/* =========================================================================== Doom 3 BFG Edition GPL Source Code Copyright (C) 1993-2012 id Software LLC, a ZeniMax Media company. This file is part of the Doom 3 BFG Edition GPL Source Code ("Doom 3 BFG Edition Source Code"). Doom 3 BFG Edition Source Code 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 3 of the License, or (at your option) any later version. Doom 3 BFG Edition Source Code 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 Doom 3 BFG Edition Source Code. If not, see . In addition, the Doom 3 BFG Edition Source Code is also subject to certain additional terms. You should have received a copy of these additional terms immediately following the terms and conditions of the GNU General Public License which accompanied the Doom 3 BFG Edition Source Code. If not, please request a copy in writing from id Software at the address below. If you have questions concerning this license or the applicable additional terms, you may contact in writing id Software LLC, c/o ZeniMax Media Inc., Suite 120, Rockville, Maryland 20850 USA. =========================================================================== */ #pragma hdrstop #include "precompiled.h" /* ================================================================================================ idBinaryImage ================================================================================================ */ #include "tr_local.h" #include "DXT/DXTCodec.h" #include "Color/ColorSpace.h" idCVar image_highQualityCompression( "image_highQualityCompression", "0", CVAR_BOOL, "Use high quality (slow) compression" ); /* ======================== idBinaryImage::Load2DFromMemory ======================== */ void idBinaryImage::Load2DFromMemory( int width, int height, const byte* pic_const, int numLevels, textureFormat_t& textureFormat, textureColor_t& colorFormat, bool gammaMips ) { fileData.textureType = TT_2D; fileData.format = textureFormat; fileData.colorFormat = colorFormat; fileData.width = width; fileData.height = height; fileData.numLevels = numLevels; byte* pic = ( byte* )Mem_Alloc( width * height * 4, TAG_TEMP ); memcpy( pic, pic_const, width * height * 4 ); if( colorFormat == CFM_YCOCG_DXT5 ) { // convert the image data to YCoCg and use the YCoCgDXT5 compressor idColorSpace::ConvertRGBToCoCg_Y( pic, pic, width, height ); } else if( colorFormat == CFM_NORMAL_DXT5 ) { // Blah, HQ swizzles automatically, Fast doesn't if( !image_highQualityCompression.GetBool() ) { for( int i = 0; i < width * height; i++ ) { pic[i * 4 + 3] = pic[i * 4 + 0]; pic[i * 4 + 0] = 0; pic[i * 4 + 2] = 0; } } } else if( colorFormat == CFM_GREEN_ALPHA ) { for( int i = 0; i < width * height; i++ ) { pic[i * 4 + 1] = pic[i * 4 + 3]; pic[i * 4 + 0] = 0; pic[i * 4 + 2] = 0; pic[i * 4 + 3] = 0; } } int scaledWidth = width; int scaledHeight = height; images.SetNum( numLevels ); for( int level = 0; level < images.Num(); level++ ) { idBinaryImageData& img = images[ level ]; // Images that are going to be DXT compressed and aren't multiples of 4 need to be // padded out before compressing. byte* dxtPic = pic; int dxtWidth = 0; int dxtHeight = 0; if( textureFormat == FMT_DXT5 || textureFormat == FMT_DXT1 ) { if( ( scaledWidth & 3 ) || ( scaledHeight & 3 ) ) { dxtWidth = ( scaledWidth + 3 ) & ~3; dxtHeight = ( scaledHeight + 3 ) & ~3; dxtPic = ( byte* )Mem_ClearedAlloc( dxtWidth * 4 * dxtHeight, TAG_IMAGE ); for( int i = 0; i < scaledHeight; i++ ) { memcpy( dxtPic + i * dxtWidth * 4, pic + i * scaledWidth * 4, scaledWidth * 4 ); } } else { dxtPic = pic; dxtWidth = scaledWidth; dxtHeight = scaledHeight; } } img.level = level; img.destZ = 0; img.width = scaledWidth; img.height = scaledHeight; // compress data or convert floats as necessary if( textureFormat == FMT_DXT1 ) { idDxtEncoder dxt; img.Alloc( dxtWidth * dxtHeight / 2 ); if( image_highQualityCompression.GetBool() ) { dxt.CompressImageDXT1HQ( dxtPic, img.data, dxtWidth, dxtHeight ); } else { dxt.CompressImageDXT1Fast( dxtPic, img.data, dxtWidth, dxtHeight ); } } else if( textureFormat == FMT_DXT5 ) { idDxtEncoder dxt; img.Alloc( dxtWidth * dxtHeight ); if( colorFormat == CFM_NORMAL_DXT5 ) { if( image_highQualityCompression.GetBool() ) { dxt.CompressNormalMapDXT5HQ( dxtPic, img.data, dxtWidth, dxtHeight ); } else { dxt.CompressNormalMapDXT5Fast( dxtPic, img.data, dxtWidth, dxtHeight ); } } else if( colorFormat == CFM_YCOCG_DXT5 ) { if( image_highQualityCompression.GetBool() ) { dxt.CompressYCoCgDXT5HQ( dxtPic, img.data, dxtWidth, dxtHeight ); } else { dxt.CompressYCoCgDXT5Fast( dxtPic, img.data, dxtWidth, dxtHeight ); } } else { fileData.colorFormat = colorFormat = CFM_DEFAULT; if( image_highQualityCompression.GetBool() ) { dxt.CompressImageDXT5HQ( dxtPic, img.data, dxtWidth, dxtHeight ); } else { dxt.CompressImageDXT5Fast( dxtPic, img.data, dxtWidth, dxtHeight ); } } } else if( textureFormat == FMT_LUM8 || textureFormat == FMT_INT8 ) { // LUM8 and INT8 just read the red channel img.Alloc( scaledWidth * scaledHeight ); for( int i = 0; i < img.dataSize; i++ ) { img.data[ i ] = pic[ i * 4 ]; } } else if( textureFormat == FMT_ALPHA ) { // ALPHA reads the alpha channel img.Alloc( scaledWidth * scaledHeight ); for( int i = 0; i < img.dataSize; i++ ) { img.data[ i ] = pic[ i * 4 + 3 ]; } } else if( textureFormat == FMT_L8A8 ) { // L8A8 reads the alpha and red channels img.Alloc( scaledWidth * scaledHeight * 2 ); for( int i = 0; i < img.dataSize / 2; i++ ) { img.data[ i * 2 + 0 ] = pic[ i * 4 + 0 ]; img.data[ i * 2 + 1 ] = pic[ i * 4 + 3 ]; } } else if( textureFormat == FMT_RGB565 ) { img.Alloc( scaledWidth * scaledHeight * 2 ); for( int i = 0; i < img.dataSize / 2; i++ ) { unsigned short color = ( ( pic[ i * 4 + 0 ] >> 3 ) << 11 ) | ( ( pic[ i * 4 + 1 ] >> 2 ) << 5 ) | ( pic[ i * 4 + 2 ] >> 3 ); img.data[ i * 2 + 0 ] = ( color >> 8 ) & 0xFF; img.data[ i * 2 + 1 ] = color & 0xFF; } } else { fileData.format = textureFormat = FMT_RGBA8; img.Alloc( scaledWidth * scaledHeight * 4 ); for( int i = 0; i < img.dataSize; i++ ) { img.data[ i ] = pic[ i ]; } } // if we had to pad to quads, free the padded version if( pic != dxtPic ) { Mem_Free( dxtPic ); dxtPic = NULL; } // downsample for the next level byte* shrunk = NULL; if( gammaMips ) { shrunk = R_MipMapWithGamma( pic, scaledWidth, scaledHeight ); } else { shrunk = R_MipMap( pic, scaledWidth, scaledHeight ); } Mem_Free( pic ); pic = shrunk; scaledWidth = Max( 1, scaledWidth >> 1 ); scaledHeight = Max( 1, scaledHeight >> 1 ); } Mem_Free( pic ); } /* ======================== PadImageTo4x4 DXT Compression requres a complete 4x4 block, even if the GPU will only be sampling a subset of it, so pad to 4x4 with replicated texels to maximize compression. ======================== */ static void PadImageTo4x4( const byte* src, int width, int height, byte dest[64] ) { // we probably will need to support this for non-square images, but I'll address // that when needed assert( width <= 4 && height <= 4 ); assert( width > 0 && height > 0 ); for( int y = 0 ; y < 4 ; y++ ) { int sy = y % height; for( int x = 0 ; x < 4 ; x++ ) { int sx = x % width; for( int c = 0 ; c < 4 ; c++ ) { dest[( y * 4 + x ) * 4 + c] = src[( sy * width + sx ) * 4 + c]; } } } } /* ======================== idBinaryImage::LoadCubeFromMemory ======================== */ void idBinaryImage::LoadCubeFromMemory( int width, const byte* pics[6], int numLevels, textureFormat_t& textureFormat, bool gammaMips ) { fileData.textureType = TT_CUBIC; fileData.format = textureFormat; fileData.colorFormat = CFM_DEFAULT; fileData.height = fileData.width = width; fileData.numLevels = numLevels; images.SetNum( fileData.numLevels * 6 ); for( int side = 0; side < 6; side++ ) { const byte* orig = pics[side]; const byte* pic = orig; int scaledWidth = fileData.width; for( int level = 0; level < fileData.numLevels; level++ ) { // compress data or convert floats as necessary idBinaryImageData& img = images[ level * 6 + side ]; // handle padding blocks less than 4x4 for the DXT compressors ALIGN16( byte padBlock[64] ); int padSize; const byte* padSrc; if( scaledWidth < 4 && ( textureFormat == FMT_DXT1 || textureFormat == FMT_DXT5 ) ) { PadImageTo4x4( pic, scaledWidth, scaledWidth, padBlock ); padSize = 4; padSrc = padBlock; } else { padSize = scaledWidth; padSrc = pic; } img.level = level; img.destZ = side; img.width = padSize; img.height = padSize; if( textureFormat == FMT_DXT1 ) { img.Alloc( padSize * padSize / 2 ); idDxtEncoder dxt; dxt.CompressImageDXT1Fast( padSrc, img.data, padSize, padSize ); } else if( textureFormat == FMT_DXT5 ) { img.Alloc( padSize * padSize ); idDxtEncoder dxt; dxt.CompressImageDXT5Fast( padSrc, img.data, padSize, padSize ); } else { fileData.format = textureFormat = FMT_RGBA8; img.Alloc( padSize * padSize * 4 ); memcpy( img.data, pic, img.dataSize ); } // downsample for the next level byte* shrunk = NULL; if( gammaMips ) { shrunk = R_MipMapWithGamma( pic, scaledWidth, scaledWidth ); } else { shrunk = R_MipMap( pic, scaledWidth, scaledWidth ); } if( pic != orig ) { Mem_Free( ( void* )pic ); pic = NULL; } pic = shrunk; scaledWidth = Max( 1, scaledWidth >> 1 ); } if( pic != orig ) { // free the down sampled version Mem_Free( ( void* )pic ); pic = NULL; } } } /* ======================== idBinaryImage::WriteGeneratedFile ======================== */ ID_TIME_T idBinaryImage::WriteGeneratedFile( ID_TIME_T sourceFileTime ) { idStr binaryFileName; MakeGeneratedFileName( binaryFileName ); idFileLocal file( fileSystem->OpenFileWrite( binaryFileName, "fs_basepath" ) ); if( file == NULL ) { idLib::Warning( "idBinaryImage: Could not open file '%s'", binaryFileName.c_str() ); return FILE_NOT_FOUND_TIMESTAMP; } idLib::Printf( "Writing %s\n", binaryFileName.c_str() ); fileData.headerMagic = BIMAGE_MAGIC; fileData.sourceFileTime = sourceFileTime; file->WriteBig( fileData.sourceFileTime ); file->WriteBig( fileData.headerMagic ); file->WriteBig( fileData.textureType ); file->WriteBig( fileData.format ); file->WriteBig( fileData.colorFormat ); file->WriteBig( fileData.width ); file->WriteBig( fileData.height ); file->WriteBig( fileData.numLevels ); for( int i = 0; i < images.Num(); i++ ) { idBinaryImageData& img = images[ i ]; file->WriteBig( img.level ); file->WriteBig( img.destZ ); file->WriteBig( img.width ); file->WriteBig( img.height ); file->WriteBig( img.dataSize ); file->Write( img.data, img.dataSize ); } return file->Timestamp(); } /* ========================== idBinaryImage::LoadFromGeneratedFile Load the preprocessed image from the generated folder. ========================== */ ID_TIME_T idBinaryImage::LoadFromGeneratedFile( ID_TIME_T sourceFileTime ) { idStr binaryFileName; MakeGeneratedFileName( binaryFileName ); idFileLocal bFile = fileSystem->OpenFileRead( binaryFileName ); if( bFile == NULL ) { return FILE_NOT_FOUND_TIMESTAMP; } if( LoadFromGeneratedFile( bFile, sourceFileTime ) ) { return bFile->Timestamp(); } return FILE_NOT_FOUND_TIMESTAMP; } /* ========================== idBinaryImage::LoadFromGeneratedFile Load the preprocessed image from the generated folder. ========================== */ bool idBinaryImage::LoadFromGeneratedFile( idFile* bFile, ID_TIME_T sourceFileTime ) { if( bFile->Read( &fileData, sizeof( fileData ) ) <= 0 ) { return false; } idSwapClass swap; swap.Big( fileData.sourceFileTime ); swap.Big( fileData.headerMagic ); swap.Big( fileData.textureType ); swap.Big( fileData.format ); swap.Big( fileData.colorFormat ); swap.Big( fileData.width ); swap.Big( fileData.height ); swap.Big( fileData.numLevels ); if( BIMAGE_MAGIC != fileData.headerMagic ) { return false; } if( fileData.sourceFileTime != sourceFileTime && sourceFileTime != 0 && com_productionMode.GetInteger() == 0 ) { return false; } int numImages = fileData.numLevels; if( fileData.textureType == TT_CUBIC ) { numImages *= 6; } images.SetNum( numImages ); for( int i = 0; i < numImages; i++ ) { idBinaryImageData& img = images[ i ]; if( bFile->Read( &img, sizeof( bimageImage_t ) ) <= 0 ) { return false; } idSwapClass swap; swap.Big( img.level ); swap.Big( img.destZ ); swap.Big( img.width ); swap.Big( img.height ); swap.Big( img.dataSize ); assert( img.level >= 0 && img.level < fileData.numLevels ); assert( img.destZ == 0 || fileData.textureType == TT_CUBIC ); assert( img.dataSize > 0 ); // DXT images need to be padded to 4x4 block sizes, but the original image // sizes are still retained, so the stored data size may be larger than // just the multiplication of dimensions assert( img.dataSize >= img.width * img.height * BitsForFormat( ( textureFormat_t )fileData.format ) / 8 ); img.Alloc( img.dataSize ); if( img.data == NULL ) { return false; } if( bFile->Read( img.data, img.dataSize ) <= 0 ) { return false; } } return true; } /* ========================== idBinaryImage::MakeGeneratedFileName ========================== */ void idBinaryImage::MakeGeneratedFileName( idStr& gfn ) { GetGeneratedFileName( gfn, GetName() ); } /* ========================== idBinaryImage::GetGeneratedFileName ========================== */ void idBinaryImage::GetGeneratedFileName( idStr& gfn, const char* name ) { gfn.Format( "generated/images/%s.bimage", name ); gfn.Replace( "(", "/" ); gfn.Replace( ",", "/" ); gfn.Replace( ")", "" ); gfn.Replace( " ", "" ); }