32 width, uint32 height, uint32 pitch);
void Super2xSaI_1555(uint16 *srcPtr, uint16 *destPtr, uint32 width, uint32 height, uint32 pitch);
void Super2xSaI_565 (uint16 *srcPtr, uint16 *destPtr, uint32 width, uint32 height, uint32 pitch);
void Super2xSaI_8 (uint8 *srcPtr, uint8 *destPtr, uint32 width, uint32 height, uint32 pitch);
void Texture2x_16(uint8 *srcPtr, uint32 srcPitch, uint8 *dstPtr, uint32 dstPitch, int width, int height);
void SharpFilter_4444(uint16 *src, uint32 srcwidth, uint32 srcheight, uint16 *dest, uint32 filter);
void SmoothFilter_4444(uint16 *src, uint32 srcwidth, uint32 srcheight, uint16 *dest, uint32 filter);
#endif
#endif /* __TEXTUREFILTERS_H__ */

155
Source/GlideHQ/TextureFilters_2xsai.cpp Normal file
View File

@ -0,0 +1,155 @@
/*
* Texture Filtering
* Version: 1.0
*
* Copyright (C) 2007 Hiroshi Morii All Rights Reserved.
* Email koolsmoky(at)users.sourceforge.net
* Web http://www.3dfxzone.it/koolsmoky
*
* this 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, or (at your option)
* any later version.
*
* this 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 GNU Make; see the file COPYING. If not, write to
* the Free Software Foundation, 675 Mass Ave, Cambridge, MA 02139, USA.
*/
/* Based on Derek Liauw Kie Fa and Rice1964 Super2xSaI code */
#include "TextureFilters.h"
#define GET_RESULT(A, B, C, D) ((A != C || A != D) - (B != C || B != D))
void Super2xSaI_8888(uint32 *srcPtr, uint32 *destPtr, uint32 width, uint32 height, uint32 pitch)
{
#define SAI_INTERPOLATE_8888(A, B) ((A & 0xFEFEFEFE) >> 1) + ((B & 0xFEFEFEFE) >> 1) + (A & B & 0x01010101)
#define SAI_Q_INTERPOLATE_8888(A, B, C, D) ((A & 0xFCFCFCFC) >> 2) + ((B & 0xFCFCFCFC) >> 2) + ((C & 0xFCFCFCFC) >> 2) + ((D & 0xFCFCFCFC) >> 2) \
+ ((((A & 0x03030303) + (B & 0x03030303) + (C & 0x03030303) + (D & 0x03030303)) >> 2) & 0x03030303)
#define SAI_INTERPOLATE SAI_INTERPOLATE_8888
#define SAI_Q_INTERPOLATE SAI_Q_INTERPOLATE_8888
uint32 destWidth = width << 1;
uint32 destHeight = height << 1;
uint32 color4, color5, color6;
uint32 color1, color2, color3;
uint32 colorA0, colorA1, colorA2, colorA3;
uint32 colorB0, colorB1, colorB2, colorB3;
uint32 colorS1, colorS2;
uint32 product1a, product1b, product2a, product2b;
#include "TextureFilters_2xsai.h"
#undef SAI_INTERPOLATE
#undef SAI_Q_INTERPOLATE
}
#if !_16BPP_HACK
void Super2xSaI_4444(uint16 *srcPtr, uint16 *destPtr, uint32 width, uint32 height, uint32 pitch)
{
#define SAI_INTERPOLATE_4444(A, B) ((A & 0xEEEE) >> 1) + ((B & 0xEEEE) >> 1) + (A & B & 0x1111)
#define SAI_Q_INTERPOLATE_4444(A, B, C, D) ((A & 0xCCCC) >> 2) + ((B & 0xCCCC) >> 2) + ((C & 0xCCCC) >> 2) + ((D & 0xCCCC) >> 2) \
+ ((((A & 0x3333) + (B & 0x3333) + (C & 0x3333) + (D & 0x3333)) >> 2) & 0x3333)
#define SAI_INTERPOLATE SAI_INTERPOLATE_4444
#define SAI_Q_INTERPOLATE SAI_Q_INTERPOLATE_4444
uint32 destWidth = width << 1;
uint32 destHeight = height << 1;
uint16 color4, color5, color6;
uint16 color1, color2, color3;
uint16 colorA0, colorA1, colorA2, colorA3;
uint16 colorB0, colorB1, colorB2, colorB3;
uint16 colorS1, colorS2;
uint16 product1a, product1b, product2a, product2b;
#include "TextureFilters_2xsai.h"
#undef SAI_INTERPOLATE
#undef SAI_Q_INTERPOLATE
}
void Super2xSaI_1555(uint16 *srcPtr, uint16 *destPtr, uint32 width, uint32 height, uint32 pitch)
{
#define SAI_INTERPOLATE_1555(A, B) ((A & 0x7BDE) >> 1) + ((B & 0x7BDE) >> 1) + (A & B & 0x8421)
#define SAI_Q_INTERPOLATE_1555(A, B, C, D) ((A & 0x739C) >> 2) + ((B & 0x739C) >> 2) + ((C & 0x739C) >> 2) + ((D & 0x739C) >> 2) \
+ ((((A & 0x8C63) + (B & 0x8C63) + (C & 0x8C63) + (D & 0x8C63)) >> 2) & 0x8C63)
#define SAI_INTERPOLATE SAI_INTERPOLATE_1555
#define SAI_Q_INTERPOLATE SAI_Q_INTERPOLATE_1555
uint32 destWidth = width << 1;
uint32 destHeight = height << 1;
uint16 color4, color5, color6;
uint16 color1, color2, color3;
uint16 colorA0, colorA1, colorA2, colorA3;
uint16 colorB0, colorB1, colorB2, colorB3;
uint16 colorS1, colorS2;
uint16 product1a, product1b, product2a, product2b;
#include "TextureFilters_2xsai.h"
#undef SAI_INTERPOLATE
#undef SAI_Q_INTERPOLATE
}
void Super2xSaI_565(uint16 *srcPtr, uint16 *destPtr, uint32 width, uint32 height, uint32 pitch)
{
#define SAI_INTERPOLATE_565(A, B) ((A & 0xF7DE) >> 1) + ((B & 0xF7DE) >> 1) + (A & B & 0x0821)
#define SAI_Q_INTERPOLATE_565(A, B, C, D) ((A & 0xE79C) >> 2) + ((B & 0xE79C) >> 2) + ((C & 0xE79C) >> 2) + ((D & 0xE79C) >> 2) \
+ ((((A & 0x1863) + (B & 0x1863) + (C & 0x1863) + (D & 0x1863)) >> 2) & 0x1863)
#define SAI_INTERPOLATE SAI_INTERPOLATE_565
#define SAI_Q_INTERPOLATE SAI_Q_INTERPOLATE_565
uint32 destWidth = width << 1;
uint32 destHeight = height << 1;
uint16 color4, color5, color6;
uint16 color1, color2, color3;
uint16 colorA0, colorA1, colorA2, colorA3;
uint16 colorB0, colorB1, colorB2, colorB3;
uint16 colorS1, colorS2;
uint16 product1a, product1b, product2a, product2b;
#include "TextureFilters_2xsai.h"
#undef SAI_INTERPOLATE
#undef SAI_Q_INTERPOLATE
}
void Super2xSaI_8(uint8 *srcPtr, uint8 *destPtr, uint32 width, uint32 height, uint32 pitch)
{
#define SAI_INTERPOLATE_8(A, B) ((A & 0xFE) >> 1) + ((B & 0xFE) >> 1) + (A & B & 0x01)
#define SAI_Q_INTERPOLATE_8(A, B, C, D) ((A & 0xFC) >> 2) + ((B & 0xFC) >> 2) + ((C & 0xFC) >> 2) + ((D & 0xFC) >> 2) \
+ ((((A & 0x03) + (B & 0x03) + (C & 0x03) + (D & 0x03)) >> 2) & 0x03)
#define SAI_INTERPOLATE SAI_INTERPOLATE_8
#define SAI_Q_INTERPOLATE SAI_Q_INTERPOLATE_8
uint32 destWidth = width << 1;
uint32 destHeight = height << 1;
uint8 color4, color5, color6;
uint8 color1, color2, color3;
uint8 colorA0, colorA1, colorA2, colorA3;
uint8 colorB0, colorB1, colorB2, colorB3;
uint8 colorS1, colorS2;
uint8 product1a, product1b, product2a, product2b;
#include "TextureFilters_2xsai.h"
#undef SAI_INTERPOLATE
#undef SAI_Q_INTERPOLATE
}
#endif /* !_16BPP_HACK */

145
Source/GlideHQ/TextureFilters_2xsai.h Normal file
View File

@ -0,0 +1,145 @@
/*
* Texture Filtering
* Version: 1.0
*
* Copyright (C) 2007 Hiroshi Morii All Rights Reserved.
* Email koolsmoky(at)users.sourceforge.net
* Web http://www.3dfxzone.it/koolsmoky
*
* this 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, or (at your option)
* any later version.
*
* this 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 GNU Make; see the file COPYING. If not, write to
* the Free Software Foundation, 675 Mass Ave, Cambridge, MA 02139, USA.
*/
/* Based on Derek Liauw Kie Fa and Rice1964 Super2xSaI code */
int row0, row1, row2, row3;
int col0, col1, col2, col3;
uint16 x;
uint16 y;
for (y = 0; y < height; y++) {
if ((y > 0) && (y < height - 1)) {
row0 = width;
row0 = -row0;
row1 = 0;
row2 = width;
row3 = (y == height - 2 ? width : width << 1);
} else {
row0 = 0;
row1 = 0;
row2 = 0;
row3 = 0;
}
for (x = 0; x < width; x++) {
//--------------------------------------- B0 B1 B2 B3
// 4 5 6 S2
// 1 2 3 S1
// A0 A1 A2 A3
if ((x > 0) && (x < width - 1)) {
col0 = -1;
col1 = 0;
col2 = 1;
col3 = (x == width - 2 ? 1 : 2);
} else {
col0 = 0;
col1 = 0;
col2 = 0;
col3 = 0;
}
colorB0 = *(srcPtr + col0 + row0);
colorB1 = *(srcPtr + col1 + row0);
colorB2 = *(srcPtr + col2 + row0);
colorB3 = *(srcPtr + col3 + row0);
color4 = *(srcPtr + col0 + row1);
color5 = *(srcPtr + col1 + row1);
color6 = *(srcPtr + col2 + row1);
colorS2 = *(srcPtr + col3 + row1);
color1 = *(srcPtr + col0 + row2);
color2 = *(srcPtr + col1 + row2);
color3 = *(srcPtr + col2 + row2);
colorS1 = *(srcPtr + col3 + row2);
colorA0 = *(srcPtr + col0 + row3);
colorA1 = *(srcPtr + col1 + row3);
colorA2 = *(srcPtr + col2 + row3);
colorA3 = *(srcPtr + col3 + row3);
//--------------------------------------
if (color2 == color6 && color5 != color3)
product2b = product1b = color2;
else if (color5 == color3 && color2 != color6)
product2b = product1b = color5;
else if (color5 == color3 && color2 == color6) {
int r = 0;
r += GET_RESULT(color6, color5, color1, colorA1);
r += GET_RESULT(color6, color5, color4, colorB1);
r += GET_RESULT(color6, color5, colorA2, colorS1);
r += GET_RESULT(color6, color5, colorB2, colorS2);
if (r > 0)
product2b = product1b = color6;
else if (r < 0)
product2b = product1b = color5;
else
product2b = product1b = SAI_INTERPOLATE(color5, color6);
} else {
if (color6 == color3 && color3 == colorA1 && color2 != colorA2 && color3 != colorA0)
product2b = SAI_Q_INTERPOLATE(color3, color3, color3, color2);
else if (color5 == color2 && color2 == colorA2 && colorA1 != color3 && color2 != colorA3)
product2b = SAI_Q_INTERPOLATE(color2, color2, color2, color3);
else
product2b = SAI_INTERPOLATE(color2, color3);
if (color6 == color3 && color6 == colorB1 && color5 != colorB2 && color6 != colorB0)
product1b = SAI_Q_INTERPOLATE(color6, color6, color6, color5);
else if (color5 == color2 && color5 == colorB2 && colorB1 != color6 && color5 != colorB3)
product1b = SAI_Q_INTERPOLATE(color6, color5, color5, color5);
else
product1b = SAI_INTERPOLATE(color5, color6);
}
if (color5 == color3 && color2 != color6 && color4 == color5 && color5 != colorA2)
product2a = SAI_INTERPOLATE(color2, color5);
else if (color5 == color1 && color6 == color5 && color4 != color2 && color5 != colorA0)
product2a = SAI_INTERPOLATE(color2, color5);
else
product2a = color2;
if (color2 == color6 && color5 != color3 && color1 == color2 && color2 != colorB2)
product1a = SAI_INTERPOLATE(color2, color5);
else if (color4 == color2 && color3 == color2 && color1 != color5 && color2 != colorB0)
product1a = SAI_INTERPOLATE(color2, color5);
else
product1a = color5;
destPtr[0] = product1a;
destPtr[1] = product1b;
destPtr[destWidth] = product2a;
destPtr[destWidth + 1] = product2b;
srcPtr++;
destPtr += 2;
}
srcPtr += (pitch-width);
destPtr += (((pitch-width)<<1)+(pitch<<1));
}

1510
Source/GlideHQ/TextureFilters_hq2x.cpp Normal file
View File

File diff suppressed because it is too large Load Diff

1847
Source/GlideHQ/TextureFilters_hq2x.h Normal file
View File

File diff suppressed because it is too large Load Diff

892
Source/GlideHQ/TextureFilters_hq4x.cpp Normal file
View File

@ -0,0 +1,892 @@
/*
* Texture Filtering
* Version: 1.0
*
* Copyright (C) 2007 Hiroshi Morii All Rights Reserved.
* Email koolsmoky(at)users.sourceforge.net
* Web http://www.3dfxzone.it/koolsmoky
*
* this 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, or (at your option)
* any later version.
*
* this 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 GNU Make; see the file COPYING. If not, write to
* the Free Software Foundation, 675 Mass Ave, Cambridge, MA 02139, USA.
*/
/* Based on Maxim Stepin and Rice1964 hq4x code */
#include <math.h>
#include <stdlib.h>
#include "TextureFilters.h"
#if !_16BPP_HACK
static uint32 RGB444toYUV[4096];
#define RGB444toYUV(val) RGB444toYUV[val & 0x0FFF] /* val = ARGB4444 */
/*inline static uint32 RGB444toYUV(uint32 val)
{
uint32 r, g, b, Y, u, v;
r = (val & 0x0F00) >> 4;
g = (val & 0x00F0);
b = val & 0x000F;
r |= r >> 4;
g |= g >> 4;
b |= b << 4;
Y = (r + g + b) >> 2;
u = 128 + ((r - b) >> 2);
v = 128 + ((2*g - r - b)>>3);
return ((Y << 16) | (u << 8) | v);
}*/
static uint32 RGB555toYUV(uint32 val)
{
uint32 r, g, b, Y, u, v;
r = (val & 0x7C00) >> 7;
g = (val & 0x03E0) >> 2;
b = (val & 0x001F) << 3;
r |= r >> 5;
g |= g >> 5;
b |= b >> 5;
Y = (r + g + b) >> 2;
u = 128 + ((r - b) >> 2);
v = 128 + ((2*g - r - b)>>3);
return ((Y << 16) | (u << 8) | v);
}
static uint32 RGB565toYUV(uint32 val)
{
uint32 r, g, b, Y, u, v;
r = (val & 0xF800) >> 8;
g = (val & 0x07E0) >> 3;
b = (val & 0x001F) << 3;
r |= r >> 5;
g |= g >> 6;
b |= b >> 5;
Y = (r + g + b) >> 2;
u = 128 + ((r - b) >> 2);
v = 128 + ((2*g - r - b)>>3);
return ((Y << 16) | (u << 8) | v);
}
#endif /* !_16BPP_HACK */
static uint32 RGB888toYUV(uint32 val)
{
#if 0
uint32 Yuv;
__asm {
mov eax, dword ptr [val];
mov ebx, eax;
mov ecx, eax;
and ebx, 0x000000ff; // b
and eax, 0x00ff0000; // r
and ecx, 0x0000ff00; // g
shl ebx, 14;
shr eax, 2;
shl ecx, 6;
mov edx, ebx;
add edx, eax;
add edx, ecx;
and edx, 0xffff0000;
sub eax, ebx;
add eax, 0x00800000;
shr eax, 8;
or edx, eax;
sub eax, 0x00800000;
and edx, 0xffffff00;
add ecx, 0x00800000;
shr ecx, 5;
shr ebx, 7;
add eax, ebx;
sub ecx, eax;
shr ecx, 11;
or edx, ecx;
mov dword ptr [Yuv], edx;
}
return Yuv;
#else
uint32 r, g, b, Y, u, v;
r = (val & 0x00ff0000) >> 16;
g = (val & 0x0000ff00) >> 8;
b = val & 0x000000ff;
Y = (r + g + b) >> 2;
u = (0x00000200 + r - b) >> 2;
v = (0x00000400 + (g << 1) - r - b) >> 3;
return ((Y << 16) | (u << 8) | v);
#endif
}
#define Ymask 0x00FF0000
#define Umask 0x0000FF00
#define Vmask 0x000000FF
#define trY 0x00300000 // ?
#define trU 0x00000700 // ??
#define trV 0x00000006 // ???
#define HQ4X_INTERP1(n, b) \
static void hq4x_Interp1_##n (uint8 * pc, uint##b p1, uint##b p2) \
{ \
/* *((uint##b*)pc) = (p1*3+p2) >> 2; */ \
*((uint##b*)pc) = INTERP_##n##_MASK_1_3((INTERP_##n##_MASK_1_3(p1)*3 + INTERP_##n##_MASK_1_3(p2)) / 4) \
| INTERP_##n##_MASK_SHIFTBACK_2_4((INTERP_##n##_MASK_SHIFT_2_4(p1)*3 + INTERP_##n##_MASK_SHIFT_2_4(p2)) / 4 ); \
}
#define HQ4X_INTERP2(n, b) \
static void hq4x_Interp2_##n (uint8 * pc, uint##b p1, uint##b p2, uint##b p3) \
{ \
/**((uint##b*)pc) = (p1*2+p2+p3) >> 2;*/ \
*((uint##b*)pc) = INTERP_##n##_MASK_1_3((INTERP_##n##_MASK_1_3(p1)*2 + INTERP_##n##_MASK_1_3(p2) + INTERP_##n##_MASK_1_3(p3)) / 4) \
| INTERP_##n##_MASK_SHIFTBACK_2_4((INTERP_##n##_MASK_SHIFT_2_4(p1)*2 + INTERP_##n##_MASK_SHIFT_2_4(p2) + INTERP_##n##_MASK_SHIFT_2_4(p3)) / 4); \
}
#define HQ4X_INTERP3(n, b) \
static void hq4x_Interp3_##n (uint8 * pc, uint##b p1, uint##b p2) \
{ \
/**((uint##b*)pc) = (p1*7+p2)/8;*/ \
*((uint##b*)pc) = INTERP_##n##_MASK_1_3((INTERP_##n##_MASK_1_3(p1)*7 + INTERP_##n##_MASK_1_3(p2)) / 8) \
| INTERP_##n##_MASK_SHIFTBACK_2_4((INTERP_##n##_MASK_SHIFT_2_4(p1)*7 + INTERP_##n##_MASK_SHIFT_2_4(p2)) / 8); \
}
#define HQ4X_INTERP5(n, b) \
static void hq4x_Interp5_##n (uint8 * pc, uint##b p1, uint##b p2) \
{ \
/**((uint##b*)pc) = (p1+p2) >> 1;*/ \
*((uint##b*)pc) = INTERP_##n##_MASK_1_3((INTERP_##n##_MASK_1_3(p1) + INTERP_##n##_MASK_1_3(p2)) / 2) \
| INTERP_##n##_MASK_SHIFTBACK_2_4((INTERP_##n##_MASK_SHIFT_2_4(p1) + INTERP_##n##_MASK_SHIFT_2_4(p2)) / 2); \
}
#define HQ4X_INTERP6(n, b) \
static void hq4x_Interp6_##n (uint8 * pc, uint##b p1, uint##b p2, uint##b p3) \
{ \
/**((uint##b*)pc) = (p1*5+p2*2+p3)/8;*/ \
*((uint##b*)pc) = INTERP_##n##_MASK_1_3((INTERP_##n##_MASK_1_3(p1)*5 + INTERP_##n##_MASK_1_3(p2)*2 + INTERP_##n##_MASK_1_3(p3)) / 8) \
| INTERP_##n##_MASK_SHIFTBACK_2_4((INTERP_##n##_MASK_SHIFT_2_4(p1)*5 + INTERP_##n##_MASK_SHIFT_2_4(p2)*2 + INTERP_##n##_MASK_SHIFT_2_4(p3)) / 8); \
}
#define HQ4X_INTERP7(n, b) \
static void hq4x_Interp7_##n (uint8 * pc, uint##b p1, uint##b p2, uint##b p3) \
{ \
/**((uint##b*)pc) = (p1*6+p2+p3)/8;*/ \
*((uint##b*)pc) = INTERP_##n##_MASK_1_3((INTERP_##n##_MASK_1_3(p1)*6 + INTERP_##n##_MASK_1_3(p2) + INTERP_##n##_MASK_1_3(p3)) / 8) \
| INTERP_##n##_MASK_SHIFTBACK_2_4((INTERP_##n##_MASK_SHIFT_2_4(p1)*6 + INTERP_##n##_MASK_SHIFT_2_4(p2) + INTERP_##n##_MASK_SHIFT_2_4(p3)) / 8); \
}
#define HQ4X_INTERP8(n, b) \
static void hq4x_Interp8_##n (uint8 * pc, uint##b p1, uint##b p2) \
{ \
/**((uint##b*)pc) = (p1*5+p2*3)/8;*/ \
*((uint##b*)pc) = INTERP_##n##_MASK_1_3((INTERP_##n##_MASK_1_3(p1)*5 + INTERP_##n##_MASK_1_3(p2)*3) / 8) \
| INTERP_##n##_MASK_SHIFTBACK_2_4((INTERP_##n##_MASK_SHIFT_2_4(p1)*5 + INTERP_##n##_MASK_SHIFT_2_4(p2)*3) / 8); \
}
#if !_16BPP_HACK
#define INTERP_4444_MASK_1_3(v) (v & 0x0F0F)
#define INTERP_4444_MASK_SHIFT_2_4(v) ((v & 0xF0F0) >> 4)
#define INTERP_4444_MASK_SHIFTBACK_2_4(v) (INTERP_4444_MASK_1_3(v) << 4)
HQ4X_INTERP1(4444, 16)
HQ4X_INTERP2(4444, 16)
HQ4X_INTERP3(4444, 16)
HQ4X_INTERP5(4444, 16)
HQ4X_INTERP6(4444, 16)
HQ4X_INTERP7(4444, 16)
HQ4X_INTERP8(4444, 16)
#define INTERP_1555_MASK_1_3(v) (v & 0x7C1F)
#define INTERP_1555_MASK_SHIFT_2_4(v) ((v & 0x83E0) >> 5)
#define INTERP_1555_MASK_SHIFTBACK_2_4(v) (INTERP_1555_MASK_1_3(v) << 5)
HQ4X_INTERP1(1555, 16)
HQ4X_INTERP2(1555, 16)
HQ4X_INTERP3(1555, 16)
HQ4X_INTERP5(1555, 16)
HQ4X_INTERP6(1555, 16)
HQ4X_INTERP7(1555, 16)
HQ4X_INTERP8(1555, 16)
#define INTERP_565_MASK_1_3(v) (v & 0xF81F)
#define INTERP_565_MASK_SHIFT_2_4(v) ((v & 0x7E0) >> 5)
#define INTERP_565_MASK_SHIFTBACK_2_4(v) (INTERP_565_MASK_1_3(v) << 5)
HQ4X_INTERP1(565, 16)
HQ4X_INTERP2(565, 16)
HQ4X_INTERP3(565, 16)
HQ4X_INTERP5(565, 16)
HQ4X_INTERP6(565, 16)
HQ4X_INTERP7(565, 16)
HQ4X_INTERP8(565, 16)
#endif /* !_16BPP_HACK */
#define INTERP_8888_MASK_1_3(v) (v & 0x00FF00FF)
#define INTERP_8888_MASK_SHIFT_2_4(v) ((v & 0xFF00FF00) >> 8)
#define INTERP_8888_MASK_SHIFTBACK_2_4(v) (INTERP_8888_MASK_1_3(v) << 8)
HQ4X_INTERP1(8888, 32)
HQ4X_INTERP2(8888, 32)
HQ4X_INTERP3(8888, 32)
HQ4X_INTERP5(8888, 32)
HQ4X_INTERP6(8888, 32)
HQ4X_INTERP7(8888, 32)
HQ4X_INTERP8(8888, 32)
#define PIXEL00_0 *((int*)(pOut)) = c[5];
#define PIXEL00_11 hq4x_Interp1(pOut, c[5], c[4]);
#define PIXEL00_12 hq4x_Interp1(pOut, c[5], c[2]);
#define PIXEL00_20 hq4x_Interp2(pOut, c[5], c[2], c[4]);
#define PIXEL00_50 hq4x_Interp5(pOut, c[2], c[4]);
#define PIXEL00_80 hq4x_Interp8(pOut, c[5], c[1]);
#define PIXEL00_81 hq4x_Interp8(pOut, c[5], c[4]);
#define PIXEL00_82 hq4x_Interp8(pOut, c[5], c[2]);
#define PIXEL01_0 *((int*)(pOut+BPP)) = c[5];
#define PIXEL01_10 hq4x_Interp1(pOut+BPP, c[5], c[1]);
#define PIXEL01_12 hq4x_Interp1(pOut+BPP, c[5], c[2]);
#define PIXEL01_14 hq4x_Interp1(pOut+BPP, c[2], c[5]);
#define PIXEL01_21 hq4x_Interp2(pOut+BPP, c[2], c[5], c[4]);
#define PIXEL01_31 hq4x_Interp3(pOut+BPP, c[5], c[4]);
#define PIXEL01_50 hq4x_Interp5(pOut+BPP, c[2], c[5]);
#define PIXEL01_60 hq4x_Interp6(pOut+BPP, c[5], c[2], c[4]);
#define PIXEL01_61 hq4x_Interp6(pOut+BPP, c[5], c[2], c[1]);
#define PIXEL01_82 hq4x_Interp8(pOut+BPP, c[5], c[2]);
#define PIXEL01_83 hq4x_Interp8(pOut+BPP, c[2], c[4]);
#define PIXEL02_0 *((int*)(pOut+BPP2)) = c[5];
#define PIXEL02_10 hq4x_Interp1(pOut+BPP2, c[5], c[3]);
#define PIXEL02_11 hq4x_Interp1(pOut+BPP2, c[5], c[2]);
#define PIXEL02_13 hq4x_Interp1(pOut+BPP2, c[2], c[5]);
#define PIXEL02_21 hq4x_Interp2(pOut+BPP2, c[2], c[5], c[6]);
#define PIXEL02_32 hq4x_Interp3(pOut+BPP2, c[5], c[6]);
#define PIXEL02_50 hq4x_Interp5(pOut+BPP2, c[2], c[5]);
#define PIXEL02_60 hq4x_Interp6(pOut+BPP2, c[5], c[2], c[6]);
#define PIXEL02_61 hq4x_Interp6(pOut+BPP2, c[5], c[2], c[3]);
#define PIXEL02_81 hq4x_Interp8(pOut+BPP2, c[5], c[2]);
#define PIXEL02_83 hq4x_Interp8(pOut+BPP2, c[2], c[6]);
#define PIXEL03_0 *((int*)(pOut+BPP3)) = c[5];
#define PIXEL03_11 hq4x_Interp1(pOut+BPP3, c[5], c[2]);
#define PIXEL03_12 hq4x_Interp1(pOut+BPP3, c[5], c[6]);
#define PIXEL03_20 hq4x_Interp2(pOut+BPP3, c[5], c[2], c[6]);
#define PIXEL03_50 hq4x_Interp5(pOut+BPP3, c[2], c[6]);
#define PIXEL03_80 hq4x_Interp8(pOut+BPP3, c[5], c[3]);
#define PIXEL03_81 hq4x_Interp8(pOut+BPP3, c[5], c[2]);
#define PIXEL03_82 hq4x_Interp8(pOut+BPP3, c[5], c[6]);
#define PIXEL10_0 *((int*)(pOut+BpL)) = c[5];
#define PIXEL10_10 hq4x_Interp1(pOut+BpL, c[5], c[1]);
#define PIXEL10_11 hq4x_Interp1(pOut+BpL, c[5], c[4]);
#define PIXEL10_13 hq4x_Interp1(pOut+BpL, c[4], c[5]);
#define PIXEL10_21 hq4x_Interp2(pOut+BpL, c[4], c[5], c[2]);
#define PIXEL10_32 hq4x_Interp3(pOut+BpL, c[5], c[2]);
#define PIXEL10_50 hq4x_Interp5(pOut+BpL, c[4], c[5]);
#define PIXEL10_60 hq4x_Interp6(pOut+BpL, c[5], c[4], c[2]);
#define PIXEL10_61 hq4x_Interp6(pOut+BpL, c[5], c[4], c[1]);
#define PIXEL10_81 hq4x_Interp8(pOut+BpL, c[5], c[4]);
#define PIXEL10_83 hq4x_Interp8(pOut+BpL, c[4], c[2]);
#define PIXEL11_0 *((int*)(pOut+BpL+BPP)) = c[5];
#define PIXEL11_30 hq4x_Interp3(pOut+BpL+BPP, c[5], c[1]);
#define PIXEL11_31 hq4x_Interp3(pOut+BpL+BPP, c[5], c[4]);
#define PIXEL11_32 hq4x_Interp3(pOut+BpL+BPP, c[5], c[2]);
#define PIXEL11_70 hq4x_Interp7(pOut+BpL+BPP, c[5], c[4], c[2]);
#define PIXEL12_0 *((int*)(pOut+BpL+BPP2)) = c[5];
#define PIXEL12_30 hq4x_Interp3(pOut+BpL+BPP2, c[5], c[3]);
#define PIXEL12_31 hq4x_Interp3(pOut+BpL+BPP2, c[5], c[2]);
#define PIXEL12_32 hq4x_Interp3(pOut+BpL+BPP2, c[5], c[6]);
#define PIXEL12_70 hq4x_Interp7(pOut+BpL+BPP2, c[5], c[6], c[2]);
#define PIXEL13_0 *((int*)(pOut+BpL+BPP3)) = c[5];
#define PIXEL13_10 hq4x_Interp1(pOut+BpL+BPP3, c[5], c[3]);
#define PIXEL13_12 hq4x_Interp1(pOut+BpL+BPP3, c[5], c[6]);
#define PIXEL13_14 hq4x_Interp1(pOut+BpL+BPP3, c[6], c[5]);
#define PIXEL13_21 hq4x_Interp2(pOut+BpL+BPP3, c[6], c[5], c[2]);
#define PIXEL13_31 hq4x_Interp3(pOut+BpL+BPP3, c[5], c[2]);
#define PIXEL13_50 hq4x_Interp5(pOut+BpL+BPP3, c[6], c[5]);
#define PIXEL13_60 hq4x_Interp6(pOut+BpL+BPP3, c[5], c[6], c[2]);
#define PIXEL13_61 hq4x_Interp6(pOut+BpL+BPP3, c[5], c[6], c[3]);
#define PIXEL13_82 hq4x_Interp8(pOut+BpL+BPP3, c[5], c[6]);
#define PIXEL13_83 hq4x_Interp8(pOut+BpL+BPP3, c[6], c[2]);
#define PIXEL20_0 *((int*)(pOut+BpL+BpL)) = c[5];
#define PIXEL20_10 hq4x_Interp1(pOut+BpL+BpL, c[5], c[7]);
#define PIXEL20_12 hq4x_Interp1(pOut+BpL+BpL, c[5], c[4]);
#define PIXEL20_14 hq4x_Interp1(pOut+BpL+BpL, c[4], c[5]);
#define PIXEL20_21 hq4x_Interp2(pOut+BpL+BpL, c[4], c[5], c[8]);
#define PIXEL20_31 hq4x_Interp3(pOut+BpL+BpL, c[5], c[8]);
#define PIXEL20_50 hq4x_Interp5(pOut+BpL+BpL, c[4], c[5]);
#define PIXEL20_60 hq4x_Interp6(pOut+BpL+BpL, c[5], c[4], c[8]);
#define PIXEL20_61 hq4x_Interp6(pOut+BpL+BpL, c[5], c[4], c[7]);
#define PIXEL20_82 hq4x_Interp8(pOut+BpL+BpL, c[5], c[4]);
#define PIXEL20_83 hq4x_Interp8(pOut+BpL+BpL, c[4], c[8]);
#define PIXEL21_0 *((int*)(pOut+BpL+BpL+BPP)) = c[5];
#define PIXEL21_30 hq4x_Interp3(pOut+BpL+BpL+BPP, c[5], c[7]);
#define PIXEL21_31 hq4x_Interp3(pOut+BpL+BpL+BPP, c[5], c[8]);
#define PIXEL21_32 hq4x_Interp3(pOut+BpL+BpL+BPP, c[5], c[4]);
#define PIXEL21_70 hq4x_Interp7(pOut+BpL+BpL+BPP, c[5], c[4], c[8]);
#define PIXEL22_0 *((int*)(pOut+BpL+BpL+BPP2)) = c[5];
#define PIXEL22_30 hq4x_Interp3(pOut+BpL+BpL+BPP2, c[5], c[9]);
#define PIXEL22_31 hq4x_Interp3(pOut+BpL+BpL+BPP2, c[5], c[6]);
#define PIXEL22_32 hq4x_Interp3(pOut+BpL+BpL+BPP2, c[5], c[8]);
#define PIXEL22_70 hq4x_Interp7(pOut+BpL+BpL+BPP2, c[5], c[6], c[8]);
#define PIXEL23_0 *((int*)(pOut+BpL+BpL+BPP3)) = c[5];
#define PIXEL23_10 hq4x_Interp1(pOut+BpL+BpL+BPP3, c[5], c[9]);
#define PIXEL23_11 hq4x_Interp1(pOut+BpL+BpL+BPP3, c[5], c[6]);
#define PIXEL23_13 hq4x_Interp1(pOut+BpL+BpL+BPP3, c[6], c[5]);
#define PIXEL23_21 hq4x_Interp2(pOut+BpL+BpL+BPP3, c[6], c[5], c[8]);
#define PIXEL23_32 hq4x_Interp3(pOut+BpL+BpL+BPP3, c[5], c[8]);
#define PIXEL23_50 hq4x_Interp5(pOut+BpL+BpL+BPP3, c[6], c[5]);
#define PIXEL23_60 hq4x_Interp6(pOut+BpL+BpL+BPP3, c[5], c[6], c[8]);
#define PIXEL23_61 hq4x_Interp6(pOut+BpL+BpL+BPP3, c[5], c[6], c[9]);
#define PIXEL23_81 hq4x_Interp8(pOut+BpL+BpL+BPP3, c[5], c[6]);
#define PIXEL23_83 hq4x_Interp8(pOut+BpL+BpL+BPP3, c[6], c[8]);
#define PIXEL30_0 *((int*)(pOut+BpL+BpL+BpL)) = c[5];
#define PIXEL30_11 hq4x_Interp1(pOut+BpL+BpL+BpL, c[5], c[8]);
#define PIXEL30_12 hq4x_Interp1(pOut+BpL+BpL+BpL, c[5], c[4]);
#define PIXEL30_20 hq4x_Interp2(pOut+BpL+BpL+BpL, c[5], c[8], c[4]);
#define PIXEL30_50 hq4x_Interp5(pOut+BpL+BpL+BpL, c[8], c[4]);
#define PIXEL30_80 hq4x_Interp8(pOut+BpL+BpL+BpL, c[5], c[7]);
#define PIXEL30_81 hq4x_Interp8(pOut+BpL+BpL+BpL, c[5], c[8]);
#define PIXEL30_82 hq4x_Interp8(pOut+BpL+BpL+BpL, c[5], c[4]);
#define PIXEL31_0 *((int*)(pOut+BpL+BpL+BpL+BPP)) = c[5];
#define PIXEL31_10 hq4x_Interp1(pOut+BpL+BpL+BpL+BPP, c[5], c[7]);
#define PIXEL31_11 hq4x_Interp1(pOut+BpL+BpL+BpL+BPP, c[5], c[8]);
#define PIXEL31_13 hq4x_Interp1(pOut+BpL+BpL+BpL+BPP, c[8], c[5]);
#define PIXEL31_21 hq4x_Interp2(pOut+BpL+BpL+BpL+BPP, c[8], c[5], c[4]);
#define PIXEL31_32 hq4x_Interp3(pOut+BpL+BpL+BpL+BPP, c[5], c[4]);
#define PIXEL31_50 hq4x_Interp5(pOut+BpL+BpL+BpL+BPP, c[8], c[5]);
#define PIXEL31_60 hq4x_Interp6(pOut+BpL+BpL+BpL+BPP, c[5], c[8], c[4]);
#define PIXEL31_61 hq4x_Interp6(pOut+BpL+BpL+BpL+BPP, c[5], c[8], c[7]);
#define PIXEL31_81 hq4x_Interp8(pOut+BpL+BpL+BpL+BPP, c[5], c[8]);
#define PIXEL31_83 hq4x_Interp8(pOut+BpL+BpL+BpL+BPP, c[8], c[4]);
#define PIXEL32_0 *((int*)(pOut+BpL+BpL+BpL+BPP2)) = c[5];
#define PIXEL32_10 hq4x_Interp1(pOut+BpL+BpL+BpL+BPP2, c[5], c[9]);
#define PIXEL32_12 hq4x_Interp1(pOut+BpL+BpL+BpL+BPP2, c[5], c[8]);
#define PIXEL32_14 hq4x_Interp1(pOut+BpL+BpL+BpL+BPP2, c[8], c[5]);
#define PIXEL32_21 hq4x_Interp2(pOut+BpL+BpL+BpL+BPP2, c[8], c[5], c[6]);
#define PIXEL32_31 hq4x_Interp3(pOut+BpL+BpL+BpL+BPP2, c[5], c[6]);
#define PIXEL32_50 hq4x_Interp5(pOut+BpL+BpL+BpL+BPP2, c[8], c[5]);
#define PIXEL32_60 hq4x_Interp6(pOut+BpL+BpL+BpL+BPP2, c[5], c[8], c[6]);
#define PIXEL32_61 hq4x_Interp6(pOut+BpL+BpL+BpL+BPP2, c[5], c[8], c[9]);
#define PIXEL32_82 hq4x_Interp8(pOut+BpL+BpL+BpL+BPP2, c[5], c[8]);
#define PIXEL32_83 hq4x_Interp8(pOut+BpL+BpL+BpL+BPP2, c[8], c[6]);
#define PIXEL33_0 *((int*)(pOut+BpL+BpL+BpL+BPP3)) = c[5];
#define PIXEL33_11 hq4x_Interp1(pOut+BpL+BpL+BpL+BPP3, c[5], c[6]);
#define PIXEL33_12 hq4x_Interp1(pOut+BpL+BpL+BpL+BPP3, c[5], c[8]);
#define PIXEL33_20 hq4x_Interp2(pOut+BpL+BpL+BpL+BPP3, c[5], c[8], c[6]);
#define PIXEL33_50 hq4x_Interp5(pOut+BpL+BpL+BpL+BPP3, c[8], c[6]);
#define PIXEL33_80 hq4x_Interp8(pOut+BpL+BpL+BpL+BPP3, c[5], c[9]);
#define PIXEL33_81 hq4x_Interp8(pOut+BpL+BpL+BpL+BPP3, c[5], c[6]);
#define PIXEL33_82 hq4x_Interp8(pOut+BpL+BpL+BpL+BPP3, c[5], c[8]);
#define HQ4X_DIFF(n, b) \
static int Diff_##n (uint##b w1, uint##b w2) \
{ \
int YUV1, YUV2; \
YUV1 = RGB##n##toYUV(w1); \
YUV2 = RGB##n##toYUV(w2); \
return ( ( abs((YUV1 & Ymask) - (YUV2 & Ymask)) > trY ) || \
( abs((YUV1 & Umask) - (YUV2 & Umask)) > trU ) || \
( abs((YUV1 & Vmask) - (YUV2 & Vmask)) > trV ) ); \
}
HQ4X_DIFF(888, 32)
#if !_16BPP_HACK
HQ4X_DIFF(444, 16)
HQ4X_DIFF(555, 16)
HQ4X_DIFF(565, 16)
void hq4x_4444(unsigned char * pIn, unsigned char * pOut, int Xres, int Yres, int SrcPPL, int BpL)
{
#define hq4x_Interp1 hq4x_Interp1_4444
#define hq4x_Interp2 hq4x_Interp2_4444
#define hq4x_Interp3 hq4x_Interp3_4444
#define hq4x_Interp4 hq4x_Interp4_4444
#define hq4x_Interp5 hq4x_Interp5_4444
#define hq4x_Interp6 hq4x_Interp6_4444
#define hq4x_Interp7 hq4x_Interp7_4444
#define hq4x_Interp8 hq4x_Interp8_4444
#define Diff Diff_444
#define BPP 2
#define BPP2 4
#define BPP3 6
int i, j, k;
int prevline, nextline;
uint16 w[10];
uint16 c[10];
int pattern;
int flag;
int YUV1, YUV2;
// +----+----+----+
// | | | |
// | w1 | w2 | w3 |
// +----+----+----+
// | | | |
// | w4 | w5 | w6 |
// +----+----+----+
// | | | |
// | w7 | w8 | w9 |
// +----+----+----+
for (j = 0; j < Yres; j++) {
if (j>0) prevline = -SrcPPL*2; else prevline = 0;
if (j<Yres-1) nextline = SrcPPL*2; else nextline = 0;
for (i=0; i<Xres; i++) {
w[2] = *((uint16*)(pIn + prevline));
w[5] = *((uint16*)pIn);
w[8] = *((uint16*)(pIn + nextline));
if (i>0) {
w[1] = *((uint16*)(pIn + prevline - 2));
w[4] = *((uint16*)(pIn - 2));
w[7] = *((uint16*)(pIn + nextline - 2));
} else {
w[1] = w[2];
w[4] = w[5];
w[7] = w[8];
}
if (i<Xres-1) {
w[3] = *((uint16*)(pIn + prevline + 2));
w[6] = *((uint16*)(pIn + 2));
w[9] = *((uint16*)(pIn + nextline + 2));
} else {
w[3] = w[2];
w[6] = w[5];
w[9] = w[8];
}
pattern = 0;
flag = 1;
YUV1 = RGB444toYUV(w[5]);
for (k=1; k<=9; k++) {
if (k==5) continue;
if ( w[k] != w[5] ) {
YUV2 = RGB444toYUV(w[k]);
if ( ( abs((YUV1 & Ymask) - (YUV2 & Ymask)) > trY ) ||
( abs((YUV1 & Umask) - (YUV2 & Umask)) > trU ) ||
( abs((YUV1 & Vmask) - (YUV2 & Vmask)) > trV ) )
pattern |= flag;
}
flag <<= 1;
}
for (k=1; k<=9; k++)
c[k] = w[k];
#include "TextureFilters_hq4x.h"
pIn+=2;
pOut+=8;
}
pIn += 2*(SrcPPL-Xres);
pOut+= 8*(SrcPPL-Xres);
pOut+=BpL;
pOut+=BpL;
pOut+=BpL;
}
#undef BPP
#undef BPP2
#undef BPP3
#undef Diff
#undef hq4x_Interp1
#undef hq4x_Interp2
#undef hq4x_Interp3
#undef hq4x_Interp4
#undef hq4x_Interp5
#undef hq4x_Interp6
#undef hq4x_Interp7
#undef hq4x_Interp8
}
void hq4x_1555(unsigned char * pIn, unsigned char * pOut, int Xres, int Yres, int SrcPPL, int BpL)
{
#define hq4x_Interp1 hq4x_Interp1_1555
#define hq4x_Interp2 hq4x_Interp2_1555
#define hq4x_Interp3 hq4x_Interp3_1555
#define hq4x_Interp4 hq4x_Interp4_1555
#define hq4x_Interp5 hq4x_Interp5_1555
#define hq4x_Interp6 hq4x_Interp6_1555
#define hq4x_Interp7 hq4x_Interp7_1555
#define hq4x_Interp8 hq4x_Interp8_1555
#define Diff Diff_555
#define BPP 2
#define BPP2 4
#define BPP3 6
int i, j, k;
int prevline, nextline;
uint16 w[10];
uint16 c[10];
int pattern;
int flag;
int YUV1, YUV2;
// +----+----+----+
// | | | |
// | w1 | w2 | w3 |
// +----+----+----+
// | | | |
// | w4 | w5 | w6 |
// +----+----+----+
// | | | |
// | w7 | w8 | w9 |
// +----+----+----+
for (j = 0; j < Yres; j++) {
if (j>0) prevline = -SrcPPL*2; else prevline = 0;
if (j<Yres-1) nextline = SrcPPL*2; else nextline = 0;
for (i=0; i<Xres; i++) {
w[2] = *((uint16*)(pIn + prevline));
w[5] = *((uint16*)pIn);
w[8] = *((uint16*)(pIn + nextline));
if (i>0) {
w[1] = *((uint16*)(pIn + prevline - 2));
w[4] = *((uint16*)(pIn - 2));
w[7] = *((uint16*)(pIn + nextline - 2));
} else {
w[1] = w[2];
w[4] = w[5];
w[7] = w[8];
}
if (i<Xres-1) {
w[3] = *((uint16*)(pIn + prevline + 2));
w[6] = *((uint16*)(pIn + 2));
w[9] = *((uint16*)(pIn + nextline + 2));
} else {
w[3] = w[2];
w[6] = w[5];
w[9] = w[8];
}
pattern = 0;
flag = 1;
YUV1 = RGB555toYUV(w[5]);
for (k=1; k<=9; k++) {
if (k==5) continue;
if ( w[k] != w[5] ) {
YUV2 = RGB555toYUV(w[k]);
if ( ( abs((YUV1 & Ymask) - (YUV2 & Ymask)) > trY ) ||
( abs((YUV1 & Umask) - (YUV2 & Umask)) > trU ) ||
( abs((YUV1 & Vmask) - (YUV2 & Vmask)) > trV ) )
pattern |= flag;
}
flag <<= 1;
}
for (k=1; k<=9; k++)
c[k] = w[k];
#include "TextureFilters_hq4x.h"
pIn+=2;
pOut+=8;
}
pIn += 2*(SrcPPL-Xres);
pOut+= 8*(SrcPPL-Xres);
pOut+=BpL;
pOut+=BpL;
pOut+=BpL;
}
#undef BPP
#undef BPP2
#undef BPP3
#undef Diff
#undef hq4x_Interp1
#undef hq4x_Interp2
#undef hq4x_Interp3
#undef hq4x_Interp4
#undef hq4x_Interp5
#undef hq4x_Interp6
#undef hq4x_Interp7
#undef hq4x_Interp8
}
void hq4x_565(unsigned char * pIn, unsigned char * pOut, int Xres, int Yres, int SrcPPL, int BpL)
{
#define hq4x_Interp1 hq4x_Interp1_565
#define hq4x_Interp2 hq4x_Interp2_565
#define hq4x_Interp3 hq4x_Interp3_565
#define hq4x_Interp4 hq4x_Interp4_565
#define hq4x_Interp5 hq4x_Interp5_565
#define hq4x_Interp6 hq4x_Interp6_565
#define hq4x_Interp7 hq4x_Interp7_565
#define hq4x_Interp8 hq4x_Interp8_565
#define Diff Diff_565
#define BPP 2
#define BPP2 4
#define BPP3 6
int i, j, k;
int prevline, nextline;
uint16 w[10];
uint16 c[10];
int pattern;
int flag;
int YUV1, YUV2;
// +----+----+----+
// | | | |
// | w1 | w2 | w3 |
// +----+----+----+
// | | | |
// | w4 | w5 | w6 |
// +----+----+----+
// | | | |
// | w7 | w8 | w9 |
// +----+----+----+
for (j = 0; j < Yres; j++) {
if (j>0) prevline = -SrcPPL*2; else prevline = 0;
if (j<Yres-1) nextline = SrcPPL*2; else nextline = 0;
for (i=0; i<Xres; i++) {
w[2] = *((uint16*)(pIn + prevline));
w[5] = *((uint16*)pIn);
w[8] = *((uint16*)(pIn + nextline));
if (i>0) {
w[1] = *((uint16*)(pIn + prevline - 2));
w[4] = *((uint16*)(pIn - 2));
w[7] = *((uint16*)(pIn + nextline - 2));
} else {
w[1] = w[2];
w[4] = w[5];
w[7] = w[8];
}
if (i<Xres-1) {
w[3] = *((uint16*)(pIn + prevline + 2));
w[6] = *((uint16*)(pIn + 2));
w[9] = *((uint16*)(pIn + nextline + 2));
} else {
w[3] = w[2];
w[6] = w[5];
w[9] = w[8];
}
pattern = 0;
flag = 1;
YUV1 = RGB565toYUV(w[5]);
for (k=1; k<=9; k++) {
if (k==5) continue;
if ( w[k] != w[5] ) {
YUV2 = RGB565toYUV(w[k]);
if ( ( abs((YUV1 & Ymask) - (YUV2 & Ymask)) > trY ) ||
( abs((YUV1 & Umask) - (YUV2 & Umask)) > trU ) ||
( abs((YUV1 & Vmask) - (YUV2 & Vmask)) > trV ) )
pattern |= flag;
}
flag <<= 1;
}
for (k=1; k<=9; k++)
c[k] = w[k];
#include "TextureFilters_hq4x.h"
pIn+=2;
pOut+=8;
}
pIn += 2*(SrcPPL-Xres);
pOut+= 8*(SrcPPL-Xres);
pOut+=BpL;
pOut+=BpL;
pOut+=BpL;
}
#undef BPP
#undef BPP2
#undef BPP3
#undef Diff
#undef hq4x_Interp1
#undef hq4x_Interp2
#undef hq4x_Interp3
#undef hq4x_Interp4
#undef hq4x_Interp5
#undef hq4x_Interp6
#undef hq4x_Interp7
#undef hq4x_Interp8
}
#endif /* !_16BPP_HACK */
void hq4x_8888(unsigned char * pIn, unsigned char * pOut, int Xres, int Yres, int SrcPPL, int BpL)
{
#define hq4x_Interp1 hq4x_Interp1_8888
#define hq4x_Interp2 hq4x_Interp2_8888
#define hq4x_Interp3 hq4x_Interp3_8888
#define hq4x_Interp4 hq4x_Interp4_8888
#define hq4x_Interp5 hq4x_Interp5_8888
#define hq4x_Interp6 hq4x_Interp6_8888
#define hq4x_Interp7 hq4x_Interp7_8888
#define hq4x_Interp8 hq4x_Interp8_8888
#define Diff Diff_888
#define BPP 4
#define BPP2 8
#define BPP3 12
int i, j, k;
int prevline, nextline;
uint32 w[10];
uint32 c[10];
int pattern;
int flag;
int YUV1, YUV2;
// +----+----+----+
// | | | |
// | w1 | w2 | w3 |
// +----+----+----+
// | | | |
// | w4 | w5 | w6 |
// +----+----+----+
// | | | |
// | w7 | w8 | w9 |
// +----+----+----+
for (j = 0; j < Yres; j++) {
if (j>0) prevline = -SrcPPL*4; else prevline = 0;
if (j<Yres-1) nextline = SrcPPL*4; else nextline = 0;
for (i=0; i<Xres; i++) {
w[2] = *((uint32*)(pIn + prevline));
w[5] = *((uint32*)pIn);
w[8] = *((uint32*)(pIn + nextline));
if (i>0) {
w[1] = *((uint32*)(pIn + prevline - 4));
w[4] = *((uint32*)(pIn - 4));
w[7] = *((uint32*)(pIn + nextline - 4));
} else {
w[1] = w[2];
w[4] = w[5];
w[7] = w[8];
}
if (i<Xres-1) {
w[3] = *((uint32*)(pIn + prevline + 4));
w[6] = *((uint32*)(pIn + 4));
w[9] = *((uint32*)(pIn + nextline + 4));
} else {
w[3] = w[2];
w[6] = w[5];
w[9] = w[8];
}
pattern = 0;
flag = 1;
YUV1 = RGB888toYUV(w[5]);
for (k=1; k<=9; k++) {
if (k==5) continue;
if ( w[k] != w[5] ) {
YUV2 = RGB888toYUV(w[k]);
if ( ( abs((YUV1 & Ymask) - (YUV2 & Ymask)) > trY ) ||
( abs((YUV1 & Umask) - (YUV2 & Umask)) > trU ) ||
( abs((YUV1 & Vmask) - (YUV2 & Vmask)) > trV ) )
pattern |= flag;
}
flag <<= 1;
}
for (k=1; k<=9; k++)
c[k] = w[k];
#include "TextureFilters_hq4x.h"
pIn+=4;
pOut+=16;
}
pIn += 4*(SrcPPL-Xres);
pOut+= 16*(SrcPPL-Xres);
pOut+=BpL;
pOut+=BpL;
pOut+=BpL;
}
#undef BPP
#undef BPP2
#undef BPP3
#undef Diff
#undef hq4x_Interp1
#undef hq4x_Interp2
#undef hq4x_Interp3
#undef hq4x_Interp4
#undef hq4x_Interp5
#undef hq4x_Interp6
#undef hq4x_Interp7
#undef hq4x_Interp8
}
#if !_16BPP_HACK
void hq4x_init(void)
{
static int done = 0;
int r, g, b, Y, u, v, i, j, k;
if (done ) return;
for (i = 0; i < 16; i++) {
for (j = 0; j < 16; j++) {
for (k = 0; k < 16; k++) {
r = (i << 4) | i;
g = (j << 4) | j;
b = (k << 4) | k;
/* Microsoft's RGB888->YUV conversion */
/*Y = ((( 66 * r + 129 * g + 25 * b + 128) >> 8) + 16) & 0xFF;
u = ((( -38 * r - 74 * g + 112 * b + 128) >> 8) + 128) & 0xFF;
v = ((( 112 * r - 94 * g - 18 * b + 128) >> 8) + 128) & 0xFF;*/
Y = (r + g + b) >> 2;
u = 128 + ((r - b) >> 2);
v = 128 + ((-r + 2*g -b)>>3);
RGB444toYUV[(i << 8) | (j << 4) | k] = (Y << 16) | (u << 8) | v;
}
}
}
done = 1;
}
#endif /* !_16BPP_HACK */

4999
Source/GlideHQ/TextureFilters_hq4x.h Normal file
View File

File diff suppressed because it is too large Load Diff

1307
Source/GlideHQ/TextureFilters_lq2x.h Normal file
View File

File diff suppressed because it is too large Load Diff

433
Source/GlideHQ/TxCache.cpp Normal file
View File

@ -0,0 +1,433 @@
/*
* Texture Filtering
* Version: 1.0
*
* Copyright (C) 2007 Hiroshi Morii All Rights Reserved.
* Email koolsmoky(at)users.sourceforge.net
* Web http://www.3dfxzone.it/koolsmoky
*
* this 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, or (at your option)
* any later version.
*
* this 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 GNU Make; see the file COPYING. If not, write to
* the Free Software Foundation, 675 Mass Ave, Cambridge, MA 02139, USA.
*/
#ifdef WIN32
#pragma warning(disable: 4786)
#endif
#include "TxCache.h"
#include "TxDbg.h"
#include <zlib/zlib.h>
#include <Common/path.h>
TxCache::~TxCache()
{
/* free memory, clean up, etc */
clear();
delete _txUtil;
}
TxCache::TxCache(int options, int cachesize, const wchar_t *path, const wchar_t *ident,
dispInfoFuncExt callback)
{
_txUtil = new TxUtil();
_options = options;
_cacheSize = cachesize;
_callback = callback;
_totalSize = 0;
/* save path name */
if (path)
_path.assign(path);
/* save ROM name */
if (ident)
_ident.assign(ident);
/* zlib memory buffers to (de)compress hires textures */
if (_options & (GZ_TEXCACHE|GZ_HIRESTEXCACHE)) {
_gzdest0 = TxMemBuf::getInstance()->get(0);
_gzdest1 = TxMemBuf::getInstance()->get(1);
_gzdestLen = (TxMemBuf::getInstance()->size_of(0) < TxMemBuf::getInstance()->size_of(1)) ?
TxMemBuf::getInstance()->size_of(0) : TxMemBuf::getInstance()->size_of(1);
if (!_gzdest0 || !_gzdest1 || !_gzdestLen) {
_options &= ~(GZ_TEXCACHE|GZ_HIRESTEXCACHE);
_gzdest0 = NULL;
_gzdest1 = NULL;
_gzdestLen = 0;
}
}
}
boolean
TxCache::add(uint64 checksum, GHQTexInfo *info, int dataSize)
{
/* NOTE: dataSize must be provided if info->data is zlib compressed. */
if (!checksum || !info->data) return 0;
uint8 *dest = info->data;
uint16 format = info->format;
if (!dataSize) {
dataSize = _txUtil->sizeofTx(info->width, info->height, info->format);
if (!dataSize) return 0;
if (_options & (GZ_TEXCACHE|GZ_HIRESTEXCACHE)) {
/* zlib compress it. compression level:1 (best speed) */
uint32 destLen = _gzdestLen;
dest = (dest == _gzdest0) ? _gzdest1 : _gzdest0;
if (compress2(dest, &destLen, info->data, dataSize, 1) != Z_OK) {
dest = info->data;
DBG_INFO(80, L"Error: zlib compression failed!\n");
} else {
DBG_INFO(80, L"zlib compressed: %.02fkb->%.02fkb\n", (float)dataSize/1000, (float)destLen/1000);
dataSize = destLen;
format |= GR_TEXFMT_GZ;
}
}
}
/* if cache size exceeds limit, remove old cache */
if (_cacheSize > 0) {
_totalSize += dataSize;
if ((_totalSize > _cacheSize) && !_cachelist.empty()) {
/* _cachelist is arranged so that frequently used textures are in the back */
std::list<uint64>::iterator itList = _cachelist.begin();
while (itList != _cachelist.end()) {
/* find it in _cache */
std::map<uint64, TXCACHE*>::iterator itMap = _cache.find(*itList);
if (itMap != _cache.end()) {
/* yep we have it. remove it. */
_totalSize -= (*itMap).second->size;
free((*itMap).second->info.data);
delete (*itMap).second;
_cache.erase(itMap);
}
itList++;
/* check if memory cache has enough space */
if (_totalSize <= _cacheSize)
break;
}
/* remove from _cachelist */
_cachelist.erase(_cachelist.begin(), itList);
DBG_INFO(80, L"+++++++++\n");
}
_totalSize -= dataSize;
}
/* cache it */
uint8 *tmpdata = (uint8*)malloc(dataSize);
if (tmpdata) {
TXCACHE *txCache = new TXCACHE;
if (txCache) {
/* we can directly write as we filter, but for now we get away
* with doing memcpy after all the filtering is done.
*/
memcpy(tmpdata, dest, dataSize);
/* copy it */
memcpy(&txCache->info, info, sizeof(GHQTexInfo));
txCache->info.data = tmpdata;
txCache->info.format = format;
txCache->size = dataSize;
/* add to cache */
if (_cacheSize > 0) {
_cachelist.push_back(checksum);
txCache->it = --(_cachelist.end());
}
/* _cache[checksum] = txCache; */
_cache.insert(std::map<uint64, TXCACHE*>::value_type(checksum, txCache));
#ifdef DEBUG
DBG_INFO(80, L"[%5d] added!! crc:%08X %08X %d x %d gfmt:%x total:%.02fmb\n",
_cache.size(), (uint32)(checksum >> 32), (uint32)(checksum & 0xffffffff),
info->width, info->height, info->format, (float)_totalSize/1000000);
DBG_INFO(80, L"smalllodlog2:%d largelodlog2:%d aspectratiolog2:%d\n",
txCache->info.smallLodLog2, txCache->info.largeLodLog2, txCache->info.aspectRatioLog2);
if (info->tiles) {
DBG_INFO(80, L"tiles:%d un-tiled size:%d x %d\n", info->tiles, info->untiled_width, info->untiled_height);
}
if (_cacheSize > 0) {
DBG_INFO(80, L"cache max config:%.02fmb\n", (float)_cacheSize/1000000);
if (_cache.size() != _cachelist.size()) {
DBG_INFO(80, L"Error: cache/cachelist mismatch! (%d/%d)\n", _cache.size(), _cachelist.size());
}
}
#endif
/* total cache size */
_totalSize += dataSize;
return 1;
}
free(tmpdata);
}
return 0;
}
boolean
TxCache::get(uint64 checksum, GHQTexInfo *info)
{
if (!checksum || _cache.empty()) return 0;
/* find a match in cache */
std::map<uint64, TXCACHE*>::iterator itMap = _cache.find(checksum);
if (itMap != _cache.end()) {
/* yep, we've got it. */
memcpy(info, &(((*itMap).second)->info), sizeof(GHQTexInfo));
/* push it to the back of the list */
if (_cacheSize > 0) {
_cachelist.erase(((*itMap).second)->it);
_cachelist.push_back(checksum);
((*itMap).second)->it = --(_cachelist.end());
}
/* zlib decompress it */
if (info->format & GR_TEXFMT_GZ) {
uint32 destLen = _gzdestLen;
uint8 *dest = (_gzdest0 == info->data) ? _gzdest1 : _gzdest0;
if (uncompress(dest, &destLen, info->data, ((*itMap).second)->size) != Z_OK) {
DBG_INFO(80, L"Error: zlib decompression failed!\n");
return 0;
}
info->data = dest;
info->format &= ~GR_TEXFMT_GZ;
DBG_INFO(80, L"zlib decompressed: %.02fkb->%.02fkb\n", (float)(((*itMap).second)->size)/1000, (float)destLen/1000);
}
return 1;
}
return 0;
}
boolean
TxCache::save(const wchar_t *path, const wchar_t *filename, int config)
{
if (!_cache.empty()) {
/* dump cache to disk */
char cbuf[MAX_PATH];
CPath cachepath(stdstr().FromUTF16(path),"");
cachepath.CreateDirectory();
/* Ugly hack to enable fopen/gzopen in Win9x */
#ifdef WIN32
wchar_t curpath[MAX_PATH];
GETCWD(MAX_PATH, curpath);
cachepath.ChangeDirectory();
#else
char curpath[MAX_PATH];
wcstombs(cbuf, cachepath.string().c_str(), MAX_PATH);
GETCWD(MAX_PATH, curpath);
CHDIR(cbuf);
#endif
#ifdef tofix
wcstombs(cbuf, filename, MAX_PATH);
gzFile gzfp = gzopen(cbuf, "wb1");
DBG_INFO(80, L"gzfp:%x file:%ls\n", gzfp, filename);
if (gzfp) {
/* write header to determine config match */
gzwrite(gzfp, &config, 4);
std::map<uint64, TXCACHE*>::iterator itMap = _cache.begin();
while (itMap != _cache.end()) {
uint8 *dest = (*itMap).second->info.data;
uint32 destLen = (*itMap).second->size;
uint16 format = (*itMap).second->info.format;
/* to keep things simple, we save the texture data in a zlib uncompressed state. */
/* sigh... for those who cannot wait the extra few seconds. changed to keep
* texture data in a zlib compressed state. if the GZ_TEXCACHE or GZ_HIRESTEXCACHE
* option is toggled, the cache will need to be rebuilt.
*/
/*if (format & GR_TEXFMT_GZ) {
dest = _gzdest0;
destLen = _gzdestLen;
if (dest && destLen) {
if (uncompress(dest, &destLen, (*itMap).second->info.data, (*itMap).second->size) != Z_OK) {
dest = NULL;
destLen = 0;
}
format &= ~GR_TEXFMT_GZ;
}
}*/
if (dest && destLen) {
/* texture checksum */
gzwrite(gzfp, &((*itMap).first), 8);
/* other texture info */
gzwrite(gzfp, &((*itMap).second->info.width), 4);
gzwrite(gzfp, &((*itMap).second->info.height), 4);
gzwrite(gzfp, &format, 2);
gzwrite(gzfp, &((*itMap).second->info.smallLodLog2), 4);
gzwrite(gzfp, &((*itMap).second->info.largeLodLog2), 4);
gzwrite(gzfp, &((*itMap).second->info.aspectRatioLog2), 4);
gzwrite(gzfp, &((*itMap).second->info.tiles), 4);
gzwrite(gzfp, &((*itMap).second->info.untiled_width), 4);
gzwrite(gzfp, &((*itMap).second->info.untiled_height), 4);
gzwrite(gzfp, &((*itMap).second->info.is_hires_tex), 1);
gzwrite(gzfp, &destLen, 4);
gzwrite(gzfp, dest, destLen);
}
itMap++;
/* not ready yet */
/*if (_callback)
(*_callback)(L"Total textures saved to HDD: %d\n", std::distance(itMap, _cache.begin()));*/
}
gzclose(gzfp);
}
CHDIR(curpath);
#endif
}
return _cache.empty();
}
boolean
TxCache::load(const wchar_t *path, const wchar_t *filename, int config)
{
/* find it on disk */
CPath cbuf(stdstr().FromUTF16(path).c_str(),stdstr().FromUTF16(filename).c_str());
gzFile gzfp = gzopen(cbuf, "rb");
DBG_INFO(80, L"gzfp:%x file:%ls\n", gzfp, filename);
if (gzfp) {
/* yep, we have it. load it into memory cache. */
int dataSize;
uint64 checksum;
GHQTexInfo tmpInfo;
int tmpconfig;
/* read header to determine config match */
gzread(gzfp, &tmpconfig, 4);
if (tmpconfig == config) {
do {
memset(&tmpInfo, 0, sizeof(GHQTexInfo));
gzread(gzfp, &checksum, 8);
gzread(gzfp, &tmpInfo.width, 4);
gzread(gzfp, &tmpInfo.height, 4);
gzread(gzfp, &tmpInfo.format, 2);
gzread(gzfp, &tmpInfo.smallLodLog2, 4);
gzread(gzfp, &tmpInfo.largeLodLog2, 4);
gzread(gzfp, &tmpInfo.aspectRatioLog2, 4);
gzread(gzfp, &tmpInfo.tiles, 4);
gzread(gzfp, &tmpInfo.untiled_width, 4);
gzread(gzfp, &tmpInfo.untiled_height, 4);
gzread(gzfp, &tmpInfo.is_hires_tex, 1);
gzread(gzfp, &dataSize, 4);
tmpInfo.data = (uint8*)malloc(dataSize);
if (tmpInfo.data) {
gzread(gzfp, tmpInfo.data, dataSize);
/* add to memory cache */
add(checksum, &tmpInfo, (tmpInfo.format & GR_TEXFMT_GZ) ? dataSize : 0);
free(tmpInfo.data);
} else {
gzseek(gzfp, dataSize, SEEK_CUR);
}
/* skip in between to prevent the loop from being tied down to vsync */
if (_callback && (!(_cache.size() % 100) || gzeof(gzfp)))
(*_callback)(L"[%d] total mem:%.02fmb - %ls\n", _cache.size(), (float)_totalSize/1000000, filename);
} while (!gzeof(gzfp));
gzclose(gzfp);
}
}
return !_cache.empty();
}
boolean
TxCache::del(uint64 checksum)
{
if (!checksum || _cache.empty()) return 0;
std::map<uint64, TXCACHE*>::iterator itMap = _cache.find(checksum);
if (itMap != _cache.end()) {
/* for texture cache (not hi-res cache) */
if (!_cachelist.empty()) _cachelist.erase(((*itMap).second)->it);
/* remove from cache */
free((*itMap).second->info.data);
_totalSize -= (*itMap).second->size;
delete (*itMap).second;
_cache.erase(itMap);
DBG_INFO(80, L"removed from cache: checksum = %08X %08X\n", (uint32)(checksum & 0xffffffff), (uint32)(checksum >> 32));
return 1;
}
return 0;
}
boolean
TxCache::is_cached(uint64 checksum)
{
std::map<uint64, TXCACHE*>::iterator itMap = _cache.find(checksum);
if (itMap != _cache.end()) return 1;
return 0;
}
void
TxCache::clear()
{
if (!_cache.empty()) {
std::map<uint64, TXCACHE*>::iterator itMap = _cache.begin();
while (itMap != _cache.end()) {
free((*itMap).second->info.data);
delete (*itMap).second;
itMap++;
}
_cache.clear();
}
if (!_cachelist.empty()) _cachelist.clear();
_totalSize = 0;
}

69
Source/GlideHQ/TxCache.h Normal file
View File

@ -0,0 +1,69 @@
/*
* Texture Filtering
* Version: 1.0
*
* Copyright (C) 2007 Hiroshi Morii All Rights Reserved.
* Email koolsmoky(at)users.sourceforge.net
* Web http://www.3dfxzone.it/koolsmoky
*
* this 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, or (at your option)
* any later version.
*
* this 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 GNU Make; see the file COPYING. If not, write to
* the Free Software Foundation, 675 Mass Ave, Cambridge, MA 02139, USA.
*/
#ifndef __TXCACHE_H__
#define __TXCACHE_H__
#include "TxInternal.h"
#include "TxUtil.h"
#include <list>
#include <map>
#include <string>
class TxCache
{
private:
std::list<uint64> _cachelist;
uint8 *_gzdest0;
uint8 *_gzdest1;
uint32 _gzdestLen;
protected:
int _options;
std::wstring _ident;
std::wstring _path;
dispInfoFuncExt _callback;
TxUtil *_txUtil;
struct TXCACHE {
int size;
GHQTexInfo info;
std::list<uint64>::iterator it;
};
int _totalSize;
int _cacheSize;
std::map<uint64, TXCACHE*> _cache;
boolean save(const wchar_t *path, const wchar_t *filename, const int config);
boolean load(const wchar_t *path, const wchar_t *filename, const int config);
boolean del(uint64 checksum); /* checksum hi:palette low:texture */
boolean is_cached(uint64 checksum); /* checksum hi:palette low:texture */
void clear();
public:
~TxCache();
TxCache(int options, int cachesize, const wchar_t *path, const wchar_t *ident,
dispInfoFuncExt callback);
boolean add(uint64 checksum, /* checksum hi:palette low:texture */
GHQTexInfo *info, int dataSize = 0);
boolean get(uint64 checksum, /* checksum hi:palette low:texture */
GHQTexInfo *info);
};
#endif /* __TXCACHE_H__ */

69
Source/GlideHQ/TxDbg.cpp Normal file
View File

@ -0,0 +1,69 @@
/*
* Texture Filtering
* Version: 1.0
*
* Copyright (C) 2007 Hiroshi Morii All Rights Reserved.
* Email koolsmoky(at)users.sourceforge.net
* Web http://www.3dfxzone.it/koolsmoky
*
* this 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, or (at your option)
* any later version.
*
* this 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 GNU Make; see the file COPYING. If not, write to
* the Free Software Foundation, 675 Mass Ave, Cambridge, MA 02139, USA.
*/
#define DBG_LEVEL 80
#include "TxDbg.h"
#include <string.h>
#include <stdarg.h>
#include <Common/std string.h>
#include <Common/path.h>
TxDbg::TxDbg()
{
_level = DBG_LEVEL;
CPath Dir(CPath::MODULE_DIRECTORY,"");
Dir.AppendDirectory("Logs");
if (!_dbgfile)
#ifdef GHQCHK
_dbgfile = fopen(CPath(Dir,"ghqchk.txt"), "w");
#else
_dbgfile = fopen(CPath((LPCSTR)Dir,"glidehq.dbg"), "w");
#endif
}
TxDbg::~TxDbg()
{
if (_dbgfile) {
fclose(_dbgfile);
_dbgfile = 0;
}
_level = DBG_LEVEL;
}
void
TxDbg::output(const int level, const wchar_t *format, ...)
{
if (level > _level)
return;
stdstr_f newformat("%d:\t%s",level,stdstr().FromUTF16(format).c_str());
va_list args;
va_start(args, format);
vfwprintf(_dbgfile, newformat.ToUTF16().c_str(), args);
fflush(_dbgfile);
va_end(args);
}

61
Source/GlideHQ/TxDbg.h Normal file
View File

@ -0,0 +1,61 @@
/*
* Texture Filtering
* Version: 1.0
*
* Copyright (C) 2007 Hiroshi Morii All Rights Reserved.
* Email koolsmoky(at)users.sourceforge.net
* Web http://www.3dfxzone.it/koolsmoky
*
* this 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, or (at your option)
* any later version.
*
* this 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 GNU Make; see the file COPYING. If not, write to
* the Free Software Foundation, 675 Mass Ave, Cambridge, MA 02139, USA.
*/
#ifndef __TXDBG_H__
#define __TXDBG_H__
#include <stdio.h>
#include "TxInternal.h"
class TxDbg
{
private:
FILE* _dbgfile;
int _level;
TxDbg();
public:
static TxDbg* getInstance() {
static TxDbg txDbg;
return &txDbg;
}
~TxDbg();
void output(const int level, const wchar_t *format, ...);
};
#ifdef DEBUG
#define DBG_INFO TxDbg::getInstance()->output
#define INFO DBG_INFO
#else
#define DBG_INFO 0 && (wchar_t)
#ifdef GHQCHK
#define INFO TxDbg::getInstance()->output
#else
#if 0 /* XXX enable this to log basic hires texture checks */
#define INFO TxDbg::getInstance()->output
#else
#define INFO DBG_INFO
#endif
#endif
#endif
#endif /* __TXDBG_H__ */

692
Source/GlideHQ/TxFilter.cpp Normal file
View File

@ -0,0 +1,692 @@
/*
* Texture Filtering
* Version: 1.0
*
* Copyright (C) 2007 Hiroshi Morii All Rights Reserved.
* Email koolsmoky(at)users.sourceforge.net
* Web http://www.3dfxzone.it/koolsmoky
*
* this 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, or (at your option)
* any later version.
*
* this 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 GNU Make; see the file COPYING. If not, write to
* the Free Software Foundation, 675 Mass Ave, Cambridge, MA 02139, USA.
*/
#ifdef WIN32
#pragma warning(disable: 4786)
#endif
#include <common/path.h>
#include "TxFilter.h"
#include "TextureFilters.h"
#include "TxDbg.h"
#include "bldno.h"
void TxFilter::clear()
{
/* clear hires texture cache */
delete _txHiResCache;
/* clear texture cache */
delete _txTexCache;
/* free memory */
TxMemBuf::getInstance()->shutdown();
/* clear other stuff */
delete _txImage;
delete _txQuantize;
delete _txUtil;
}
TxFilter::~TxFilter()
{
clear();
}
TxFilter::TxFilter(int maxwidth, int maxheight, int maxbpp, int options,
int cachesize, wchar_t *path, wchar_t *ident,
dispInfoFuncExt callback) :
_numcore(0),
_tex1(NULL),
_tex2(NULL),
_maxwidth(0),
_maxheight(0),
_maxbpp(0),
_options(0),
_cacheSize(0),
_txQuantize(NULL),
_txTexCache(NULL),
_txHiResCache(NULL),
_txUtil(NULL),
_txImage(NULL),
_initialized(false)
{
/* HACKALERT: the emulator misbehaves and sometimes forgets to shutdown */
if ((ident && wcscmp(ident, L"DEFAULT") != 0 && _ident.compare(ident) == 0) &&
_maxwidth == maxwidth &&
_maxheight == maxheight &&
_maxbpp == maxbpp &&
_options == options &&
_cacheSize == cachesize) return;
clear(); /* gcc does not allow the destructor to be called */
/* shamelessness :P this first call to the debug output message creates
* a file in the executable directory. */
INFO(0, L"------------------------------------------------------------------\n");
#ifdef GHQCHK
INFO(0, L" GlideHQ Hires Texture Checker 1.02.00.%d\n", BUILD_NUMBER);
#else
// INFO(0, L" GlideHQ version 1.02.00.%d\n", BUILD_NUMBER);
#endif
INFO(0, L" Copyright (C) 2010 Hiroshi Morii All Rights Reserved\n");
INFO(0, L" email : koolsmoky(at)users.sourceforge.net\n");
INFO(0, L" website : http://www.3dfxzone.it/koolsmoky\n");
INFO(0, L"\n");
INFO(0, L" Glide64 official website : http://glide64.emuxhaven.net\n");
INFO(0, L"------------------------------------------------------------------\n");
_options = options;
_txImage = new TxImage();
_txQuantize = new TxQuantize();
_txUtil = new TxUtil();
/* get number of CPU cores. */
_numcore = _txUtil->getNumberofProcessors();
_initialized = 0;
_tex1 = NULL;
_tex2 = NULL;
/* XXX: anything larger than 1024 * 1024 is overkill */
_maxwidth = maxwidth > 1024 ? 1024 : maxwidth;
_maxheight = maxheight > 1024 ? 1024 : maxheight;
_maxbpp = maxbpp;
_cacheSize = cachesize;
/* TODO: validate options and do overrides here*/
/* save path name */
if (path)
_path.assign(path);
/* save ROM name */
if (ident && wcscmp(ident, L"DEFAULT") != 0)
_ident.assign(ident);
/* check for dxtn extensions */
if (!TxLoadLib::getInstance()->getdxtCompressTexFuncExt())
_options &= ~S3TC_COMPRESSION;
if (!TxLoadLib::getInstance()->getfxtCompressTexFuncExt())
_options &= ~FXT1_COMPRESSION;
switch (options & COMPRESSION_MASK) {
case FXT1_COMPRESSION:
case S3TC_COMPRESSION:
break;
case NCC_COMPRESSION:
default:
_options &= ~COMPRESSION_MASK;
}
if (TxMemBuf::getInstance()->init(_maxwidth, _maxheight)) {
if (!_tex1)
_tex1 = TxMemBuf::getInstance()->get(0);
if (!_tex2)
_tex2 = TxMemBuf::getInstance()->get(1);
}
#if !_16BPP_HACK
/* initialize hq4x filter */
hq4x_init();
#endif
/* initialize texture cache in bytes. 128Mb will do nicely in most cases */
_txTexCache = new TxTexCache(_options, _cacheSize, _path.c_str(), _ident.c_str(), callback);
/* hires texture */
#if HIRES_TEXTURE
_txHiResCache = new TxHiResCache(_maxwidth, _maxheight, _maxbpp, _options, _path.c_str(), _ident.c_str(), callback);
if (_txHiResCache->empty())
_options &= ~HIRESTEXTURES_MASK;
#endif
if (!(_options & COMPRESS_TEX))
_options &= ~COMPRESSION_MASK;
if (_tex1 && _tex2)
_initialized = 1;
}
boolean
TxFilter::filter(uint8 *src, int srcwidth, int srcheight, uint16 srcformat, uint64 g64crc, GHQTexInfo *info)
{
uint8 *texture = src;
uint8 *tmptex = _tex1;
uint16 destformat = srcformat;
/* We need to be initialized first! */
if (!_initialized) return 0;
/* find cached textures */
if (_cacheSize) {
/* calculate checksum of source texture */
if (!g64crc)
g64crc = (uint64)(_txUtil->checksumTx(texture, srcwidth, srcheight, srcformat));
DBG_INFO(80, L"filter: crc:%08X %08X %d x %d gfmt:%x\n",
(uint32)(g64crc >> 32), (uint32)(g64crc & 0xffffffff), srcwidth, srcheight, srcformat);
#if 0 /* use hirestex to retrieve cached textures. */
/* check if we have it in cache */
if (!(g64crc & 0xffffffff00000000) && /* we reach here only when there is no hires texture for this crc */
_txTexCache->get(g64crc, info)) {
DBG_INFO(80, L"cache hit: %d x %d gfmt:%x\n", info->width, info->height, info->format);
return 1; /* yep, we've got it */
}
#endif
}
/* Leave small textures alone because filtering makes little difference.
* Moreover, some filters require at least 4 * 4 to work.
* Bypass _options to do ARGB8888->16bpp if _maxbpp=16 or forced color reduction.
*/
if ((srcwidth >= 4 && srcheight >= 4) &&
((_options & (FILTER_MASK|ENHANCEMENT_MASK|COMPRESSION_MASK)) ||
(srcformat == GR_TEXFMT_ARGB_8888 && (_maxbpp < 32 || _options & FORCE16BPP_TEX)))) {
#if !_16BPP_HACK
/* convert textures to a format that the compressor accepts (ARGB8888) */
if (_options & COMPRESSION_MASK) {
#endif
if (srcformat != GR_TEXFMT_ARGB_8888) {
if (!_txQuantize->quantize(texture, tmptex, srcwidth, srcheight, srcformat, GR_TEXFMT_ARGB_8888)) {
DBG_INFO(80, L"Error: unsupported format! gfmt:%x\n", srcformat);
return 0;
}
texture = tmptex;
destformat = GR_TEXFMT_ARGB_8888;
}
#if !_16BPP_HACK
}
#endif
switch (destformat) {
case GR_TEXFMT_ARGB_8888:
/*
* prepare texture enhancements (x2, x4 scalers)
*/
int scale_shift = 0, num_filters = 0;
uint32 filter = 0;
if ((_options & ENHANCEMENT_MASK) == HQ4X_ENHANCEMENT) {
if (srcwidth <= (_maxwidth >> 2) && srcheight <= (_maxheight >> 2)) {
filter |= HQ4X_ENHANCEMENT;
scale_shift = 2;
num_filters++;
} else if (srcwidth <= (_maxwidth >> 1) && srcheight <= (_maxheight >> 1)) {
filter |= HQ2X_ENHANCEMENT;
scale_shift = 1;
num_filters++;
}
} else if (_options & ENHANCEMENT_MASK) {
if (srcwidth <= (_maxwidth >> 1) && srcheight <= (_maxheight >> 1)) {
filter |= (_options & ENHANCEMENT_MASK);
scale_shift = 1;
num_filters++;
}
}
/*
* prepare texture filters
*/
if (_options & (SMOOTH_FILTER_MASK|SHARP_FILTER_MASK)) {
filter |= (_options & (SMOOTH_FILTER_MASK|SHARP_FILTER_MASK));
num_filters++;
}
/*
* execute texture enhancements and filters
*/
while (num_filters > 0) {
tmptex = (texture == _tex1) ? _tex2 : _tex1;
uint8 *_texture = texture;
uint8 *_tmptex = tmptex;
unsigned int numcore = _numcore;
unsigned int blkrow = 0;
while (numcore > 1 && blkrow == 0) {
blkrow = (srcheight >> 2) / numcore;
numcore--;
}
if (blkrow > 0 && numcore > 1) {
#ifdef tofix
boost::thread *thrd[MAX_NUMCORE];
unsigned int i;
int blkheight = blkrow << 2;
unsigned int srcStride = (srcwidth * blkheight) << 2;
unsigned int destStride = srcStride << scale_shift << scale_shift;
for (i = 0; i < numcore - 1; i++) {
thrd[i] = new boost::thread(boost::bind(filter_8888,
(uint32*)_texture,
srcwidth,
blkheight,
(uint32*)_tmptex,
filter));
_texture += srcStride;
_tmptex += destStride;
}
thrd[i] = new boost::thread(boost::bind(filter_8888,
(uint32*)_texture,
srcwidth,
srcheight - blkheight * i,
(uint32*)_tmptex,
filter));
for (i = 0; i < numcore; i++) {
thrd[i]->join();
delete thrd[i];
}
#endif
} else {
filter_8888((uint32*)_texture, srcwidth, srcheight, (uint32*)_tmptex, filter);
}
if (filter & ENHANCEMENT_MASK) {
srcwidth <<= scale_shift;
srcheight <<= scale_shift;
filter &= ~ENHANCEMENT_MASK;
scale_shift = 0;
}
texture = tmptex;
num_filters--;
}
/*
* texture compression
*/
/* ignored if we only have texture compression option on.
* only done when texture enhancer is used. see constructor. */
if ((_options & COMPRESSION_MASK) &&
(srcwidth >= 64 && srcheight >= 64) /* Texture compression is not suitable for low pixel coarse detail
* textures. The assumption here is that textures larger than 64x64
* have enough detail to produce decent quality when compressed. The
* down side is that narrow stripped textures that the N64 often use
* for large background textures are also ignored. It would be more
* reasonable if decisions are made based on fourier-transform
* spectrum or RMS error.
*/
) {
int compressionType = _options & COMPRESSION_MASK;
int tmpwidth, tmpheight;
uint16 tmpformat;
/* XXX: textures that use 8bit alpha channel look bad with the current
* fxt1 library, so we substitute it with dxtn for now. afaik all gfx
* cards that support fxt1 also support dxtn. (3dfx and Intel) */
if ((destformat == GR_TEXFMT_ALPHA_INTENSITY_88) ||
(destformat == GR_TEXFMT_ARGB_8888) ||
(destformat == GR_TEXFMT_ALPHA_8)) {
compressionType = S3TC_COMPRESSION;
}
tmptex = (texture == _tex1) ? _tex2 : _tex1;
if (_txQuantize->compress(texture, tmptex,
srcwidth, srcheight, srcformat,
&tmpwidth, &tmpheight, &tmpformat,
compressionType)) {
srcwidth = tmpwidth;
srcheight = tmpheight;
destformat = tmpformat;
texture = tmptex;
}
}
/*
* texture (re)conversions
*/
if (destformat == GR_TEXFMT_ARGB_8888) {
if (srcformat == GR_TEXFMT_ARGB_8888 && (_maxbpp < 32 || _options & FORCE16BPP_TEX)) srcformat = GR_TEXFMT_ARGB_4444;
if (srcformat != GR_TEXFMT_ARGB_8888) {
tmptex = (texture == _tex1) ? _tex2 : _tex1;
if (!_txQuantize->quantize(texture, tmptex, srcwidth, srcheight, GR_TEXFMT_ARGB_8888, srcformat)) {
DBG_INFO(80, L"Error: unsupported format! gfmt:%x\n", srcformat);
return 0;
}
texture = tmptex;
destformat = srcformat;
}
}
break;
#if !_16BPP_HACK
case GR_TEXFMT_ARGB_4444:
int scale_shift = 0;
tmptex = (texture == _tex1) ? _tex2 : _tex1;
switch (_options & ENHANCEMENT_MASK) {
case HQ4X_ENHANCEMENT:
if (srcwidth <= (_maxwidth >> 2) && srcheight <= (_maxheight >> 2)) {
hq4x_4444((uint8*)texture, (uint8*)tmptex, srcwidth, srcheight, srcwidth, srcwidth * 4 * 2);
scale_shift = 2;
}/* else if (srcwidth <= (_maxwidth >> 1) && srcheight <= (_maxheight >> 1)) {
hq2x_16((uint8*)texture, srcwidth * 2, (uint8*)tmptex, srcwidth * 2 * 2, srcwidth, srcheight);
scale_shift = 1;
}*/
break;
case HQ2X_ENHANCEMENT:
if (srcwidth <= (_maxwidth >> 1) && srcheight <= (_maxheight >> 1)) {
hq2x_16((uint8*)texture, srcwidth * 2, (uint8*)tmptex, srcwidth * 2 * 2, srcwidth, srcheight);
scale_shift = 1;
}
break;
case HQ2XS_ENHANCEMENT:
if (srcwidth <= (_maxwidth >> 1) && srcheight <= (_maxheight >> 1)) {
hq2xS_16((uint8*)texture, srcwidth * 2, (uint8*)tmptex, srcwidth * 2 * 2, srcwidth, srcheight);
scale_shift = 1;
}
break;
case LQ2X_ENHANCEMENT:
if (srcwidth <= (_maxwidth >> 1) && srcheight <= (_maxheight >> 1)) {
lq2x_16((uint8*)texture, srcwidth * 2, (uint8*)tmptex, srcwidth * 2 * 2, srcwidth, srcheight);
scale_shift = 1;
}
break;
case LQ2XS_ENHANCEMENT:
if (srcwidth <= (_maxwidth >> 1) && srcheight <= (_maxheight >> 1)) {
lq2xS_16((uint8*)texture, srcwidth * 2, (uint8*)tmptex, srcwidth * 2 * 2, srcwidth, srcheight);
scale_shift = 1;
}
break;
case X2SAI_ENHANCEMENT:
if (srcwidth <= (_maxwidth >> 1) && srcheight <= (_maxheight >> 1)) {
Super2xSaI_4444((uint16*)texture, (uint16*)tmptex, srcwidth, srcheight, srcwidth);
scale_shift = 1;
}
break;
case X2_ENHANCEMENT:
if (srcwidth <= (_maxwidth >> 1) && srcheight <= (_maxheight >> 1)) {
Texture2x_16((uint8*)texture, srcwidth * 2, (uint8*)tmptex, srcwidth * 2 * 2, srcwidth, srcheight);
scale_shift = 1;
}
}
if (scale_shift) {
srcwidth <<= scale_shift;
srcheight <<= scale_shift;
texture = tmptex;
}
if (_options & SMOOTH_FILTER_MASK) {
tmptex = (texture == _tex1) ? _tex2 : _tex1;
SmoothFilter_4444((uint16*)texture, srcwidth, srcheight, (uint16*)tmptex, (_options & SMOOTH_FILTER_MASK));
texture = tmptex;
} else if (_options & SHARP_FILTER_MASK) {
tmptex = (texture == _tex1) ? _tex2 : _tex1;
SharpFilter_4444((uint16*)texture, srcwidth, srcheight, (uint16*)tmptex, (_options & SHARP_FILTER_MASK));
texture = tmptex;
}
break;
case GR_TEXFMT_ARGB_1555:
break;
case GR_TEXFMT_RGB_565:
break;
case GR_TEXFMT_ALPHA_8:
break;
#endif /* _16BPP_HACK */
}
}
/* fill in the texture info. */
info->data = texture;
info->width = srcwidth;
info->height = srcheight;
info->format = destformat;
info->smallLodLog2 = _txUtil->grLodLog2(srcwidth, srcheight);
info->largeLodLog2 = info->smallLodLog2;
info->aspectRatioLog2 = _txUtil->grAspectRatioLog2(srcwidth, srcheight);
info->is_hires_tex = 0;
/* cache the texture. */
if (_cacheSize) _txTexCache->add(g64crc, info);
DBG_INFO(80, L"filtered texture: %d x %d gfmt:%x\n", info->width, info->height, info->format);
return 1;
}
boolean
TxFilter::hirestex(uint64 g64crc, uint64 r_crc64, uint16 *palette, GHQTexInfo *info)
{
/* NOTE: Rice CRC32 sometimes return the same value for different textures.
* As a workaround, Glide64 CRC32 is used for the key for NON-hires
* texture cache.
*
* r_crc64 = hi:palette low:texture
* (separate crc. doesn't necessary have to be rice crc)
* g64crc = texture + palette glide64 crc32
* (can be any other crc if robust)
*/
DBG_INFO(80, L"hirestex: r_crc64:%08X %08X, g64crc:%08X %08X\n",
(uint32)(r_crc64 >> 32), (uint32)(r_crc64 & 0xffffffff),
(uint32)(g64crc >> 32), (uint32)(g64crc & 0xffffffff));
#if HIRES_TEXTURE
/* check if we have it in hires memory cache. */
if ((_options & HIRESTEXTURES_MASK) && r_crc64) {
if (_txHiResCache->get(r_crc64, info)) {
DBG_INFO(80, L"hires hit: %d x %d gfmt:%x\n", info->width, info->height, info->format);
/* TODO: Enable emulation for special N64 combiner modes. There are few ways
* to get this done. Also applies for CI textures below.
*
* Solution 1. Load the hiresolution textures in ARGB8888 (or A8, IA88) format
* to cache. When a cache is hit, then we take the modes passed in from Glide64
* (also TODO) and apply the modification. Then we do color reduction or format
* conversion or compression if desired and stuff it into the non-hires texture
* cache.
*
* Solution 2. When a cache is hit and if the combiner modes are present,
* convert the texture to ARGB4444 and pass it back to Glide64 to process.
* If a texture is compressed, it needs to be decompressed first. Then add
* the processed texture to the non-hires texture cache.
*
* Solution 3. Hybrid of the above 2. Load the textures in ARGB8888 (A8, IA88)
* format. Convert the texture to ARGB4444 and pass it back to Glide64 when
* the combiner modes are present. Get the processed texture back from Glide64
* and compress if desired and add it to the non-hires texture cache.
*
* Solution 4. Take the easy way out and forget about this whole thing.
*/
return 1; /* yep, got it */
}
if (_txHiResCache->get((r_crc64 & 0xffffffff), info)) {
DBG_INFO(80, L"hires hit: %d x %d gfmt:%x\n", info->width, info->height, info->format);
/* for true CI textures, we use the passed in palette to convert to
* ARGB1555 and add it to memory cache.
*
* NOTE: we do this AFTER all other texture cache searches because
* only a few texture packs actually use true CI textures.
*
* NOTE: the pre-converted palette from Glide64 is in RGBA5551 format.
* A comp comes before RGB comp.
*/
if (palette && info->format == GR_TEXFMT_P_8) {
DBG_INFO(80, L"found GR_TEXFMT_P_8 format. Need conversion!!\n");
int width = info->width;
int height = info->height;
uint16 format = info->format;
/* XXX: avoid collision with zlib compression buffer in TxHiResTexture::get */
uint8 *texture = info->data;
uint8 *tmptex = (texture == _tex1) ? _tex2 : _tex1;
/* use palette and convert to 16bit format */
_txQuantize->P8_16BPP((uint32*)texture, (uint32*)tmptex, info->width, info->height, (uint32*)palette);
texture = tmptex;
format = GR_TEXFMT_ARGB_1555;
#if 1
/* XXX: compressed if memory cache compression is ON */
if (_options & COMPRESSION_MASK) {
tmptex = (texture == _tex1) ? _tex2 : _tex1;
if (_txQuantize->quantize(texture, tmptex, info->width, info->height, format, GR_TEXFMT_ARGB_8888)) {
texture = tmptex;
format = GR_TEXFMT_ARGB_8888;
}
if (format == GR_TEXFMT_ARGB_8888) {
tmptex = (texture == _tex1) ? _tex2 : _tex1;
if (_txQuantize->compress(texture, tmptex,
info->width, info->height, GR_TEXFMT_ARGB_1555,
&width, &height, &format,
_options & COMPRESSION_MASK)) {
texture = tmptex;
} else {
/*if (!_txQuantize->quantize(texture, tmptex, info->width, info->height, GR_TEXFMT_ARGB_8888, GR_TEXFMT_ARGB_1555)) {
DBG_INFO(80, L"Error: unsupported format! gfmt:%x\n", format);
return 0;
}*/
texture = tmptex;
format = GR_TEXFMT_ARGB_1555;
}
}
}
#endif
/* fill in the required info to return */
info->data = texture;
info->width = width;
info->height = height;
info->format = format;
info->smallLodLog2 = _txUtil->grLodLog2(width, height);
info->largeLodLog2 = info->smallLodLog2;
info->aspectRatioLog2 = _txUtil->grAspectRatioLog2(width, height);
info->is_hires_tex = 1;
/* XXX: add to hires texture cache!!! */
_txHiResCache->add(r_crc64, info);
DBG_INFO(80, L"GR_TEXFMT_P_8 loaded as gfmt:%x!\n", format);
}
return 1;
}
}
#endif
/* check if we have it in memory cache */
if (_cacheSize && g64crc) {
if (_txTexCache->get(g64crc, info)) {
DBG_INFO(80, L"cache hit: %d x %d gfmt:%x\n", info->width, info->height, info->format);
return 1; /* yep, we've got it */
}
}
DBG_INFO(80, L"no cache hits.\n");
return 0;
}
uint64
TxFilter::checksum64(uint8 *src, int width, int height, int size, int rowStride, uint8 *palette)
{
if (_options & (HIRESTEXTURES_MASK|DUMP_TEX))
return _txUtil->checksum64(src, width, height, size, rowStride, palette);
return 0;
}
boolean
TxFilter::dmptx(uint8 *src, int width, int height, int rowStridePixel, uint16 gfmt, uint16 n64fmt, uint64 r_crc64)
{
if (!_initialized)
return 0;
if (!(_options & DUMP_TEX))
return 0;
DBG_INFO(80, L"gfmt = %02x n64fmt = %02x\n", gfmt, n64fmt);
DBG_INFO(80, L"hirestex: r_crc64:%08X %08X\n",
(uint32)(r_crc64 >> 32), (uint32)(r_crc64 & 0xffffffff));
if (!_txQuantize->quantize(src, _tex1, rowStridePixel, height, (gfmt & 0x00ff), GR_TEXFMT_ARGB_8888))
return 0;
src = _tex1;
if (!_path.empty() && !_ident.empty()) {
/* dump it to disk */
FILE *fp = NULL;
CPath tmpbuf(stdstr().FromUTF16(_path.c_str()).c_str(),"");
/* create directories */
tmpbuf.AppendDirectory("texture_dump");
if (!tmpbuf.DirectoryExists() && !tmpbuf.CreateDirectory())
return 0;
tmpbuf.AppendDirectory(stdstr().FromUTF16(_ident.c_str()).c_str());
if (!tmpbuf.DirectoryExists() && !tmpbuf.CreateDirectory())
return 0;
tmpbuf.AppendDirectory("GlideHQ");
if (!tmpbuf.DirectoryExists() && !tmpbuf.CreateDirectory())
return 0;
if ((n64fmt >> 8) == 0x2) {
tmpbuf.SetNameExtension(stdstr_f("%ls#%08X#%01X#%01X#%08X_ciByRGBA.png",_ident.c_str(),(uint32)(r_crc64 & 0xffffffff), (n64fmt >> 8),(n64fmt & 0xf),(uint32)(r_crc64 >> 32)).c_str());
} else {
tmpbuf.SetNameExtension(stdstr_f("%ls#%08X#%01X#%01X_all.png",_ident.c_str(),(uint32)(r_crc64 & 0xffffffff),(n64fmt >> 8),(n64fmt & 0xf)).c_str());
}
#ifdef WIN32
if ((fp = fopen(tmpbuf, "wb")) != NULL) {
#else
char cbuf[MAX_PATH];
wcstombs(cbuf, tmpbuf.c_str(), MAX_PATH);
if ((fp = fopen(cbuf, "wb")) != NULL) {
#endif
_txImage->writePNG(src, fp, width, height, (rowStridePixel << 2), 0x0003, 0);
fclose(fp);
return 1;
}
}
return 0;
}
boolean
TxFilter::reloadhirestex()
{
DBG_INFO(80, L"Reload hires textures from texture pack.\n");
if (_txHiResCache->load(0)) {
if (_txHiResCache->empty()) _options &= ~HIRESTEXTURES_MASK;
else _options |= HIRESTEXTURES_MASK;
return 1;
}
return 0;
}

81
Source/GlideHQ/TxFilter.h Normal file
View File

@ -0,0 +1,81 @@
/*
* Texture Filtering
* Version: 1.0
*
* Copyright (C) 2007 Hiroshi Morii All Rights Reserved.
* Email koolsmoky(at)users.sourceforge.net
* Web http://www.3dfxzone.it/koolsmoky
*
* this 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, or (at your option)
* any later version.
*
* this 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 GNU Make; see the file COPYING. If not, write to
* the Free Software Foundation, 675 Mass Ave, Cambridge, MA 02139, USA.
*/
#ifndef __TXFILTER_H__
#define __TXFILTER_H__
#include "TxInternal.h"
#include "TxQuantize.h"
#include "TxHiResCache.h"
#include "TxTexCache.h"
#include "TxUtil.h"
#include "TxImage.h"
#include <string>
class TxFilter
{
private:
int _numcore;
uint8 *_tex1;
uint8 *_tex2;
int _maxwidth;
int _maxheight;
int _maxbpp;
int _options;
int _cacheSize;
std::wstring _ident;
std::wstring _path;
TxQuantize *_txQuantize;
TxTexCache *_txTexCache;
TxHiResCache *_txHiResCache;
TxUtil *_txUtil;
TxImage *_txImage;
boolean _initialized;
void clear();
public:
~TxFilter();
TxFilter(int maxwidth,
int maxheight,
int maxbpp,
int options,
int cachesize,
wchar_t *path,
wchar_t *ident,
dispInfoFuncExt callback);
boolean filter(uint8 *src,
int srcwidth,
int srcheight,
uint16 srcformat,
uint64 g64crc, /* glide64 crc, 64bit for future use */
GHQTexInfo *info);
boolean hirestex(uint64 g64crc, /* glide64 crc, 64bit for future use */
uint64 r_crc64, /* checksum hi:palette low:texture */
uint16 *palette,
GHQTexInfo *info);
uint64 checksum64(uint8 *src, int width, int height, int size, int rowStride, uint8 *palette);
boolean dmptx(uint8 *src, int width, int height, int rowStridePixel, uint16 gfmt, uint16 n64fmt, uint64 r_crc64);
boolean reloadhirestex();
};
#endif /* __TXFILTER_H__ */

105
Source/GlideHQ/TxFilterExport.cpp Normal file
View File

@ -0,0 +1,105 @@
/*
* Texture Filtering
* Version: 1.0
*
* Copyright (C) 2007 Hiroshi Morii All Rights Reserved.
* Email koolsmoky(at)users.sourceforge.net
* Web http://www.3dfxzone.it/koolsmoky
*
* this 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, or (at your option)
* any later version.
*
* this 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 GNU Make; see the file COPYING. If not, write to
* the Free Software Foundation, 675 Mass Ave, Cambridge, MA 02139, USA.
*/
#ifdef WIN32
#pragma warning(disable: 4786)
#endif
#include "TxFilter.h"
TxFilter *txFilter = NULL;
#ifdef __cplusplus
extern "C"{
#endif
TAPI boolean TAPIENTRY
txfilter_init(int maxwidth, int maxheight, int maxbpp, int options, int cachesize,
wchar_t *path, wchar_t*ident,
dispInfoFuncExt callback)
{
if (txFilter) return 0;
txFilter = new TxFilter(maxwidth, maxheight, maxbpp, options, cachesize,
path, ident, callback);
return (txFilter ? 1 : 0);
}
void txfilter_shutdown(void)
{
if (txFilter) delete txFilter;
txFilter = NULL;
}
TAPI boolean TAPIENTRY
txfilter(uint8 *src, int srcwidth, int srcheight, uint16 srcformat,
uint64 g64crc, GHQTexInfo *info)
{
if (txFilter)
return txFilter->filter(src, srcwidth, srcheight, srcformat,
g64crc, info);
return 0;
}
TAPI boolean TAPIENTRY
txfilter_hirestex(uint64 g64crc, uint64 r_crc64, uint16 *palette, GHQTexInfo *info)
{
if (txFilter)
return txFilter->hirestex(g64crc, r_crc64, palette, info);
return 0;
}
TAPI uint64 TAPIENTRY
txfilter_checksum(uint8 *src, int width, int height, int size, int rowStride, uint8 *palette)
{
if (txFilter)
return txFilter->checksum64(src, width, height, size, rowStride, palette);
return 0;
}
TAPI boolean TAPIENTRY
txfilter_dmptx(uint8 *src, int width, int height, int rowStridePixel, uint16 gfmt, uint16 n64fmt, uint64 r_crc64)
{
if (txFilter)
return txFilter->dmptx(src, width, height, rowStridePixel, gfmt, n64fmt, r_crc64);
return 0;
}
TAPI boolean TAPIENTRY
txfilter_reloadhirestex()
{
if (txFilter)
return txFilter->reloadhirestex();
return 0;
}
#ifdef __cplusplus
}
#endif

1084
Source/GlideHQ/TxHiResCache.cpp Normal file
View File

File diff suppressed because it is too large Load Diff

59
Source/GlideHQ/TxHiResCache.h Normal file
View File

@ -0,0 +1,59 @@
/*
* Texture Filtering
* Version: 1.0
*
* Copyright (C) 2007 Hiroshi Morii All Rights Reserved.
* Email koolsmoky(at)users.sourceforge.net
* Web http://www.3dfxzone.it/koolsmoky
*
* this 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, or (at your option)
* any later version.
*
* this 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 GNU Make; see the file COPYING. If not, write to
* the Free Software Foundation, 675 Mass Ave, Cambridge, MA 02139, USA.
*/
#ifndef __TXHIRESCACHE_H__
#define __TXHIRESCACHE_H__
/* support hires textures
* 0: disable
* 1: enable
*/
#define HIRES_TEXTURE 1
#include "TxCache.h"
#include "TxQuantize.h"
#include "TxImage.h"
#include "TxReSample.h"
class TxHiResCache : public TxCache
{
private:
int _maxwidth;
int _maxheight;
int _maxbpp;
boolean _haveCache;
boolean _abortLoad;
TxImage *_txImage;
TxQuantize *_txQuantize;
TxReSample *_txReSample;
boolean loadHiResTextures(LPCSTR dir_path, boolean replace);
public:
~TxHiResCache();
TxHiResCache(int maxwidth, int maxheight, int maxbpp, int options,
const wchar_t *path, const wchar_t *ident,
dispInfoFuncExt callback);
boolean empty();
boolean load(boolean replace);
};
#endif /* __TXHIRESCACHE_H__ */

799
Source/GlideHQ/TxImage.cpp Normal file
View File

@ -0,0 +1,799 @@
/*
* Texture Filtering
* Version: 1.0
*
* Copyright (C) 2007 Hiroshi Morii All Rights Reserved.
* Email koolsmoky(at)users.sourceforge.net
* Web http://www.3dfxzone.it/koolsmoky
*
* this 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, or (at your option)
* any later version.
*
* this 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 GNU Make; see the file COPYING. If not, write to
* the Free Software Foundation, 675 Mass Ave, Cambridge, MA 02139, USA.
*/
/* use power of 2 texture size
* (0:disable, 1:enable, 2:3dfx) */
#define POW2_TEXTURES 0
/* check 8 bytes. use a larger value if needed. */
#define PNG_CHK_BYTES 8
#include "TxImage.h"
#include "TxReSample.h"
#include "TxDbg.h"
#include <stdlib.h>
boolean
TxImage::getPNGInfo(FILE *fp, png_structp *png_ptr, png_infop *info_ptr)
{
unsigned char sig[PNG_CHK_BYTES];
/* check for valid file pointer */
if (!fp)
return 0;
/* check if file is PNG */
if (fread(sig, 1, PNG_CHK_BYTES, fp) != PNG_CHK_BYTES)
return 0;
if (png_sig_cmp(sig, 0, PNG_CHK_BYTES) != 0)
return 0;
/* get PNG file info */
*png_ptr = png_create_read_struct(PNG_LIBPNG_VER_STRING, NULL, NULL, NULL);
if (!*png_ptr)
return 0;
*info_ptr = png_create_info_struct(*png_ptr);
if (!*info_ptr) {
png_destroy_read_struct(png_ptr, NULL, NULL);
return 0;
}
if (setjmp(png_jmpbuf(*png_ptr))) {
DBG_INFO(80, L"error reading png!\n");
png_destroy_read_struct(png_ptr, info_ptr, NULL);
return 0;
}
png_init_io(*png_ptr, fp);
png_set_sig_bytes(*png_ptr, PNG_CHK_BYTES);
png_read_info(*png_ptr, *info_ptr);
return 1;
}
uint8*
TxImage::readPNG(FILE* fp, int* width, int* height, uint16* format)
{
/* NOTE: returned image format is GR_TEXFMT_ARGB_8888 */
png_structp png_ptr;
png_infop info_ptr;
uint8 *image = NULL;
int bit_depth, color_type, interlace_type, compression_type, filter_type,
row_bytes, o_width, o_height, num_pas;
/* initialize */
*width = 0;
*height = 0;
*format = 0;
/* check if we have a valid png file */
if (!fp)
return NULL;
if (!getPNGInfo(fp, &png_ptr, &info_ptr)) {
INFO(80, L"error reading png file! png image is corrupt.\n");
return NULL;
}
png_get_IHDR(png_ptr, info_ptr,
(png_uint_32*)&o_width, (png_uint_32*)&o_height, &bit_depth, &color_type,
&interlace_type, &compression_type, &filter_type);
DBG_INFO(80, L"png format %d x %d bitdepth:%d color:%x interlace:%x compression:%x filter:%x\n",
o_width, o_height, bit_depth, color_type,
interlace_type, compression_type, filter_type);
/* transformations */
/* Rice hi-res textures
* _all.png
* _rgb.png, _a.png
* _ciByRGBA.png
* _allciByRGBA.png
*/
/* strip if color channel is larger than 8 bits */
if (bit_depth > 8) {
png_set_strip_16(png_ptr);
bit_depth = 8;
}
#if 1
/* These are not really required per Rice format spec,
* but is done just in case someone uses them.
*/
/* convert palette color to rgb color */
if (color_type == PNG_COLOR_TYPE_PALETTE) {
png_set_palette_to_rgb(png_ptr);
color_type = PNG_COLOR_TYPE_RGB;
}
/* expand 1,2,4 bit gray scale to 8 bit gray scale */
if (color_type == PNG_COLOR_TYPE_GRAY && bit_depth < 8)
png_set_expand_gray_1_2_4_to_8(png_ptr);
/* convert gray scale or gray scale + alpha to rgb color */
if (color_type == PNG_COLOR_TYPE_GRAY ||
color_type == PNG_COLOR_TYPE_GRAY_ALPHA) {
png_set_gray_to_rgb(png_ptr);
color_type = PNG_COLOR_TYPE_RGB;
}
#endif
/* add alpha channel if any */
if (png_get_valid(png_ptr, info_ptr, PNG_INFO_tRNS)) {
png_set_tRNS_to_alpha(png_ptr);
color_type = PNG_COLOR_TYPE_RGB_ALPHA;
}
/* convert rgb to rgba */
if (color_type == PNG_COLOR_TYPE_RGB) {
png_set_filler(png_ptr, 0xff, PNG_FILLER_AFTER);
color_type = PNG_COLOR_TYPE_RGB_ALPHA;
}
/* punt invalid formats */
if (color_type != PNG_COLOR_TYPE_RGB_ALPHA) {
png_destroy_read_struct(&png_ptr, &info_ptr, NULL);
DBG_INFO(80, L"Error: not PNG_COLOR_TYPE_RGB_ALPHA format!\n");
return NULL;
}
/*png_color_8p sig_bit;
if (png_get_sBIT(png_ptr, info_ptr, &sig_bit))
png_set_shift(png_ptr, sig_bit);*/
/* convert rgba to bgra */
png_set_bgr(png_ptr);
/* turn on interlace handling to cope with the weirdness
* of texture authors using interlaced format */
num_pas = png_set_interlace_handling(png_ptr);
/* update info structure */
png_read_update_info(png_ptr, info_ptr);
/* we only get here if ARGB8888 */
row_bytes = png_get_rowbytes(png_ptr, info_ptr);
/* allocate memory to read in image */
image = (uint8*)malloc(row_bytes * o_height);
/* read in image */
if (image) {
int pas, i;
uint8* tmpimage;
for (pas = 0; pas < num_pas; pas++) { /* deal with interlacing */
tmpimage = image;
for (i = 0; i < o_height; i++) {
/* copy row */
png_read_rows(png_ptr, &tmpimage, NULL, 1);
tmpimage += row_bytes;
}
}
/* read rest of the info structure */
png_read_end(png_ptr, info_ptr);
*width = (row_bytes >> 2);
*height = o_height;
*format = GR_TEXFMT_ARGB_8888;
#if POW2_TEXTURES
/* next power of 2 size conversions */
/* NOTE: I can do this in the above loop for faster operations, but some
* texture packs require a workaround. see HACKALERT in nextPow2().
*/
TxReSample txReSample = new TxReSample; // XXX: temporary. move to a better place.
#if (POW2_TEXTURES == 2)
if (!txReSample->nextPow2(&image, width, height, 32, 1)) {
#else
if (!txReSample->nextPow2(&image, width, height, 32, 0)) {
#endif
if (image) {
free(image);
image = NULL;
}
*width = 0;
*height = 0;
*format = 0;
}
delete txReSample;
#endif /* POW2_TEXTURES */
}
/* clean up */
png_destroy_read_struct(&png_ptr, &info_ptr, NULL);
#ifdef DEBUG
if (!image) {
DBG_INFO(80, L"Error: failed to load png image!\n");
}
#endif
return image;
}
boolean
TxImage::writePNG(uint8* src, FILE* fp, int width, int height, int rowStride, uint16 format, uint8 *palette)
{
png_structp png_ptr;
png_infop info_ptr;
png_color_8 sig_bit;
png_colorp palette_ptr;
png_bytep trans_ptr;//, tex_ptr;
int bit_depth, color_type, row_bytes, num_palette;
int i;
//uint16 srcfmt, destfmt;
if (!src || !fp)
return 0;
png_ptr = png_create_write_struct(PNG_LIBPNG_VER_STRING, NULL, NULL, NULL);
if (png_ptr == NULL)
return 0;
info_ptr = png_create_info_struct(png_ptr);
if (info_ptr == NULL) {
png_destroy_write_struct(&png_ptr, NULL);
return 0;
}
/*if (setjmp(png_ptr->jmpbuf)) {
png_destroy_write_struct(&png_ptr, &info_ptr);
return 0;
}*/
png_init_io(png_ptr, fp);
/* TODO: images must be converted to RGBA8888 or CI8,
* palettes need to be separated to A and RGB. */
/* N64 formats
* Format: 0 - RGBA, 1 - YUV, 2 - CI, 3 - IA, 4 - I
* Size: 0 - 4bit, 1 - 8bit, 2 - 16bit, 3 - 32 bit
* format = (Format << 8 | Size);
*/
/* each channel is saved in 8bits for consistency */
switch (format) {
case 0x0002:/* RGBA5551 */
bit_depth = 8;
sig_bit.red = 5;
sig_bit.green = 5;
sig_bit.blue = 5;
sig_bit.alpha = 1;
color_type = PNG_COLOR_TYPE_RGB_ALPHA;
break;
case 0x0003:/* RGBA8888 */
case 0x0302:/* IA88 */
bit_depth = 8;
sig_bit.red = 8;
sig_bit.green = 8;
sig_bit.blue = 8;
sig_bit.alpha = 8;
color_type = PNG_COLOR_TYPE_RGB_ALPHA;
break;
case 0x0300:/* IA31 */
bit_depth = 8;
sig_bit.red = 3;
sig_bit.green = 3;
sig_bit.blue = 3;
sig_bit.alpha = 1;
color_type = PNG_COLOR_TYPE_RGB_ALPHA;
break;
case 0x0301:/* IA44 */
bit_depth = 8;
sig_bit.red = 4;
sig_bit.green = 4;
sig_bit.blue = 4;
sig_bit.alpha = 4;
color_type = PNG_COLOR_TYPE_RGB_ALPHA;
break;
case 0x0400:/* I4 */
bit_depth = 8;
sig_bit.red = 4;
sig_bit.green = 4;
sig_bit.blue = 4;
color_type = PNG_COLOR_TYPE_RGB;
break;
case 0x0401:/* I8 */
case 0x0402:/* I16 */
bit_depth = 8;
sig_bit.red = 8;
sig_bit.green = 8;
sig_bit.blue = 8;
color_type = PNG_COLOR_TYPE_RGB;
break;
case 0x0200:/* CI4 */
bit_depth = 8;
num_palette = 16;
color_type = PNG_COLOR_TYPE_PALETTE;
break;
case 0x0201:/* CI8 */
bit_depth = 8;
num_palette = 256;
color_type = PNG_COLOR_TYPE_PALETTE;
break;
case 0x0102:/* YUV ? */
case 0x0103:
default:
/* unsupported format */
png_destroy_write_struct(&png_ptr, &info_ptr);
return 0;
}
switch (color_type) {
case PNG_COLOR_TYPE_RGB_ALPHA:
case PNG_COLOR_TYPE_RGB:
//row_bytes = (bit_depth * width) >> 1;
row_bytes = rowStride;
png_set_bgr(png_ptr);
png_set_sBIT(png_ptr, info_ptr, &sig_bit);
break;
case PNG_COLOR_TYPE_PALETTE:
//row_bytes = (bit_depth * width) >> 3;
row_bytes = rowStride;
png_set_PLTE(png_ptr, info_ptr, palette_ptr, num_palette);
png_set_tRNS(png_ptr, info_ptr, trans_ptr, num_palette, 0);
}
//png_set_filter(png_ptr, 0, PNG_ALL_FILTERS);
//if (bit_depth == 16)
// png_set_swap(png_ptr);
//if (bit_depth < 8)
// png_set_packswap(png_ptr);
png_set_IHDR(png_ptr, info_ptr, width, height,
bit_depth, color_type, PNG_INTERLACE_NONE,
PNG_COMPRESSION_TYPE_DEFAULT, PNG_FILTER_TYPE_DEFAULT);
//png_set_gAMA(png_ptr, info_ptr, 1.0);
png_write_info(png_ptr, info_ptr);
for (i = 0; i < height; i++) {
png_write_row(png_ptr, (png_bytep)src);
src += row_bytes;
}
png_write_end(png_ptr, info_ptr);
png_destroy_write_struct(&png_ptr, &info_ptr);
//if (tex_ptr) delete [] tex_ptr;
return 1;
}
boolean
TxImage::getBMPInfo(FILE* fp, BITMAPFILEHEADER* bmp_fhdr, BITMAPINFOHEADER* bmp_ihdr)
{
/*
* read in BITMAPFILEHEADER
*/
/* is this a BMP file? */
if (fread(&bmp_fhdr->bfType, 2, 1, fp) != 1)
return 0;
if (memcmp(&bmp_fhdr->bfType, "BM", 2) != 0)
return 0;
/* get file size */
if (fread(&bmp_fhdr->bfSize, 4, 1, fp) != 1)
return 0;
/* reserved 1 */
if (fread(&bmp_fhdr->bfReserved1, 2, 1, fp) != 1)
return 0;
/* reserved 2 */
if (fread(&bmp_fhdr->bfReserved2, 2, 1, fp) != 1)
return 0;
/* offset to the image data */
if (fread(&bmp_fhdr->bfOffBits, 4, 1, fp) != 1)
return 0;
/*
* read in BITMAPINFOHEADER
*/
/* size of BITMAPINFOHEADER */
if (fread(&bmp_ihdr->biSize, 4, 1, fp) != 1)
return 0;
/* is this a Windows BMP? */
if (bmp_ihdr->biSize != 40)
return 0;
/* width of the bitmap in pixels */
if (fread(&bmp_ihdr->biWidth, 4, 1, fp) != 1)
return 0;
/* height of the bitmap in pixels */
if (fread(&bmp_ihdr->biHeight, 4, 1, fp) != 1)
return 0;
/* number of planes (always 1) */
if (fread(&bmp_ihdr->biPlanes, 2, 1, fp) != 1)
return 0;
/* number of bits-per-pixel. (1, 4, 8, 16, 24, 32) */
if (fread(&bmp_ihdr->biBitCount, 2, 1, fp) != 1)
return 0;
/* compression for a compressed bottom-up bitmap
* 0 : uncompressed format
* 1 : run-length encoded 4 bpp format
* 2 : run-length encoded 8 bpp format
* 3 : bitfield
*/
if (fread(&bmp_ihdr->biCompression, 4, 1, fp) != 1)
return 0;
/* size of the image in bytes */
if (fread(&bmp_ihdr->biSizeImage, 4, 1, fp) != 1)
return 0;
/* horizontal resolution in pixels-per-meter */
if (fread(&bmp_ihdr->biXPelsPerMeter, 4, 1, fp) != 1)
return 0;
/* vertical resolution in pixels-per-meter */
if (fread(&bmp_ihdr->biYPelsPerMeter, 4, 1, fp) != 1)
return 0;
/* number of color indexes in the color table that are actually used */
if (fread(&bmp_ihdr->biClrUsed, 4, 1, fp) != 1)
return 0;
/* the number of color indexes that are required for displaying */
if (fread(&bmp_ihdr->biClrImportant, 4, 1, fp) != 1)
return 0;
return 1;
}
uint8*
TxImage::readBMP(FILE* fp, int* width, int* height, uint16* format)
{
/* NOTE: returned image format;
* 4, 8bit palette bmp -> GR_TEXFMT_P_8
* 24, 32bit bmp -> GR_TEXFMT_ARGB_8888
*/
uint8 *image = NULL;
uint8 *image_row = NULL;
uint8 *tmpimage = NULL;
int row_bytes, pos, i, j;
/* Windows Bitmap */
BITMAPFILEHEADER bmp_fhdr;
BITMAPINFOHEADER bmp_ihdr;
/* initialize */
*width = 0;
*height = 0;
*format = 0;
/* check if we have a valid bmp file */
if (!fp)
return NULL;
if (!getBMPInfo(fp, &bmp_fhdr, &bmp_ihdr)) {
INFO(80, L"error reading bitmap file! bitmap image is corrupt.\n");
return NULL;
}
DBG_INFO(80, L"bmp format %d x %d bitdepth:%d compression:%x offset:%d\n",
bmp_ihdr.biWidth, bmp_ihdr.biHeight, bmp_ihdr.biBitCount,
bmp_ihdr.biCompression, bmp_fhdr.bfOffBits);
/* rowStride in bytes */
row_bytes = (bmp_ihdr.biWidth * bmp_ihdr.biBitCount) >> 3;
/* align to 4bytes boundary */
row_bytes = (row_bytes + 3) & ~3;
/* Rice hi-res textures */
if (!(bmp_ihdr.biBitCount == 8 || bmp_ihdr.biBitCount == 4 || bmp_ihdr.biBitCount == 32 || bmp_ihdr.biBitCount == 24) ||
bmp_ihdr.biCompression != 0) {
DBG_INFO(80, L"Error: incompatible bitmap format!\n");
return NULL;
}
switch (bmp_ihdr.biBitCount) {
case 8:
case 32:
/* 8 bit, 32 bit bitmap */
image = (uint8*)malloc(row_bytes * bmp_ihdr.biHeight);
if (image) {
tmpimage = image;
pos = bmp_fhdr.bfOffBits + row_bytes * (bmp_ihdr.biHeight - 1);
for (i = 0; i < bmp_ihdr.biHeight; i++) {
/* read in image */
fseek(fp, pos, SEEK_SET);
fread(tmpimage, row_bytes, 1, fp);
tmpimage += row_bytes;
pos -= row_bytes;
}
}
break;
case 4:
/* 4bit bitmap */
image = (uint8*)malloc((row_bytes * bmp_ihdr.biHeight) << 1);
image_row = (uint8*)malloc(row_bytes);
if (image && image_row) {
tmpimage = image;
pos = bmp_fhdr.bfOffBits + row_bytes * (bmp_ihdr.biHeight - 1);
for (i = 0; i < bmp_ihdr.biHeight; i++) {
/* read in image */
fseek(fp, pos, SEEK_SET);
fread(image_row, row_bytes, 1, fp);
/* expand 4bpp to 8bpp. stuff 4bit values into 8bit comps. */
for (j = 0; j < row_bytes; j++) {
tmpimage[j << 1] = image_row[j] & 0x0f;
tmpimage[(j << 1) + 1] = (image_row[j] & 0xf0) >> 4;
}
tmpimage += (row_bytes << 1);
pos -= row_bytes;
}
free(image_row);
} else {
if (image_row) free(image_row);
if (image) free(image);
image = NULL;
}
break;
case 24:
/* 24 bit bitmap */
image = (uint8*)malloc((bmp_ihdr.biWidth * bmp_ihdr.biHeight) << 2);
image_row = (uint8*)malloc(row_bytes);
if (image && image_row) {
tmpimage = image;
pos = bmp_fhdr.bfOffBits + row_bytes * (bmp_ihdr.biHeight - 1);
for (i = 0; i < bmp_ihdr.biHeight; i++) {
/* read in image */
fseek(fp, pos, SEEK_SET);
fread(image_row, row_bytes, 1, fp);
/* convert 24bpp to 32bpp. */
for (j = 0; j < bmp_ihdr.biWidth; j++) {
tmpimage[(j << 2)] = image_row[j * 3];
tmpimage[(j << 2) + 1] = image_row[j * 3 + 1];
tmpimage[(j << 2) + 2] = image_row[j * 3 + 2];
tmpimage[(j << 2) + 3] = 0xFF;
}
tmpimage += (bmp_ihdr.biWidth << 2);
pos -= row_bytes;
}
free(image_row);
} else {
if (image_row) free(image_row);
if (image) free(image);
image = NULL;
}
}
if (image) {
*width = (row_bytes << 3) / bmp_ihdr.biBitCount;
*height = bmp_ihdr.biHeight;
switch (bmp_ihdr.biBitCount) {
case 8:
case 4:
*format = GR_TEXFMT_P_8;
break;
case 32:
case 24:
*format = GR_TEXFMT_ARGB_8888;
}
#if POW2_TEXTURES
/* next power of 2 size conversions */
/* NOTE: I can do this in the above loop for faster operations, but some
* texture packs require a workaround. see HACKALERT in nextPow2().
*/
TxReSample txReSample = new TxReSample; // XXX: temporary. move to a better place.
#if (POW2_TEXTURES == 2)
if (!txReSample->nextPow2(&image, width, height, 8, 1)) {
#else
if (!txReSample->nextPow2(&image, width, height, 8, 0)) {
#endif
if (image) {
free(image);
image = NULL;
}
*width = 0;
*height = 0;
*format = 0;
}
delete txReSample;
#endif /* POW2_TEXTURES */
}
#ifdef DEBUG
if (!image) {
DBG_INFO(80, L"Error: failed to load bmp image!\n");
}
#endif
return image;
}
boolean
TxImage::getDDSInfo(FILE *fp, DDSFILEHEADER *dds_fhdr)
{
/*
* read in DDSFILEHEADER
*/
/* is this a DDS file? */
if (fread(&dds_fhdr->dwMagic, 4, 1, fp) != 1)
return 0;
if (memcmp(&dds_fhdr->dwMagic, "DDS ", 4) != 0)
return 0;
if (fread(&dds_fhdr->dwSize, 4, 1, fp) != 1)
return 0;
/* get file flags */
if (fread(&dds_fhdr->dwFlags, 4, 1, fp) != 1)
return 0;
/* height of dds in pixels */
if (fread(&dds_fhdr->dwHeight, 4, 1, fp) != 1)
return 0;
/* width of dds in pixels */
if (fread(&dds_fhdr->dwWidth, 4, 1, fp) != 1)
return 0;
if (fread(&dds_fhdr->dwLinearSize, 4, 1, fp) != 1)
return 0;
if (fread(&dds_fhdr->dwDepth, 4, 1, fp) != 1)
return 0;
if (fread(&dds_fhdr->dwMipMapCount, 4, 1, fp) != 1)
return 0;
if (fread(&dds_fhdr->dwReserved1, 4 * 11, 1, fp) != 1)
return 0;
if (fread(&dds_fhdr->ddpf.dwSize, 4, 1, fp) != 1)
return 0;
if (fread(&dds_fhdr->ddpf.dwFlags, 4, 1, fp) != 1)
return 0;
if (fread(&dds_fhdr->ddpf.dwFourCC, 4, 1, fp) != 1)
return 0;
if (fread(&dds_fhdr->ddpf.dwRGBBitCount, 4, 1, fp) != 1)
return 0;
if (fread(&dds_fhdr->ddpf.dwRBitMask, 4, 1, fp) != 1)
return 0;
if (fread(&dds_fhdr->ddpf.dwGBitMask, 4, 1, fp) != 1)
return 0;
if (fread(&dds_fhdr->ddpf.dwBBitMask, 4, 1, fp) != 1)
return 0;
if (fread(&dds_fhdr->ddpf.dwRGBAlphaBitMask, 4, 1, fp) != 1)
return 0;
if (fread(&dds_fhdr->dwCaps1, 4, 1, fp) != 1)
return 0;
if (fread(&dds_fhdr->dwCaps2, 4, 1, fp) != 1)
return 0;
return 1;
}
uint8*
TxImage::readDDS(FILE* fp, int* width, int* height, uint16* format)
{
uint8 *image = NULL;
DDSFILEHEADER dds_fhdr;
uint16 tmpformat = 0;
/* initialize */
*width = 0;
*height = 0;
*format = 0;
/* check if we have a valid dds file */
if (!fp)
return NULL;
if (!getDDSInfo(fp, &dds_fhdr)) {
INFO(80, L"error reading dds file! dds image is corrupt.\n");
return NULL;
}
DBG_INFO(80, L"dds format %d x %d HeaderSize %d LinearSize %d\n",
dds_fhdr.dwWidth, dds_fhdr.dwHeight, dds_fhdr.dwSize, dds_fhdr.dwLinearSize);
if (!(dds_fhdr.dwFlags & (DDSD_CAPS|DDSD_WIDTH|DDSD_HEIGHT|DDSD_PIXELFORMAT|DDSD_LINEARSIZE))) {
DBG_INFO(80, L"Error: incompatible dds format!\n");
return NULL;
}
if ((dds_fhdr.dwFlags & DDSD_MIPMAPCOUNT) && dds_fhdr.dwMipMapCount != 1) {
DBG_INFO(80, L"Error: mipmapped dds not supported!\n");
return NULL;
}
if (!((dds_fhdr.ddpf.dwFlags & DDPF_FOURCC) && dds_fhdr.dwCaps2 == 0)) {
DBG_INFO(80, L"Error: not fourcc standard texture!\n");
return NULL;
}
if (memcmp(&dds_fhdr.ddpf.dwFourCC, "DXT1", 4) == 0) {
DBG_INFO(80, L"DXT1 format\n");
/* compensate for missing LinearSize */
dds_fhdr.dwLinearSize = (dds_fhdr.dwWidth * dds_fhdr.dwHeight) >> 1;
tmpformat = GR_TEXFMT_ARGB_CMP_DXT1;
} else if (memcmp(&dds_fhdr.ddpf.dwFourCC, "DXT3", 4) == 0) {
DBG_INFO(80, L"DXT3 format\n");
dds_fhdr.dwLinearSize = dds_fhdr.dwWidth * dds_fhdr.dwHeight;
tmpformat = GR_TEXFMT_ARGB_CMP_DXT3;
} else if (memcmp(&dds_fhdr.ddpf.dwFourCC, "DXT5", 4) == 0) {
DBG_INFO(80, L"DXT5 format\n");
dds_fhdr.dwLinearSize = dds_fhdr.dwWidth * dds_fhdr.dwHeight;
tmpformat = GR_TEXFMT_ARGB_CMP_DXT5;
} else {
DBG_INFO(80, L"Error: not DXT1 or DXT3 or DXT5 format!\n");
return NULL;
}
/* read in image */
image = (uint8*)malloc(dds_fhdr.dwLinearSize);
if (image) {
*width = dds_fhdr.dwWidth;
*height = dds_fhdr.dwHeight;
*format = tmpformat;
fseek(fp, 128, SEEK_SET); /* size of header is 128 bytes */
fread(image, dds_fhdr.dwLinearSize, 1, fp);
}
return image;
}

116
Source/GlideHQ/TxImage.h Normal file
View File

@ -0,0 +1,116 @@
/*
* Texture Filtering
* Version: 1.0
*
* Copyright (C) 2007 Hiroshi Morii All Rights Reserved.
* Email koolsmoky(at)users.sourceforge.net
* Web http://www.3dfxzone.it/koolsmoky
*
* this 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, or (at your option)
* any later version.
*
* this 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 GNU Make; see the file COPYING. If not, write to
* the Free Software Foundation, 675 Mass Ave, Cambridge, MA 02139, USA.
*/
#ifndef __TXIMAGE_H__
#define __TXIMAGE_H__
#include <stdio.h>
#include <png/png.h>
#include "TxInternal.h"
#ifndef WIN32
typedef struct tagBITMAPFILEHEADER {
unsigned short bfType;
unsigned long bfSize;
unsigned short bfReserved1;
unsigned short bfReserved2;
unsigned long bfOffBits;
} BITMAPFILEHEADER;
typedef struct tagBITMAPINFOHEADER {
unsigned long biSize;
long biWidth;
long biHeight;
unsigned short biPlanes;
unsigned short biBitCount;
unsigned long biCompression;
unsigned long biSizeImage;
long biXPelsPerMeter;
long biYPelsPerMeter;
unsigned long biClrUsed;
unsigned long biClrImportant;
} BITMAPINFOHEADER;
#else
typedef struct tagBITMAPFILEHEADER BITMAPFILEHEADER;
typedef struct tagBITMAPINFOHEADER BITMAPINFOHEADER;
#endif
#define DDSD_CAPS 0x00000001
#define DDSD_HEIGHT 0x00000002
#define DDSD_WIDTH 0x00000004
#define DDSD_PITCH 0x00000008
#define DDSD_PIXELFORMAT 0x00001000
#define DDSD_MIPMAPCOUNT 0x00020000
#define DDSD_LINEARSIZE 0x00080000
#define DDSD_DEPTH 0x00800000
#define DDPF_ALPHAPIXELS 0x00000001
#define DDPF_FOURCC 0x00000004
#define DDPF_RGB 0x00000040
#define DDSCAPS_COMPLEX 0x00000008
#define DDSCAPS_TEXTURE 0x00001000
#define DDSCAPS_MIPMAP 0x00400000
typedef struct tagDDSPIXELFORMAT {
unsigned long dwSize;
unsigned long dwFlags;
unsigned long dwFourCC;
unsigned long dwRGBBitCount;
unsigned long dwRBitMask;
unsigned long dwGBitMask;
unsigned long dwBBitMask;
unsigned long dwRGBAlphaBitMask;
} DDSPIXELFORMAT;
typedef struct tagDDSFILEHEADER {
unsigned long dwMagic;
unsigned long dwSize;
unsigned long dwFlags;
unsigned long dwHeight;
unsigned long dwWidth;
unsigned long dwLinearSize;
unsigned long dwDepth;
unsigned long dwMipMapCount;
unsigned long dwReserved1[11];
DDSPIXELFORMAT ddpf;
unsigned long dwCaps1;
unsigned long dwCaps2;
} DDSFILEHEADER;
class TxImage
{
private:
boolean getPNGInfo(FILE *fp, png_structp *png_ptr, png_infop *info_ptr);
boolean getBMPInfo(FILE *fp, BITMAPFILEHEADER *bmp_fhdr, BITMAPINFOHEADER *bmp_ihdr);
boolean getDDSInfo(FILE *fp, DDSFILEHEADER *dds_fhdr);
public:
TxImage() {}
~TxImage() {}
uint8* readPNG(FILE* fp, int* width, int* height, uint16* format);
boolean writePNG(uint8* src, FILE* fp, int width, int height, int rowStride, uint16 format, uint8 *palette);
uint8* readBMP(FILE* fp, int* width, int* height, uint16* format);
uint8* readDDS(FILE* fp, int* width, int* height, uint16* format);
};
#endif /* __TXIMAGE_H__ */

100
Source/GlideHQ/TxInternal.h Normal file
View File

@ -0,0 +1,100 @@
/*
* Texture Filtering
* Version: 1.0
*
* Copyright (C) 2007 Hiroshi Morii All Rights Reserved.
* Email koolsmoky(at)users.sourceforge.net
* Web http://www.3dfxzone.it/koolsmoky
*
* this 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, or (at your option)
* any later version.
*
* this 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 GNU Make; see the file COPYING. If not, write to
* the Free Software Foundation, 675 Mass Ave, Cambridge, MA 02139, USA.
*/
#ifndef __INTERNAL_H__
#define __INTERNAL_H__
#include "Ext_TxFilter.h"
/* dll exports */
#ifdef TXFILTER_DLL
#define TAPI __declspec(dllexport)
#define TAPIENTRY
#else
#define TAPI
#define TAPIENTRY
#endif
typedef unsigned char uint8;
typedef unsigned short uint16;
typedef unsigned long uint32;
#ifdef WIN32
#define KBHIT(key) ((GetAsyncKeyState(key) & 0x8001) == 0x8001)
#else
#define KBHIT(key) (0)
#endif
/* from OpenGL glext.h */
#define GL_COMPRESSED_RGB_S3TC_DXT1_EXT 0x83F0
#define GL_COMPRESSED_RGBA_S3TC_DXT1_EXT 0x83F1
#define GL_COMPRESSED_RGBA_S3TC_DXT3_EXT 0x83F2
#define GL_COMPRESSED_RGBA_S3TC_DXT5_EXT 0x83F3
/* for explicit fxt1 compression */
#define CC_CHROMA 0x0
#define CC_HI 0x1
#define CC_ALPHA 0x2
/* in-memory zlib texture compression */
#define GR_TEXFMT_GZ 0x8000
#if 0 /* this is here to remind me of other formats */
/* from 3Dfx Interactive Inc. glide.h */
#define GR_TEXFMT_8BIT 0x0
#define GR_TEXFMT_RGB_332 GR_TEXFMT_8BIT
#define GR_TEXFMT_YIQ_422 0x1
#define GR_TEXFMT_ALPHA_8 0x2 /* (0..0xFF) alpha */
#define GR_TEXFMT_INTENSITY_8 0x3 /* (0..0xFF) intensity */
#define GR_TEXFMT_ALPHA_INTENSITY_44 0x4
#define GR_TEXFMT_P_8 0x5 /* 8-bit palette */
#define GR_TEXFMT_RSVD0 0x6 /* GR_TEXFMT_P_8_RGBA */
#define GR_TEXFMT_P_8_6666 GR_TEXFMT_RSVD0
#define GR_TEXFMT_P_8_6666_EXT GR_TEXFMT_RSVD0
#define GR_TEXFMT_RSVD1 0x7
#define GR_TEXFMT_16BIT 0x8
#define GR_TEXFMT_ARGB_8332 GR_TEXFMT_16BIT
#define GR_TEXFMT_AYIQ_8422 0x9
#define GR_TEXFMT_RGB_565 0xa
#define GR_TEXFMT_ARGB_1555 0xb
#define GR_TEXFMT_ARGB_4444 0xc
#define GR_TEXFMT_ALPHA_INTENSITY_88 0xd
#define GR_TEXFMT_AP_88 0xe /* 8-bit alpha 8-bit palette */
#define GR_TEXFMT_RSVD2 0xf
#define GR_TEXFMT_RSVD4 GR_TEXFMT_RSVD2
/* from 3Dfx Interactive Inc. g3ext.h */
#define GR_TEXFMT_ARGB_CMP_FXT1 0x11
#define GR_TEXFMT_ARGB_8888 0x12
#define GR_TEXFMT_YUYV_422 0x13
#define GR_TEXFMT_UYVY_422 0x14
#define GR_TEXFMT_AYUV_444 0x15
#define GR_TEXFMT_ARGB_CMP_DXT1 0x16
#define GR_TEXFMT_ARGB_CMP_DXT2 0x17
#define GR_TEXFMT_ARGB_CMP_DXT3 0x18
#define GR_TEXFMT_ARGB_CMP_DXT4 0x19
#define GR_TEXFMT_ARGB_CMP_DXT5 0x1A
#define GR_TEXTFMT_RGB_888 0xFF
#endif
#endif /* __INTERNAL_H__ */

2405
Source/GlideHQ/TxQuantize.cpp Normal file
View File

File diff suppressed because it is too large Load Diff

99
Source/GlideHQ/TxQuantize.h Normal file
View File

@ -0,0 +1,99 @@
/*
* Texture Filtering
* Version: 1.0
*
* Copyright (C) 2007 Hiroshi Morii All Rights Reserved.
* Email koolsmoky(at)users.sourceforge.net
* Web http://www.3dfxzone.it/koolsmoky
*
* this 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, or (at your option)
* any later version.
*
* this 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 GNU Make; see the file COPYING. If not, write to
* the Free Software Foundation, 675 Mass Ave, Cambridge, MA 02139, USA.
*/
#ifndef __TXQUANTIZE_H__
#define __TXQUANTIZE_H__
/* Glide64 DXTn workaround
* (0:disable, 1:enable) */
#define GLIDE64_DXTN 1
#include "TxInternal.h"
#include "TxUtil.h"
class TxQuantize
{
private:
TxUtil *_txUtil;
int _numcore;
fxtCompressTexFuncExt _tx_compress_fxt1;
dxtCompressTexFuncExt _tx_compress_dxtn;
/* fast optimized... well, sort of. */
void ARGB1555_ARGB8888(uint32* src, uint32* dst, int width, int height);
void ARGB4444_ARGB8888(uint32* src, uint32* dst, int width, int height);
void RGB565_ARGB8888(uint32* src, uint32* dst, int width, int height);
void A8_ARGB8888(uint32* src, uint32* dst, int width, int height);
void AI44_ARGB8888(uint32* src, uint32* dst, int width, int height);
void AI88_ARGB8888(uint32* src, uint32* dst, int width, int height);
void ARGB8888_ARGB1555(uint32* src, uint32* dst, int width, int height);
void ARGB8888_ARGB4444(uint32* src, uint32* dst, int width, int height);
void ARGB8888_RGB565(uint32* src, uint32* dst, int width, int height);
void ARGB8888_A8(uint32* src, uint32* dst, int width, int height);
void ARGB8888_AI44(uint32* src, uint32* dst, int width, int height);
void ARGB8888_AI88(uint32* src, uint32* dst, int width, int height);
/* quality */
void ARGB8888_RGB565_ErrD(uint32* src, uint32* dst, int width, int height);
void ARGB8888_ARGB1555_ErrD(uint32* src, uint32* dst, int width, int height);
void ARGB8888_ARGB4444_ErrD(uint32* src, uint32* dst, int width, int height);
void ARGB8888_AI44_ErrD(uint32* src, uint32* dst, int width, int height);
void ARGB8888_AI88_Slow(uint32* src, uint32* dst, int width, int height);
void ARGB8888_I8_Slow(uint32* src, uint32* dst, int width, int height);
/* compressors */
boolean FXT1(uint8 *src, uint8 *dest,
int srcwidth, int srcheight, uint16 srcformat,
int *destwidth, int *destheight, uint16 *destformat);
boolean DXTn(uint8 *src, uint8 *dest,
int srcwidth, int srcheight, uint16 srcformat,
int *destwidth, int *destheight, uint16 *destformat);
public:
TxQuantize();
~TxQuantize();
/* others */
void P8_16BPP(uint32* src, uint32* dst, int width, int height, uint32* palette);
boolean quantize(uint8* src, uint8* dest, int width, int height, uint16 srcformat, uint16 destformat, boolean fastQuantizer = 1);
boolean compress(uint8 *src, uint8 *dest,
int srcwidth, int srcheight, uint16 srcformat,
int *destwidth, int *destheight, uint16 *destformat,
int compressionType);
#if 0 /* unused */
void ARGB8888_I8(uint32* src, uint32* dst, int width, int height);
void I8_ARGB8888(uint32* src, uint32* dst, int width, int height);
void ARGB1555_ABGR8888(uint32* src, uint32* dst, int width, int height);
void ARGB4444_ABGR8888(uint32* src, uint32* dst, int width, int height);
void ARGB8888_ABGR8888(uint32* src, uint32* dst, int width, int height);
#endif
};
#endif /* __TXQUANTIZE_H__ */

417
Source/GlideHQ/TxReSample.cpp Normal file
View File

@ -0,0 +1,417 @@
/*
* Texture Filtering
* Version: 1.0
*
* Copyright (C) 2007 Hiroshi Morii All Rights Reserved.
* Email koolsmoky(at)users.sourceforge.net
* Web http://www.3dfxzone.it/koolsmoky
*
* this 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, or (at your option)
* any later version.
*
* this 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 GNU Make; see the file COPYING. If not, write to
* the Free Software Foundation, 675 Mass Ave, Cambridge, MA 02139, USA.
*/
#include "TxReSample.h"
#include "TxDbg.h"
#include <stdlib.h>
#include <memory.h>
#define _USE_MATH_DEFINES
#include <math.h>
#ifndef M_PI
#define M_PI 3.14159265358979323846
#endif
int
TxReSample::nextPow2(int num)
{
num = num - 1;
num = num | (num >> 1);
num = num | (num >> 2);
num = num | (num >> 4);
num = num | (num >> 8);
num = num | (num >> 16);
/*num = num | (num >> 32);*//* for 64bit architecture */
num = num + 1;
return num;
}
boolean
TxReSample::nextPow2(uint8** image, int* width, int* height, int bpp, boolean use_3dfx = 0)
{
/* NOTE: bpp must be one of the follwing: 8, 16, 24, 32 bits per pixel */
if (!*image || !*width || !*height || !bpp)
return 0;
int row_bytes = ((*width * bpp) >> 3);
int o_row_bytes = row_bytes;
int o_width = *width;
int n_width = *width;
int o_height = *height;
int n_height = *height;
/* HACKALERT: I have explicitly subtracted (n) from width/height to
* adjust textures that have (n) pixel larger width/height than
* power of 2 size. This is a dirty hack for textures that have
* munged aspect ratio by (n) pixel to the original.
*/
if (n_width > 64) n_width -= 4;
else if (n_width > 16) n_width -= 2;
else if (n_width > 4) n_width -= 1;
if (n_height > 64) n_height -= 4;
else if (n_height > 16) n_height -= 2;
else if (n_height > 4) n_height -= 1;
n_width = nextPow2(n_width);
n_height = nextPow2(n_height);
row_bytes = (n_width * bpp) >> 3;
/* 3dfx Glide3 format, W:H aspect ratio range (8:1 - 1:8) */
if (use_3dfx) {
if (n_width > n_height) {
if (n_width > (n_height << 3))
n_height = n_width >> 3;
} else {
if (n_height > (n_width << 3)) {
n_width = n_height >> 3;
row_bytes = (n_width * bpp) >> 3;
}
}
DBG_INFO(80, L"using 3dfx W:H aspect ratio range (8:1 - 1:8).\n");
}
/* do we really need to do this ? */
if (o_width == n_width && o_height == n_height)
return 1; /* nope */
DBG_INFO(80, L"expand image to next power of 2 dimensions. %d x %d -> %d x %d\n",
o_width, o_height, n_width, n_height);
if (o_width > n_width)
o_width = n_width;
if (o_height > n_height)
o_height = n_height;
/* allocate memory to read in image */
uint8 *pow2image = (uint8*)malloc(row_bytes * n_height);
/* read in image */
if (pow2image) {
int i, j;
uint8 *tmpimage = *image, *tmppow2image = pow2image;
for (i = 0; i < o_height; i++) {
/* copy row */
memcpy(tmppow2image, tmpimage, ((o_width * bpp) >> 3));
/* expand to pow2 size by replication */
for(j = ((o_width * bpp) >> 3); j < row_bytes; j++)
tmppow2image[j] = tmppow2image[j - (bpp >> 3)];
tmppow2image += row_bytes;
tmpimage += o_row_bytes;
}
/* expand to pow2 size by replication */
for (i = o_height; i < n_height; i++)
memcpy(&pow2image[row_bytes * i], &pow2image[row_bytes * (i - 1)], row_bytes);
free(*image);
*image = pow2image;
*height = n_height;
*width = n_width;
return 1;
}
return 0;
}
/* Ken Turkowski
* Filters for Common Resampling Tasks
* Apple Computer 1990
*/
double
TxReSample::tent(double x)
{
if (x < 0.0) x = -x;
if (x < 1.0) return (1.0 - x);
return 0.0;
}
double
TxReSample::gaussian(double x)
{
if (x < 0) x = -x;
if (x < 2.0) return pow(2.0, -2.0 * x * x);
return 0.0;
}
double
TxReSample::sinc(double x)
{
if (x == 0) return 1.0;
x *= M_PI;
return (sin(x) / x);
}
double
TxReSample::lanczos3(double x)
{
if (x < 0) x = -x;
if (x < 3.0) return (sinc(x) * sinc(x/3.0));
return 0.0;
}
/* Don P. Mitchell and Arun N. Netravali
* Reconstruction Filters in Computer Graphics
* SIGGRAPH '88
* Proceedings of the 15th annual conference on Computer
* graphics and interactive techniques, pp221-228, 1988
*/
double
TxReSample::mitchell(double x)
{
if (x < 0) x = -x;
if (x < 2.0) {
const double B = 1.0 / 3.0;
const double C = 1.0 / 3.0;
if (x < 1.0) {
x = (((12.0 - 9.0 * B - 6.0 * C) * (x * x * x))
+ ((-18.0 + 12.0 * B + 6.0 * C) * (x * x))
+ (6.0 - 2.0 * B));
} else {
x = (((-1.0 * B - 6.0 * C) * (x * x * x))
+ ((6.0 * B + 30.0 * C) * (x * x))
+ ((-12.0 * B - 48.0 * C) * x)
+ (8.0 * B + 24.0 * C));
}
return (x / 6.0);
}
return 0.0;
}
/* J. F. Kaiser and W. A. Reed
* Data smoothing using low-pass digital filters
* Rev. Sci. instrum. 48 (11), pp1447-1457, 1977
*/
double
TxReSample::besselI0(double x)
{
/* zero-order modified bessel function of the first kind */
const double eps_coeff = 1E-16; /* small enough */
double xh, sum, pow, ds;
xh = 0.5 * x;
sum = 1.0;
pow = 1.0;
ds = 1.0;
int k = 0;
while (ds > sum * eps_coeff) {
k++;
pow *= (xh / k);
ds = pow * pow;
sum = sum + ds;
}
return sum;
}
double
TxReSample::kaiser(double x)
{
const double alpha = 4.0;
const double half_window = 5.0;
const double ratio = x / half_window;
return sinc(x) * besselI0(alpha * sqrt(1 - ratio * ratio)) / besselI0(alpha);
}
boolean
TxReSample::minify(uint8 **src, int *width, int *height, int ratio)
{
/* NOTE: src must be ARGB8888, ratio is the inverse representation */
#if 0
if (!*src || ratio < 2) return 0;
/* Box filtering.
* It would be nice to do Kaiser filtering.
* N64 uses narrow strip textures which makes it hard to filter effectively.
*/
int x, y, x2, y2, offset, numtexel;
uint32 A, R, G, B, texel;
int tmpwidth = *width / ratio;
int tmpheight = *height / ratio;
uint8 *tmptex = (uint8*)malloc((tmpwidth * tmpheight) << 2);
if (tmptex) {
numtexel = ratio * ratio;
for (y = 0; y < tmpheight; y++) {
offset = ratio * y * *width;
for (x = 0; x < tmpwidth; x++) {
A = R = G = B = 0;
for (y2 = 0; y2 < ratio; y2++) {
for (x2 = 0; x2 < ratio; x2++) {
texel = ((uint32*)*src)[offset + *width * y2 + x2];
A += (texel >> 24);
R += ((texel >> 16) & 0x000000ff);
G += ((texel >> 8) & 0x000000ff);
B += (texel & 0x000000ff);
}
}
A = (A + ratio) / numtexel;
R = (R + ratio) / numtexel;
G = (G + ratio) / numtexel;
B = (B + ratio) / numtexel;
((uint32*)tmptex)[y * tmpwidth + x] = ((A << 24) | (R << 16) | (G << 8) | B);
offset += ratio;
}
}
free(*src);
*src = tmptex;
*width = tmpwidth;
*height = tmpheight;
DBG_INFO(80, L"minification ratio:%d -> %d x %d\n", ratio, *width, *height);
return 1;
}
DBG_INFO(80, L"Error: failed minification!\n");
return 0;
#else
if (!*src || ratio < 2) return 0;
/* Image Resampling */
/* half width of filter window.
* NOTE: must be 1.0 or larger.
*
* kaiser-bessel 5, lanczos3 3, mitchell 2, gaussian 1.5, tent 1
*/
double half_window = 5.0;
int x, y, x2, y2, z;
double A, R, G, B;
uint32 texel;
int tmpwidth = *width / ratio;
int tmpheight = *height / ratio;
/* resampled destination */
uint8 *tmptex = (uint8*)malloc((tmpwidth * tmpheight) << 2);
if (!tmptex) return 0;
/* work buffer. single row */
uint8 *workbuf = (uint8*)malloc(*width << 2);
if (!workbuf) {
free(tmptex);
return 0;
}
/* prepare filter lookup table. only half width required for symetric filters. */
double *weight = (double*)malloc((int)((half_window * ratio) * sizeof(double)));
if (!weight) {
free(tmptex);
free(workbuf);
return 0;
}
for (x = 0; x < half_window * ratio; x++) {
//weight[x] = tent((double)x / ratio) / ratio;
//weight[x] = gaussian((double)x / ratio) / ratio;
//weight[x] = lanczos3((double)x / ratio) / ratio;
//weight[x] = mitchell((double)x / ratio) / ratio;
weight[x] = kaiser((double)x / ratio) / ratio;
}
/* linear convolution */
for (y = 0; y < tmpheight; y++) {
for (x = 0; x < *width; x++) {
texel = ((uint32*)*src)[y * ratio * *width + x];
A = (double)(texel >> 24) * weight[0];
R = (double)((texel >> 16) & 0xff) * weight[0];
G = (double)((texel >> 8) & 0xff) * weight[0];
B = (double)((texel ) & 0xff) * weight[0];
for (y2 = 1; y2 < half_window * ratio; y2++) {
z = y * ratio + y2;
if (z >= *height) z = *height - 1;
texel = ((uint32*)*src)[z * *width + x];
A += (double)(texel >> 24) * weight[y2];
R += (double)((texel >> 16) & 0xff) * weight[y2];
G += (double)((texel >> 8) & 0xff) * weight[y2];
B += (double)((texel ) & 0xff) * weight[y2];
z = y * ratio - y2;
if (z < 0) z = 0;
texel = ((uint32*)*src)[z * *width + x];
A += (double)(texel >> 24) * weight[y2];
R += (double)((texel >> 16) & 0xff) * weight[y2];
G += (double)((texel >> 8) & 0xff) * weight[y2];
B += (double)((texel ) & 0xff) * weight[y2];
}
if (A < 0) A = 0; else if (A > 255) A = 255;
if (R < 0) R = 0; else if (R > 255) R = 255;
if (G < 0) G = 0; else if (G > 255) G = 255;
if (B < 0) B = 0; else if (B > 255) B = 255;
((uint32*)workbuf)[x] = (((uint32)A << 24) | ((uint32)R << 16) | ((uint32)G << 8) | (uint32)B);
}
for (x = 0; x < tmpwidth; x++) {
texel = ((uint32*)workbuf)[x * ratio];
A = (double)(texel >> 24) * weight[0];
R = (double)((texel >> 16) & 0xff) * weight[0];
G = (double)((texel >> 8) & 0xff) * weight[0];
B = (double)((texel ) & 0xff) * weight[0];
for (x2 = 1; x2 < half_window * ratio; x2++) {
z = x * ratio + x2;
if (z >= *width) z = *width - 1;
texel = ((uint32*)workbuf)[z];
A += (double)(texel >> 24) * weight[x2];
R += (double)((texel >> 16) & 0xff) * weight[x2];
G += (double)((texel >> 8) & 0xff) * weight[x2];
B += (double)((texel ) & 0xff) * weight[x2];
z = x * ratio - x2;
if (z < 0) z = 0;
texel = ((uint32*)workbuf)[z];
A += (double)(texel >> 24) * weight[x2];
R += (double)((texel >> 16) & 0xff) * weight[x2];
G += (double)((texel >> 8) & 0xff) * weight[x2];
B += (double)((texel ) & 0xff) * weight[x2];
}
if (A < 0) A = 0; else if (A > 255) A = 255;
if (R < 0) R = 0; else if (R > 255) R = 255;
if (G < 0) G = 0; else if (G > 255) G = 255;
if (B < 0) B = 0; else if (B > 255) B = 255;
((uint32*)tmptex)[y * tmpwidth + x] = (((uint32)A << 24) | ((uint32)R << 16) | ((uint32)G << 8) | (uint32)B);
}
}
free(*src);
*src = tmptex;
free(weight);
free(workbuf);
*width = tmpwidth;
*height = tmpheight;
DBG_INFO(80, L"minification ratio:%d -> %d x %d\n", ratio, *width, *height);
return 1;
#endif
}

45
Source/GlideHQ/TxReSample.h Normal file
View File

@ -0,0 +1,45 @@
/*
* Texture Filtering
* Version: 1.0
*
* Copyright (C) 2007 Hiroshi Morii All Rights Reserved.
* Email koolsmoky(at)users.sourceforge.net
* Web http://www.3dfxzone.it/koolsmoky
*
* this 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, or (at your option)
* any later version.
*
* this 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 GNU Make; see the file COPYING. If not, write to
* the Free Software Foundation, 675 Mass Ave, Cambridge, MA 02139, USA.
*/
#ifndef __TXRESAMPLE_H__
#define __TXRESAMPLE_H__
#include "TxInternal.h"
class TxReSample
{
private:
double tent(double x);
double gaussian(double x);
double sinc(double x);
double lanczos3(double x);
double mitchell(double x);
double besselI0(double x);
double kaiser(double x);
public:
boolean minify(uint8 **src, int *width, int *height, int ratio);
boolean nextPow2(uint8** image, int* width, int* height, int bpp, boolean use_3dfx);
int nextPow2(int num);
};
#endif /* __TXRESAMPLE_H__ */

80
Source/GlideHQ/TxTexCache.cpp Normal file
View File

@ -0,0 +1,80 @@
/*
* Texture Filtering
* Version: 1.0
*
* Copyright (C) 2007 Hiroshi Morii All Rights Reserved.
* Email koolsmoky(at)users.sourceforge.net
* Web http://www.3dfxzone.it/koolsmoky
*
* this 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, or (at your option)
* any later version.
*
* this 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 GNU Make; see the file COPYING. If not, write to
* the Free Software Foundation, 675 Mass Ave, Cambridge, MA 02139, USA.
*/
#ifdef WIN32
#pragma warning(disable: 4786)
#endif
/* dump cache to disk (0:disable, 1:enable) */
#define DUMP_CACHE 1
#include "TxTexCache.h"
#include "TxDbg.h"
#include <zlib/zlib.h>
#include <string>
#include <Common/path.h>
TxTexCache::~TxTexCache()
{
#if DUMP_CACHE
if (_options & DUMP_TEXCACHE) {
/* dump cache to disk */
std::wstring filename = _ident + L"_MEMORYCACHE.dat";
CPath cachepath(stdstr().FromUTF16(_path.c_str()).c_str(),"");
cachepath.AppendDirectory("cache");
int config = _options & (FILTER_MASK|ENHANCEMENT_MASK|COMPRESS_TEX|COMPRESSION_MASK|FORCE16BPP_TEX|GZ_TEXCACHE);
TxCache::save(stdstr(cachepath).ToUTF16().c_str(), filename.c_str(), config);
}
#endif
}
TxTexCache::TxTexCache(int options, int cachesize, const wchar_t *path, const wchar_t *ident,
dispInfoFuncExt callback
) : TxCache((options & ~GZ_HIRESTEXCACHE), cachesize, path, ident, callback)
{
/* assert local options */
if (_path.empty() || _ident.empty() || !_cacheSize)
_options &= ~DUMP_TEXCACHE;
#if DUMP_CACHE
if (_options & DUMP_TEXCACHE) {
/* find it on disk */
std::wstring filename = _ident + L"_MEMORYCACHE.dat";
CPath cachepath(stdstr().FromUTF16(_path.c_str()),"");
cachepath.AppendDirectory("cache");
int config = _options & (FILTER_MASK|ENHANCEMENT_MASK|COMPRESS_TEX|COMPRESSION_MASK|FORCE16BPP_TEX|GZ_TEXCACHE);
TxCache::load(stdstr(cachepath).ToUTF16().c_str(), filename.c_str(), config);
}
#endif
}
boolean
TxTexCache::add(uint64 checksum, GHQTexInfo *info)
{
if (_cacheSize <= 0) return 0;
return TxCache::add(checksum, info);
}

39
Source/GlideHQ/TxTexCache.h Normal file
View File

@ -0,0 +1,39 @@
/*
* Texture Filtering
* Version: 1.0
*
* Copyright (C) 2007 Hiroshi Morii All Rights Reserved.
* Email koolsmoky(at)users.sourceforge.net
* Web http://www.3dfxzone.it/koolsmoky
*
* this 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, or (at your option)
* any later version.
*
* this 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 GNU Make; see the file COPYING. If not, write to
* the Free Software Foundation, 675 Mass Ave, Cambridge, MA 02139, USA.
*/
#ifndef __TXTEXCACHE_H__
#define __TXTEXCACHE_H__
#include "TxCache.h"
class TxTexCache : public TxCache
{
public:
~TxTexCache();
TxTexCache(int options, int cachesize, const wchar_t *path, const wchar_t *ident,
dispInfoFuncExt callback);
boolean add(uint64 checksum, /* checksum hi:palette low:texture */
GHQTexInfo *info);
};
#endif /* __TXTEXCACHE_H__ */

1006
Source/GlideHQ/TxUtil.cpp Normal file
View File

File diff suppressed because it is too large Load Diff

121
Source/GlideHQ/TxUtil.h Normal file
View File

@ -0,0 +1,121 @@
/*
* Texture Filtering
* Version: 1.0
*
* Copyright (C) 2007 Hiroshi Morii All Rights Reserved.
* Email koolsmoky(at)users.sourceforge.net
* Web http://www.3dfxzone.it/koolsmoky
*
* this 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, or (at your option)
* any later version.
*
* this 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 GNU Make; see the file COPYING. If not, write to
* the Free Software Foundation, 675 Mass Ave, Cambridge, MA 02139, USA.
*/
#ifndef __TXUTIL_H__
#define __TXUTIL_H__
/* maximum number of CPU cores allowed */
#define MAX_NUMCORE 8
#include "TxInternal.h"
#include <string>
#ifndef DXTN_DLL
#ifdef __cplusplus
extern "C"{
#endif
void tx_compress_dxtn(int srccomps, int width, int height,
const void *source, int destformat, void *dest,
int destRowStride);
int fxt1_encode(int width, int height, int comps,
const void *source, int srcRowStride,
void *dest, int destRowStride);
#ifdef __cplusplus
}
#endif
#endif /* DXTN_DLL */
typedef void (*dxtCompressTexFuncExt)(int srccomps, int width,
int height, const void *srcPixData,
int destformat, void *dest,
int dstRowStride);
typedef int (*fxtCompressTexFuncExt)(int width, int height, int comps,
const void *source, int srcRowStride,
void *dest, int destRowStride);
class TxLoadLib
{
private:
#ifdef DXTN_DLL
HMODULE _dxtnlib;
#endif
fxtCompressTexFuncExt _tx_compress_fxt1;
dxtCompressTexFuncExt _tx_compress_dxtn;
TxLoadLib();
public:
static TxLoadLib* getInstance() {
static TxLoadLib txLoadLib;
return &txLoadLib;
}
~TxLoadLib();
fxtCompressTexFuncExt getfxtCompressTexFuncExt();
dxtCompressTexFuncExt getdxtCompressTexFuncExt();
};
class TxUtil
{
private:
uint32 Adler32(const uint8* data, int Len, uint32 Adler);
uint32 Adler32(const uint8* src, int width, int height, int size, int rowStride);
uint32 RiceCRC32(const uint8* src, int width, int height, int size, int rowStride);
boolean RiceCRC32_CI4(const uint8* src, int width, int height, int size, int rowStride,
uint32* crc32, uint32* cimax);
boolean RiceCRC32_CI8(const uint8* src, int width, int height, int size, int rowStride,
uint32* crc32, uint32* cimax);
int log2(int num);
public:
TxUtil() { }
~TxUtil() { }
int sizeofTx(int width, int height, uint16 format);
uint32 checksumTx(uint8 *data, int width, int height, uint16 format);
#if 0 /* unused */
uint32 chkAlpha(uint32* src, int width, int height);
#endif
uint32 checksum(uint8 *src, int width, int height, int size, int rowStride);
uint64 checksum64(uint8 *src, int width, int height, int size, int rowStride, uint8 *palette);
int grLodLog2(int w, int h);
int grAspectRatioLog2(int w, int h);
int getNumberofProcessors();
};
class TxMemBuf
{
private:
uint8 *_tex[2];
uint32 _size[2];
TxMemBuf();
public:
static TxMemBuf* getInstance() {
static TxMemBuf txMemBuf;
return &txMemBuf;
}
~TxMemBuf();
boolean init(int maxwidth, int maxheight);
void shutdown(void);
uint8 *get(unsigned int num);
uint32 size_of(unsigned int num);
};
#endif /* __TXUTIL_H__ */

0
Source/GlideHQ/bldno.h Normal file
View File

884
Source/GlideHQ/tc-1.1+/dxtn.c Normal file
View File

@ -0,0 +1,884 @@
/*
* DXTn codec
* Version: 1.1
*
* Copyright (C) 2004 Daniel Borca All Rights Reserved.
*
* this 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, or (at your option)
* any later version.
*
* this 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 GNU Make; see the file COPYING. If not, write to
* the Free Software Foundation, 675 Mass Ave, Cambridge, MA 02139, USA.
*/
/* Copyright (C) 2007 Hiroshi Morii <koolsmoky(at)users.sourceforge.net>
* Added support for ARGB inputs, DXT3,5 workaround for ATI Radeons, and
* YUV conversions to determine representative colors.
*/
#include <stdlib.h>
#include <string.h>
#include <assert.h>
#include <stdio.h>
#include "types.h"
#include "internal.h"
#include "dxtn.h"
/***************************************************************************\
* DXTn encoder
*
* The encoder was built by reversing the decoder,
* and is vaguely based on FXT1 codec. Note that this code
* is merely a proof of concept, since it is highly UNoptimized!
\***************************************************************************/
#define MAX_COMP 4 /* ever needed maximum number of components in texel */
#define MAX_VECT 4 /* ever needed maximum number of base vectors to find */
#define N_TEXELS 16 /* number of texels in a block (always 16) */
#define COLOR565(v) (word)((((v)[RCOMP] & 0xf8) << 8) | (((v)[GCOMP] & 0xfc) << 3) | ((v)[BCOMP] >> 3))
static const int dxtn_color_tlat[2][4] = {
{ 0, 2, 3, 1 },
{ 0, 2, 1, 3 }
};
static const int dxtn_alpha_tlat[2][8] = {
{ 0, 2, 3, 4, 5, 6, 7, 1 },
{ 0, 2, 3, 4, 5, 1, 6, 7 }
};
static void
dxt1_rgb_quantize (dword *cc, const byte *lines[], int comps)
{
float b, iv[MAX_COMP]; /* interpolation vector */
dword hi; /* high doubleword */
int color0, color1;
int n_vect;
const int n_comp = 3;
int black = 0;
#ifndef YUV
int minSum = 2000; /* big enough */
#else
int minSum = 2000000;
#endif
int maxSum = -1; /* small enough */
int minCol = 0; /* phoudoin: silent compiler! */
int maxCol = 0; /* phoudoin: silent compiler! */
byte input[N_TEXELS][MAX_COMP];
int i, k, l;
/* make the whole block opaque */
/* we will NEVER reference ACOMP of any pixel */
/* 4 texels each line */
#ifndef ARGB
for (l = 0; l < 4; l++) {
for (k = 0; k < 4; k++) {
for (i = 0; i < comps; i++) {
input[k + l * 4][i] = *lines[l]++;
}
}
}
#else
/* H.Morii - support for ARGB inputs */
for (l = 0; l < 4; l++) {
for (k = 0; k < 4; k++) {
input[k + l * 4][2] = *lines[l]++;
input[k + l * 4][1] = *lines[l]++;
input[k + l * 4][0] = *lines[l]++;
if (comps == 4) input[k + l * 4][3] = *lines[l]++;
}
}
#endif
/* Our solution here is to find the darkest and brightest colors in
* the 4x4 tile and use those as the two representative colors.
* There are probably better algorithms to use (histogram-based).
*/
for (k = 0; k < N_TEXELS; k++) {
int sum = 0;
#ifndef YUV
for (i = 0; i < n_comp; i++) {
sum += input[k][i];
}
#else
/* RGB to YUV conversion according to CCIR 601 specs
* Y = 0.299R+0.587G+0.114B
* U = 0.713(R - Y) = 0.500R-0.419G-0.081B
* V = 0.564(B - Y) = -0.169R-0.331G+0.500B
*/
sum = 299 * input[k][RCOMP] + 587 * input[k][GCOMP] + 114 * input[k][BCOMP];
#endif
if (minSum > sum) {
minSum = sum;
minCol = k;
}
if (maxSum < sum) {
maxSum = sum;
maxCol = k;
}
if (sum == 0) {
black = 1;
}
}
color0 = COLOR565(input[minCol]);
color1 = COLOR565(input[maxCol]);
if (color0 == color1) {
/* we'll use 3-vector */
cc[0] = color0 | (color1 << 16);
hi = black ? -1 : 0;
} else {
if (black && ((color0 == 0) || (color1 == 0))) {
/* we still can use 4-vector */
black = 0;
}
if (black ^ (color0 <= color1)) {
int aux;
aux = color0;
color0 = color1;
color1 = aux;
aux = minCol;
minCol = maxCol;
maxCol = aux;
}
n_vect = (color0 <= color1) ? 2 : 3;
MAKEIVEC(n_vect, n_comp, iv, b, input[minCol], input[maxCol]);
/* add in texels */
cc[0] = color0 | (color1 << 16);
hi = 0;
for (k = N_TEXELS - 1; k >= 0; k--) {
int texel = 3;
int sum = 0;
if (black) {
for (i = 0; i < n_comp; i++) {
sum += input[k][i];
}
}
if (!black || sum) {
/* interpolate color */
CALCCDOT(texel, n_vect, n_comp, iv, b, input[k]);
texel = dxtn_color_tlat[black][texel];
}
/* add in texel */
hi <<= 2;
hi |= texel;
}
}
cc[1] = hi;
}
static void
dxt1_rgba_quantize (dword *cc, const byte *lines[], int comps)
{
float b, iv[MAX_COMP]; /* interpolation vector */
dword hi; /* high doubleword */
int color0, color1;
int n_vect;
const int n_comp = 3;
int transparent = 0;
#ifndef YUV
int minSum = 2000; /* big enough */
#else
int minSum = 2000000;
#endif
int maxSum = -1; /* small enough */
int minCol = 0; /* phoudoin: silent compiler! */
int maxCol = 0; /* phoudoin: silent compiler! */
byte input[N_TEXELS][MAX_COMP];
int i, k, l;
if (comps == 3) {
/* make the whole block opaque */
memset(input, -1, sizeof(input));
}
/* 4 texels each line */
#ifndef ARGB
for (l = 0; l < 4; l++) {
for (k = 0; k < 4; k++) {
for (i = 0; i < comps; i++) {
input[k + l * 4][i] = *lines[l]++;
}
}
}
#else
/* H.Morii - support for ARGB inputs */
for (l = 0; l < 4; l++) {
for (k = 0; k < 4; k++) {
input[k + l * 4][2] = *lines[l]++;
input[k + l * 4][1] = *lines[l]++;
input[k + l * 4][0] = *lines[l]++;
if (comps == 4) input[k + l * 4][3] = *lines[l]++;
}
}
#endif
/* Our solution here is to find the darkest and brightest colors in
* the 4x4 tile and use those as the two representative colors.
* There are probably better algorithms to use (histogram-based).
*/
for (k = 0; k < N_TEXELS; k++) {
int sum = 0;
#ifndef YUV
for (i = 0; i < n_comp; i++) {
sum += input[k][i];
}
#else
sum = 299 * input[k][RCOMP] + 587 * input[k][GCOMP] + 114 * input[k][BCOMP];
#endif
if (minSum > sum) {
minSum = sum;
minCol = k;
}
if (maxSum < sum) {
maxSum = sum;
maxCol = k;
}
if (input[k][ACOMP] < 128) {
transparent = 1;
}
}
color0 = COLOR565(input[minCol]);
color1 = COLOR565(input[maxCol]);
if (color0 == color1) {
/* we'll use 3-vector */
cc[0] = color0 | (color1 << 16);
hi = transparent ? -1 : 0;
} else {
if (transparent ^ (color0 <= color1)) {
int aux;
aux = color0;
color0 = color1;
color1 = aux;
aux = minCol;
minCol = maxCol;
maxCol = aux;
}
n_vect = (color0 <= color1) ? 2 : 3;
MAKEIVEC(n_vect, n_comp, iv, b, input[minCol], input[maxCol]);
/* add in texels */
cc[0] = color0 | (color1 << 16);
hi = 0;
for (k = N_TEXELS - 1; k >= 0; k--) {
int texel = 3;
if (input[k][ACOMP] >= 128) {
/* interpolate color */
CALCCDOT(texel, n_vect, n_comp, iv, b, input[k]);
texel = dxtn_color_tlat[transparent][texel];
}
/* add in texel */
hi <<= 2;
hi |= texel;
}
}
cc[1] = hi;
}
static void
dxt3_rgba_quantize (dword *cc, const byte *lines[], int comps)
{
float b, iv[MAX_COMP]; /* interpolation vector */
dword lolo, lohi; /* low quadword: lo dword, hi dword */
dword hihi; /* high quadword: high dword */
int color0, color1;
const int n_vect = 3;
const int n_comp = 3;
#ifndef YUV
int minSum = 2000; /* big enough */
#else
int minSum = 2000000;
#endif
int maxSum = -1; /* small enough */
int minCol = 0; /* phoudoin: silent compiler! */
int maxCol = 0; /* phoudoin: silent compiler! */
byte input[N_TEXELS][MAX_COMP];
int i, k, l;
if (comps == 3) {
/* make the whole block opaque */
memset(input, -1, sizeof(input));
}
/* 4 texels each line */
#ifndef ARGB
for (l = 0; l < 4; l++) {
for (k = 0; k < 4; k++) {
for (i = 0; i < comps; i++) {
input[k + l * 4][i] = *lines[l]++;
}
}
}
#else
/* H.Morii - support for ARGB inputs */
for (l = 0; l < 4; l++) {
for (k = 0; k < 4; k++) {
input[k + l * 4][2] = *lines[l]++;
input[k + l * 4][1] = *lines[l]++;
input[k + l * 4][0] = *lines[l]++;
if (comps == 4) input[k + l * 4][3] = *lines[l]++;
}
}
#endif
/* Our solution here is to find the darkest and brightest colors in
* the 4x4 tile and use those as the two representative colors.
* There are probably better algorithms to use (histogram-based).
*/
for (k = 0; k < N_TEXELS; k++) {
int sum = 0;
#ifndef YUV
for (i = 0; i < n_comp; i++) {
sum += input[k][i];
}
#else
sum = 299 * input[k][RCOMP] + 587 * input[k][GCOMP] + 114 * input[k][BCOMP];
#endif
if (minSum > sum) {
minSum = sum;
minCol = k;
}
if (maxSum < sum) {
maxSum = sum;
maxCol = k;
}
}
/* add in alphas */
lolo = lohi = 0;
for (k = N_TEXELS - 1; k >= N_TEXELS / 2; k--) {
/* add in alpha */
lohi <<= 4;
lohi |= input[k][ACOMP] >> 4;
}
cc[1] = lohi;
for (; k >= 0; k--) {
/* add in alpha */
lolo <<= 4;
lolo |= input[k][ACOMP] >> 4;
}
cc[0] = lolo;
color0 = COLOR565(input[minCol]);
color1 = COLOR565(input[maxCol]);
#ifdef RADEON
/* H.Morii - Workaround for ATI Radeon
* According to the OpenGL EXT_texture_compression_s3tc specs,
* the encoding of the RGB components for DXT3 and DXT5 formats
* use the non-transparent encodings of DXT1 but treated as
* though color0 > color1, regardless of the actual values of
* color0 and color1. ATI Radeons however require the values to
* be color0 > color1.
*/
if (color0 < color1) {
int aux;
aux = color0;
color0 = color1;
color1 = aux;
aux = minCol;
minCol = maxCol;
maxCol = aux;
}
#endif
cc[2] = color0 | (color1 << 16);
hihi = 0;
if (color0 != color1) {
MAKEIVEC(n_vect, n_comp, iv, b, input[minCol], input[maxCol]);
/* add in texels */
for (k = N_TEXELS - 1; k >= 0; k--) {
int texel;
/* interpolate color */
CALCCDOT(texel, n_vect, n_comp, iv, b, input[k]);
texel = dxtn_color_tlat[0][texel];
/* add in texel */
hihi <<= 2;
hihi |= texel;
}
}
cc[3] = hihi;
}
static void
dxt5_rgba_quantize (dword *cc, const byte *lines[], int comps)
{
float b, iv[MAX_COMP]; /* interpolation vector */
qword lo; /* low quadword */
dword hihi; /* high quadword: high dword */
int color0, color1;
const int n_vect = 3;
const int n_comp = 3;
#ifndef YUV
int minSum = 2000; /* big enough */
#else
int minSum = 2000000;
#endif
int maxSum = -1; /* small enough */
int minCol = 0; /* phoudoin: silent compiler! */
int maxCol = 0; /* phoudoin: silent compiler! */
int alpha0 = 2000; /* big enough */
int alpha1 = -1; /* small enough */
int anyZero = 0, anyOne = 0;
int a_vect;
byte input[N_TEXELS][MAX_COMP];
int i, k, l;
if (comps == 3) {
/* make the whole block opaque */
memset(input, -1, sizeof(input));
}
/* 4 texels each line */
#ifndef ARGB
for (l = 0; l < 4; l++) {
for (k = 0; k < 4; k++) {
for (i = 0; i < comps; i++) {
input[k + l * 4][i] = *lines[l]++;
}
}
}
#else
/* H.Morii - support for ARGB inputs */
for (l = 0; l < 4; l++) {
for (k = 0; k < 4; k++) {
input[k + l * 4][2] = *lines[l]++;
input[k + l * 4][1] = *lines[l]++;
input[k + l * 4][0] = *lines[l]++;
if (comps == 4) input[k + l * 4][3] = *lines[l]++;
}
}
#endif
/* Our solution here is to find the darkest and brightest colors in
* the 4x4 tile and use those as the two representative colors.
* There are probably better algorithms to use (histogram-based).
*/
for (k = 0; k < N_TEXELS; k++) {
int sum = 0;
#ifndef YUV
for (i = 0; i < n_comp; i++) {
sum += input[k][i];
}
#else
sum = 299 * input[k][RCOMP] + 587 * input[k][GCOMP] + 114 * input[k][BCOMP];
#endif
if (minSum > sum) {
minSum = sum;
minCol = k;
}
if (maxSum < sum) {
maxSum = sum;
maxCol = k;
}
if (alpha0 > input[k][ACOMP]) {
alpha0 = input[k][ACOMP];
}
if (alpha1 < input[k][ACOMP]) {
alpha1 = input[k][ACOMP];
}
if (input[k][ACOMP] == 0) {
anyZero = 1;
}
if (input[k][ACOMP] == 255) {
anyOne = 1;
}
}
/* add in alphas */
if (alpha0 == alpha1) {
/* we'll use 6-vector */
cc[0] = alpha0 | (alpha1 << 8);
cc[1] = 0;
} else {
if (anyZero && ((alpha0 == 0) || (alpha1 == 0))) {
/* we still might use 8-vector */
anyZero = 0;
}
if (anyOne && ((alpha0 == 255) || (alpha1 == 255))) {
/* we still might use 8-vector */
anyOne = 0;
}
if ((anyZero | anyOne) ^ (alpha0 <= alpha1)) {
int aux;
aux = alpha0;
alpha0 = alpha1;
alpha1 = aux;
}
a_vect = (alpha0 <= alpha1) ? 5 : 7;
/* compute interpolation vector */
iv[ACOMP] = (float)a_vect / (alpha1 - alpha0);
b = -iv[ACOMP] * alpha0 + 0.5F;
/* add in alphas */
Q_MOV32(lo, 0);
for (k = N_TEXELS - 1; k >= 0; k--) {
int texel = -1;
if (anyZero | anyOne) {
if (input[k][ACOMP] == 0) {
texel = 6;
} else if (input[k][ACOMP] == 255) {
texel = 7;
}
}
/* interpolate alpha */
if (texel == -1) {
float dot = input[k][ACOMP] * iv[ACOMP];
texel = (int)(dot + b);
#if SAFECDOT
if (texel < 0) {
texel = 0;
} else if (texel > a_vect) {
texel = a_vect;
}
#endif
texel = dxtn_alpha_tlat[anyZero | anyOne][texel];
}
/* add in texel */
Q_SHL(lo, 3);
Q_OR32(lo, texel);
}
Q_SHL(lo, 16);
Q_OR32(lo, alpha0 | (alpha1 << 8));
((qword *)cc)[0] = lo;
}
color0 = COLOR565(input[minCol]);
color1 = COLOR565(input[maxCol]);
#ifdef RADEON /* H.Morii - Workaround for ATI Radeon */
if (color0 < color1) {
int aux;
aux = color0;
color0 = color1;
color1 = aux;
aux = minCol;
minCol = maxCol;
maxCol = aux;
}
#endif
cc[2] = color0 | (color1 << 16);
hihi = 0;
if (color0 != color1) {
MAKEIVEC(n_vect, n_comp, iv, b, input[minCol], input[maxCol]);
/* add in texels */
for (k = N_TEXELS - 1; k >= 0; k--) {
int texel;
/* interpolate color */
CALCCDOT(texel, n_vect, n_comp, iv, b, input[k]);
texel = dxtn_color_tlat[0][texel];
/* add in texel */
hihi <<= 2;
hihi |= texel;
}
}
cc[3] = hihi;
}
#define ENCODER(dxtn, n) \
int TAPIENTRY \
dxtn##_encode (int width, int height, int comps, \
const void *source, int srcRowStride, \
void *dest, int destRowStride) \
{ \
int x, y; \
const byte *data; \
dword *encoded = (dword *)dest; \
void *newSource = NULL; \
\
/* Replicate image if width is not M4 or height is not M4 */ \
if ((width & 3) | (height & 3)) { \
int newWidth = (width + 3) & ~3; \
int newHeight = (height + 3) & ~3; \
newSource = malloc(comps * newWidth * newHeight * sizeof(byte *));\
_mesa_upscale_teximage2d(width, height, newWidth, newHeight, \
comps, (const byte *)source, \
srcRowStride, (byte *)newSource); \
source = newSource; \
width = newWidth; \
height = newHeight; \
srcRowStride = comps * newWidth; \
} \
\
data = (const byte *)source; \
destRowStride = (destRowStride - width * n) / 4; \
for (y = 0; y < height; y += 4) { \
unsigned int offs = 0 + (y + 0) * srcRowStride; \
for (x = 0; x < width; x += 4) { \
const byte *lines[4]; \
lines[0] = &data[offs]; \
lines[1] = lines[0] + srcRowStride; \
lines[2] = lines[1] + srcRowStride; \
lines[3] = lines[2] + srcRowStride; \
offs += 4 * comps; \
dxtn##_quantize(encoded, lines, comps); \
/* 4x4 block */ \
encoded += n; \
} \
encoded += destRowStride; \
} \
\
if (newSource != NULL) { \
free(newSource); \
} \
\
return 0; \
}
ENCODER(dxt1_rgb, 2)
ENCODER(dxt1_rgba, 2)
ENCODER(dxt3_rgba, 4)
ENCODER(dxt5_rgba, 4)
/***************************************************************************\
* DXTn decoder
*
* The decoder is based on GL_EXT_texture_compression_s3tc
* specification and serves as a concept for the encoder.
\***************************************************************************/
/* lookup table for scaling 4 bit colors up to 8 bits */
static const byte _rgb_scale_4[] = {
0, 17, 34, 51, 68, 85, 102, 119,
136, 153, 170, 187, 204, 221, 238, 255
};
/* lookup table for scaling 5 bit colors up to 8 bits */
static const byte _rgb_scale_5[] = {
0, 8, 16, 25, 33, 41, 49, 58,
66, 74, 82, 90, 99, 107, 115, 123,
132, 140, 148, 156, 165, 173, 181, 189,
197, 206, 214, 222, 230, 239, 247, 255
};
/* lookup table for scaling 6 bit colors up to 8 bits */
static const byte _rgb_scale_6[] = {
0, 4, 8, 12, 16, 20, 24, 28,
32, 36, 40, 45, 49, 53, 57, 61,
65, 69, 73, 77, 81, 85, 89, 93,
97, 101, 105, 109, 113, 117, 121, 125,
130, 134, 138, 142, 146, 150, 154, 158,
162, 166, 170, 174, 178, 182, 186, 190,
194, 198, 202, 206, 210, 215, 219, 223,
227, 231, 235, 239, 243, 247, 251, 255
};
#define CC_SEL(cc, which) (((dword *)(cc))[(which) / 32] >> ((which) & 31))
#define UP4(c) _rgb_scale_4[(c) & 15]
#define UP5(c) _rgb_scale_5[(c) & 31]
#define UP6(c) _rgb_scale_6[(c) & 63]
#define ZERO_4UBV(v) *((dword *)(v)) = 0
void TAPIENTRY
dxt1_rgb_decode_1 (const void *texture, int stride,
int i, int j, byte *rgba)
{
const byte *src = (const byte *)texture
+ ((j / 4) * ((stride + 3) / 4) + i / 4) * 8;
const int code = (src[4 + (j & 3)] >> ((i & 3) * 2)) & 0x3;
if (code == 0) {
rgba[RCOMP] = UP5(CC_SEL(src, 11));
rgba[GCOMP] = UP6(CC_SEL(src, 5));
rgba[BCOMP] = UP5(CC_SEL(src, 0));
} else if (code == 1) {
rgba[RCOMP] = UP5(CC_SEL(src, 27));
rgba[GCOMP] = UP6(CC_SEL(src, 21));
rgba[BCOMP] = UP5(CC_SEL(src, 16));
} else {
const word col0 = src[0] | (src[1] << 8);
const word col1 = src[2] | (src[3] << 8);
if (col0 > col1) {
if (code == 2) {
rgba[RCOMP] = (UP5(col0 >> 11) * 2 + UP5(col1 >> 11)) / 3;
rgba[GCOMP] = (UP6(col0 >> 5) * 2 + UP6(col1 >> 5)) / 3;
rgba[BCOMP] = (UP5(col0 ) * 2 + UP5(col1 )) / 3;
} else {
rgba[RCOMP] = (UP5(col0 >> 11) + 2 * UP5(col1 >> 11)) / 3;
rgba[GCOMP] = (UP6(col0 >> 5) + 2 * UP6(col1 >> 5)) / 3;
rgba[BCOMP] = (UP5(col0 ) + 2 * UP5(col1 )) / 3;
}
} else {
if (code == 2) {
rgba[RCOMP] = (UP5(col0 >> 11) + UP5(col1 >> 11)) / 2;
rgba[GCOMP] = (UP6(col0 >> 5) + UP6(col1 >> 5)) / 2;
rgba[BCOMP] = (UP5(col0 ) + UP5(col1 )) / 2;
} else {
ZERO_4UBV(rgba);
}
}
}
rgba[ACOMP] = 255;
}
void TAPIENTRY
dxt1_rgba_decode_1 (const void *texture, int stride,
int i, int j, byte *rgba)
{
/* Same as rgb_dxt1 above, except alpha=0 if col0<=col1 and code=3. */
const byte *src = (const byte *)texture
+ ((j / 4) * ((stride + 3) / 4) + i / 4) * 8;
const int code = (src[4 + (j & 3)] >> ((i & 3) * 2)) & 0x3;
if (code == 0) {
rgba[RCOMP] = UP5(CC_SEL(src, 11));
rgba[GCOMP] = UP6(CC_SEL(src, 5));
rgba[BCOMP] = UP5(CC_SEL(src, 0));
rgba[ACOMP] = 255;
} else if (code == 1) {
rgba[RCOMP] = UP5(CC_SEL(src, 27));
rgba[GCOMP] = UP6(CC_SEL(src, 21));
rgba[BCOMP] = UP5(CC_SEL(src, 16));
rgba[ACOMP] = 255;
} else {
const word col0 = src[0] | (src[1] << 8);
const word col1 = src[2] | (src[3] << 8);
if (col0 > col1) {
if (code == 2) {
rgba[RCOMP] = (UP5(col0 >> 11) * 2 + UP5(col1 >> 11)) / 3;
rgba[GCOMP] = (UP6(col0 >> 5) * 2 + UP6(col1 >> 5)) / 3;
rgba[BCOMP] = (UP5(col0 ) * 2 + UP5(col1 )) / 3;
} else {
rgba[RCOMP] = (UP5(col0 >> 11) + 2 * UP5(col1 >> 11)) / 3;
rgba[GCOMP] = (UP6(col0 >> 5) + 2 * UP6(col1 >> 5)) / 3;
rgba[BCOMP] = (UP5(col0 ) + 2 * UP5(col1 )) / 3;
}
rgba[ACOMP] = 255;
} else {
if (code == 2) {
rgba[RCOMP] = (UP5(col0 >> 11) + UP5(col1 >> 11)) / 2;
rgba[GCOMP] = (UP6(col0 >> 5) + UP6(col1 >> 5)) / 2;
rgba[BCOMP] = (UP5(col0 ) + UP5(col1 )) / 2;
rgba[ACOMP] = 255;
} else {
ZERO_4UBV(rgba);
}
}
}
}
void TAPIENTRY
dxt3_rgba_decode_1 (const void *texture, int stride,
int i, int j, byte *rgba)
{
const byte *src = (const byte *)texture
+ ((j / 4) * ((stride + 3) / 4) + i / 4) * 16;
const int code = (src[12 + (j & 3)] >> ((i & 3) * 2)) & 0x3;
const dword *cc = (const dword *)(src + 8);
if (code == 0) {
rgba[RCOMP] = UP5(CC_SEL(cc, 11));
rgba[GCOMP] = UP6(CC_SEL(cc, 5));
rgba[BCOMP] = UP5(CC_SEL(cc, 0));
} else if (code == 1) {
rgba[RCOMP] = UP5(CC_SEL(cc, 27));
rgba[GCOMP] = UP6(CC_SEL(cc, 21));
rgba[BCOMP] = UP5(CC_SEL(cc, 16));
} else if (code == 2) {
/* (col0 * (4 - code) + col1 * (code - 1)) / 3 */
rgba[RCOMP] = (UP5(CC_SEL(cc, 11)) * 2 + UP5(CC_SEL(cc, 27))) / 3;
rgba[GCOMP] = (UP6(CC_SEL(cc, 5)) * 2 + UP6(CC_SEL(cc, 21))) / 3;
rgba[BCOMP] = (UP5(CC_SEL(cc, 0)) * 2 + UP5(CC_SEL(cc, 16))) / 3;
} else {
rgba[RCOMP] = (UP5(CC_SEL(cc, 11)) + 2 * UP5(CC_SEL(cc, 27))) / 3;
rgba[GCOMP] = (UP6(CC_SEL(cc, 5)) + 2 * UP6(CC_SEL(cc, 21))) / 3;
rgba[BCOMP] = (UP5(CC_SEL(cc, 0)) + 2 * UP5(CC_SEL(cc, 16))) / 3;
}
rgba[ACOMP] = UP4(src[((j & 3) * 4 + (i & 3)) / 2] >> ((i & 1) * 4));
}
void TAPIENTRY
dxt5_rgba_decode_1 (const void *texture, int stride,
int i, int j, byte *rgba)
{
const byte *src = (const byte *)texture
+ ((j / 4) * ((stride + 3) / 4) + i / 4) * 16;
const int code = (src[12 + (j & 3)] >> ((i & 3) * 2)) & 0x3;
const dword *cc = (const dword *)(src + 8);
const byte alpha0 = src[0];
const byte alpha1 = src[1];
const int alphaShift = (((j & 3) * 4) + (i & 3)) * 3 + 16;
const int acode = ((alphaShift == 31)
? CC_SEL(src + 2, alphaShift - 16)
: CC_SEL(src, alphaShift)) & 0x7;
if (code == 0) {
rgba[RCOMP] = UP5(CC_SEL(cc, 11));
rgba[GCOMP] = UP6(CC_SEL(cc, 5));
rgba[BCOMP] = UP5(CC_SEL(cc, 0));
} else if (code == 1) {
rgba[RCOMP] = UP5(CC_SEL(cc, 27));
rgba[GCOMP] = UP6(CC_SEL(cc, 21));
rgba[BCOMP] = UP5(CC_SEL(cc, 16));
} else if (code == 2) {
/* (col0 * (4 - code) + col1 * (code - 1)) / 3 */
rgba[RCOMP] = (UP5(CC_SEL(cc, 11)) * 2 + UP5(CC_SEL(cc, 27))) / 3;
rgba[GCOMP] = (UP6(CC_SEL(cc, 5)) * 2 + UP6(CC_SEL(cc, 21))) / 3;
rgba[BCOMP] = (UP5(CC_SEL(cc, 0)) * 2 + UP5(CC_SEL(cc, 16))) / 3;
} else {
rgba[RCOMP] = (UP5(CC_SEL(cc, 11)) + 2 * UP5(CC_SEL(cc, 27))) / 3;
rgba[GCOMP] = (UP6(CC_SEL(cc, 5)) + 2 * UP6(CC_SEL(cc, 21))) / 3;
rgba[BCOMP] = (UP5(CC_SEL(cc, 0)) + 2 * UP5(CC_SEL(cc, 16))) / 3;
}
if (acode == 0) {
rgba[ACOMP] = alpha0;
} else if (acode == 1) {
rgba[ACOMP] = alpha1;
} else if (alpha0 > alpha1) {
rgba[ACOMP] = ((8 - acode) * alpha0 + (acode - 1) * alpha1) / 7;
} else if (acode == 6) {
rgba[ACOMP] = 0;
} else if (acode == 7) {
rgba[ACOMP] = 255;
} else {
rgba[ACOMP] = ((6 - acode) * alpha0 + (acode - 1) * alpha1) / 5;
}
}

62
Source/GlideHQ/tc-1.1+/dxtn.h Normal file
View File

@ -0,0 +1,62 @@
/*
* DXTn codec
* Version: 1.1
*
* Copyright (C) 2004 Daniel Borca All Rights Reserved.
*
* this 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, or (at your option)
* any later version.
*
* this 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 GNU Make; see the file COPYING. If not, write to
* the Free Software Foundation, 675 Mass Ave, Cambridge, MA 02139, USA.
*/
#ifndef DXTN_H_included
#define DXTN_H_included
TAPI int TAPIENTRY
dxt1_rgb_encode (int width, int height, int comps,
const void *source, int srcRowStride,
void *dest, int destRowStride);
TAPI int TAPIENTRY
dxt1_rgba_encode (int width, int height, int comps,
const void *source, int srcRowStride,
void *dest, int destRowStride);
TAPI int TAPIENTRY
dxt3_rgba_encode (int width, int height, int comps,
const void *source, int srcRowStride,
void *dest, int destRowStride);
TAPI int TAPIENTRY
dxt5_rgba_encode (int width, int height, int comps,
const void *source, int srcRowStride,
void *dest, int destRowStride);
TAPI void TAPIENTRY
dxt1_rgb_decode_1 (const void *texture, int stride /* in pixels */,
int i, int j, byte *rgba);
TAPI void TAPIENTRY
dxt1_rgba_decode_1 (const void *texture, int stride /* in pixels */,
int i, int j, byte *rgba);
TAPI void TAPIENTRY
dxt3_rgba_decode_1 (const void *texture, int stride /* in pixels */,
int i, int j, byte *rgba);
TAPI void TAPIENTRY
dxt5_rgba_decode_1 (const void *texture, int stride /* in pixels */,
int i, int j, byte *rgba);
#endif

1459
Source/GlideHQ/tc-1.1+/fxt1.c Normal file
View File

File diff suppressed because it is too large Load Diff

38
Source/GlideHQ/tc-1.1+/fxt1.h Normal file
View File

@ -0,0 +1,38 @@
/*
* FXT1 codec
* Version: 1.1
*
* Copyright (C) 2004 Daniel Borca All Rights Reserved.
*
* Permission is hereby granted, free of charge, to any person obtaining a
* copy of this software and associated documentation files (the "Software"),
* to deal in the Software without restriction, including without limitation
* the rights to use, copy, modify, merge, publish, distribute, sublicense,
* and/or sell copies of the Software, and to permit persons to whom the
* Software is furnished to do so, subject to the following conditions:
*
* The above copyright notice and this permission notice shall be included
* in all copies or substantial portions of the Software.
*
* THE SOFTWARE IS PROVIDED "AS IS", WITHOUT WARRANTY OF ANY KIND, EXPRESS
* OR IMPLIED, INCLUDING BUT NOT LIMITED TO THE WARRANTIES OF MERCHANTABILITY,
* FITNESS FOR A PARTICULAR PURPOSE AND NONINFRINGEMENT. IN NO EVENT SHALL
* DANIEL BORCA BE LIABLE FOR ANY CLAIM, DAMAGES OR OTHER LIABILITY, WHETHER IN
* AN ACTION OF CONTRACT, TORT OR OTHERWISE, ARISING FROM, OUT OF OR IN
* CONNECTION WITH THE SOFTWARE OR THE USE OR OTHER DEALINGS IN THE SOFTWARE.
*/
#ifndef FXT1_H_included
#define FXT1_H_included
TAPI int TAPIENTRY
fxt1_encode (int width, int height, int comps,
const void *source, int srcRowStride,
void *dest, int destRowStride);
TAPI void TAPIENTRY
fxt1_decode_1 (const void *texture, int stride /* in pixels */,
int i, int j, byte *rgba);
#endif

137
Source/GlideHQ/tc-1.1+/internal.h Normal file
View File

@ -0,0 +1,137 @@
/*
* Texture compression
* Version: 1.0
*
* Copyright (C) 2004 Daniel Borca All Rights Reserved.
*
* this 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, or (at your option)
* any later version.
*
* this 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 GNU Make; see the file COPYING. If not, write to
* the Free Software Foundation, 675 Mass Ave, Cambridge, MA 02139, USA.
*/
#ifndef INTERNAL_H_included
#define INTERNAL_H_included
/*****************************************************************************\
* DLL stuff
\*****************************************************************************/
#ifdef __WIN32__
#define TAPI __declspec(dllexport)
#define TAPIENTRY /*__stdcall*/
#else
#define TAPI
#define TAPIENTRY
#endif
/*****************************************************************************\
* 64bit types on 32bit machine
\*****************************************************************************/
#if (defined(__GNUC__) && !defined(__cplusplus)) || defined(WIN32)
typedef unsigned long long qword;
#define Q_MOV32(a, b) a = b
#define Q_OR32(a, b) a |= b
#define Q_SHL(a, c) a <<= c
#else /* !__GNUC__ */
typedef struct {
dword lo, hi;
} qword;
#define Q_MOV32(a, b) a.lo = b
#define Q_OR32(a, b) a.lo |= b
#define Q_SHL(a, c) \
do { \
if ((c) >= 32) { \
a.hi = a.lo << ((c) - 32); \
a.lo = 0; \
} else { \
a.hi = (a.hi << (c)) | (a.lo >> (32 - (c)));\
a.lo <<= c; \
} \
} while (0)
#endif /* !__GNUC__ */
/*****************************************************************************\
* Config
\*****************************************************************************/
#define RCOMP 0
#define GCOMP 1
#define BCOMP 2
#define ACOMP 3
/*****************************************************************************\
* Metric
\*****************************************************************************/
#define F(i) (float)1 /* can be used to obtain an oblong metric: 0.30 / 0.59 / 0.11 */
#define SAFECDOT 1 /* for paranoids */
#define MAKEIVEC(NV, NC, IV, B, V0, V1) \
do { \
/* compute interpolation vector */\
float d2 = 0.0F; \
float rd2; \
\
for (i = 0; i < NC; i++) { \
IV[i] = (V1[i] - V0[i]) * F(i);\
d2 += IV[i] * IV[i]; \
} \
rd2 = (float)NV / d2; \
B = 0; \
for (i = 0; i < NC; i++) { \
IV[i] *= F(i); \
B -= IV[i] * V0[i]; \
IV[i] *= rd2; \
} \
B = B * rd2 + 0.5F; \
} while (0)
#define CALCCDOT(TEXEL, NV, NC, IV, B, V)\
do { \
float dot = 0.0F; \
for (i = 0; i < NC; i++) { \
dot += V[i] * IV[i]; \
} \
TEXEL = (int)(dot + B); \
if (SAFECDOT) { \
if (TEXEL < 0) { \
TEXEL = 0; \
} else if (TEXEL > NV) { \
TEXEL = NV; \
} \
} \
} while (0)
/*****************************************************************************\
* Utility functions
\*****************************************************************************/
void
_mesa_upscale_teximage2d (unsigned int inWidth, unsigned int inHeight,
unsigned int outWidth, unsigned int outHeight,
unsigned int comps,
const byte *src, int srcRowStride,
unsigned char *dest);
#endif

93
Source/GlideHQ/tc-1.1+/texstore.c Normal file
View File

@ -0,0 +1,93 @@
/*
* Mesa 3-D graphics library
* Version: 6.3
*
* Copyright (C) 1999-2004 Brian Paul All Rights Reserved.
*
* Permission is hereby granted, free of charge, to any person obtaining a
* copy of this software and associated documentation files (the "Software"),
* to deal in the Software without restriction, including without limitation
* the rights to use, copy, modify, merge, publish, distribute, sublicense,
* and/or sell copies of the Software, and to permit persons to whom the
* Software is furnished to do so, subject to the following conditions:
*
* The above copyright notice and this permission notice shall be included
* in all copies or substantial portions of the Software.
*
* THE SOFTWARE IS PROVIDED "AS IS", WITHOUT WARRANTY OF ANY KIND, EXPRESS
* OR IMPLIED, INCLUDING BUT NOT LIMITED TO THE WARRANTIES OF MERCHANTABILITY,
* FITNESS FOR A PARTICULAR PURPOSE AND NONINFRINGEMENT. IN NO EVENT SHALL
* BRIAN PAUL BE LIABLE FOR ANY CLAIM, DAMAGES OR OTHER LIABILITY, WHETHER IN
* AN ACTION OF CONTRACT, TORT OR OTHERWISE, ARISING FROM, OUT OF OR IN
* CONNECTION WITH THE SOFTWARE OR THE USE OR OTHER DEALINGS IN THE SOFTWARE.
*/
/* Copyright (C) 2007 Hiroshi Morii <koolsmoky(at)users.sourceforge.net>
* _mesa_upscale_teximage2d speedup
*/
#include <assert.h>
#include "types.h"
#include "internal.h"
void
_mesa_upscale_teximage2d (unsigned int inWidth, unsigned int inHeight,
unsigned int outWidth, unsigned int outHeight,
unsigned int comps,
const byte *src, int srcRowStride,
byte *dest)
{
unsigned int i, j, k;
assert(outWidth >= inWidth);
assert(outHeight >= inHeight);
#if 1 /* H.Morii - faster loops */
for (i = 0; i < inHeight; i++) {
for (j = 0; j < inWidth; j++) {
const int aa = (i * outWidth + j) * comps;
const int bb = i * srcRowStride + j * comps;
for (k = 0; k < comps; k++) {
dest[aa + k] = src[bb + k];
}
}
for (; j < outWidth; j++) {
const int aa = (i * outWidth + j) * comps;
const int bb = i * srcRowStride + (j - inWidth) * comps;
for (k = 0; k < comps; k++) {
dest[aa + k] = src[bb + k];
}
}
}
for (; i < outHeight; i++) {
for (j = 0; j < inWidth; j++) {
const int aa = (i * outWidth + j) * comps;
const int bb = (i - inHeight) * srcRowStride + j * comps;
for (k = 0; k < comps; k++) {
dest[aa + k] = src[bb + k];
}
}
for (; j < outWidth; j++) {
const int aa = (i * outWidth + j) * comps;
const int bb = (i - inHeight) * srcRowStride + (j - inWidth) * comps;
for (k = 0; k < comps; k++) {
dest[aa + k] = src[bb + k];
}
}
}
#else
for (i = 0; i < outHeight; i++) {
const int ii = i % inHeight;
for (j = 0; j < outWidth; j++) {
const int jj = j % inWidth;
const int aa = (i * outWidth + j) * comps;
const int bb = ii * srcRowStride + jj * comps;
for (k = 0; k < comps; k++) {
dest[aa + k] = src[bb + k];
}
}
}
#endif
}

11
Source/GlideHQ/tc-1.1+/types.h Normal file
View File

@ -0,0 +1,11 @@
#ifndef TYPES_H_included
#define TYPES_H_included
/*****************************************************************************\
* 32bit types
\*****************************************************************************/
typedef unsigned char byte; /* 8-bit */
typedef unsigned short word; /* 16-bit */
typedef unsigned int dword; /* 32-bit */
#endif

110
Source/GlideHQ/tc-1.1+/wrapper.c Normal file
View File

@ -0,0 +1,110 @@
/*
* Texture compression
* Version: 1.0
*
* Copyright (C) 2004 Daniel Borca All Rights Reserved.
*
* this 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, or (at your option)
* any later version.
*
* this 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 GNU Make; see the file COPYING. If not, write to
* the Free Software Foundation, 675 Mass Ave, Cambridge, MA 02139, USA.
*/
#include <assert.h>
#include "types.h"
#include "internal.h"
#include "dxtn.h"
#define GL_COMPRESSED_RGB_S3TC_DXT1_EXT 0x83F0
#define GL_COMPRESSED_RGBA_S3TC_DXT1_EXT 0x83F1
#define GL_COMPRESSED_RGBA_S3TC_DXT3_EXT 0x83F2
#define GL_COMPRESSED_RGBA_S3TC_DXT5_EXT 0x83F3
TAPI void TAPIENTRY
fetch_2d_texel_rgb_dxt1 (int texImage_RowStride,
const byte *texImage_Data,
int i, int j,
byte *texel)
{
dxt1_rgb_decode_1(texImage_Data, texImage_RowStride, i, j, texel);
}
TAPI void TAPIENTRY
fetch_2d_texel_rgba_dxt1 (int texImage_RowStride,
const byte *texImage_Data,
int i, int j,
byte *texel)
{
dxt1_rgba_decode_1(texImage_Data, texImage_RowStride, i, j, texel);
}
TAPI void TAPIENTRY
fetch_2d_texel_rgba_dxt3 (int texImage_RowStride,
const byte *texImage_Data,
int i, int j,
byte *texel)
{
dxt3_rgba_decode_1(texImage_Data, texImage_RowStride, i, j, texel);
}
TAPI void TAPIENTRY
fetch_2d_texel_rgba_dxt5 (int texImage_RowStride,
const byte *texImage_Data,
int i, int j,
byte *texel)
{
dxt5_rgba_decode_1(texImage_Data, texImage_RowStride, i, j, texel);
}
TAPI void TAPIENTRY
tx_compress_dxtn (int srccomps, int width, int height,
const byte *source, int destformat, byte *dest,
int destRowStride)
{
int srcRowStride = width * srccomps;
int rv;
switch (destformat) {
case GL_COMPRESSED_RGB_S3TC_DXT1_EXT:
rv = dxt1_rgb_encode(width, height, srccomps,
source, srcRowStride,
dest, destRowStride);
break;
case GL_COMPRESSED_RGBA_S3TC_DXT1_EXT:
rv = dxt1_rgba_encode(width, height, srccomps,
source, srcRowStride,
dest, destRowStride);
break;
case GL_COMPRESSED_RGBA_S3TC_DXT3_EXT:
rv = dxt3_rgba_encode(width, height, srccomps,
source, srcRowStride,
dest, destRowStride);
break;
case GL_COMPRESSED_RGBA_S3TC_DXT5_EXT:
rv = dxt5_rgba_encode(width, height, srccomps,
source, srcRowStride,
dest, destRowStride);
break;
default:
assert(0);
}
/*return rv;*/
}