1 /* -----------------------------------------------------------------------------
2 
3 	Copyright (c) 2006 Simon Brown                          si@sjbrown.co.uk
4 
5 	Permission is hereby granted, free of charge, to any person obtaining
6 	a copy of this software and associated documentation files (the
7 	"Software"), to	deal in the Software without restriction, including
8 	without limitation the rights to use, copy, modify, merge, publish,
9 	distribute, sublicense, and/or sell copies of the Software, and to
10 	permit persons to whom the Software is furnished to do so, subject to
11 	the following conditions:
12 
13 	The above copyright notice and this permission notice shall be included
14 	in all copies or substantial portions of the Software.
15 
16 	THE SOFTWARE IS PROVIDED "AS IS", WITHOUT WARRANTY OF ANY KIND, EXPRESS
17 	OR IMPLIED, INCLUDING BUT NOT LIMITED TO THE WARRANTIES OF
18 	MERCHANTABILITY, FITNESS FOR A PARTICULAR PURPOSE AND NONINFRINGEMENT.
19 	IN NO EVENT SHALL THE AUTHORS OR COPYRIGHT HOLDERS BE LIABLE FOR ANY
20 	CLAIM, DAMAGES OR OTHER LIABILITY, WHETHER IN AN ACTION OF CONTRACT,
21 	TORT OR OTHERWISE, ARISING FROM, OUT OF OR IN CONNECTION WITH THE
22 	SOFTWARE OR THE USE OR OTHER DEALINGS IN THE SOFTWARE.
23 
24    -------------------------------------------------------------------------- */
25 
26 #include "squish.h"
27 #include "colourset.h"
28 #include "maths.h"
29 #include "rangefit.h"
30 #include "clusterfit.h"
31 #include "colourblock.h"
32 #include "alpha.h"
33 #include "singlecolourfit.h"
34 
35 namespace squish {
36 
FixFlags(int flags)37 static int FixFlags( int flags )
38 {
39 	// grab the flag bits
40 	int method = flags & ( kDxt1 | kDxt3 | kDxt5 | kBc4 | kBc5 );
41 	int fit = flags & ( kColourIterativeClusterFit | kColourClusterFit | kColourRangeFit );
42 	int extra = flags & kWeightColourByAlpha;
43 
44 	// set defaults
45 	if ( method != kDxt3
46 	&&   method != kDxt5
47 	&&   method != kBc4
48 	&&   method != kBc5 )
49 	{
50 		method = kDxt1;
51 	}
52 	if( fit != kColourRangeFit && fit != kColourIterativeClusterFit )
53 		fit = kColourClusterFit;
54 
55 	// done
56 	return method | fit | extra;
57 }
58 
CompressMasked(u8 const * rgba,int mask,void * block,int flags,float * metric)59 void CompressMasked( u8 const* rgba, int mask, void* block, int flags, float* metric )
60 {
61 	// fix any bad flags
62 	flags = FixFlags( flags );
63 
64 	if ( ( flags & ( kBc4 | kBc5 ) ) != 0 )
65 	{
66 		u8 alpha[16*4];
67 		for( int i = 0; i < 16; ++i )
68 		{
69 			alpha[i*4 + 3] = rgba[i*4 + 0]; // copy R to A
70 		}
71 
72 		u8* rBlock = reinterpret_cast< u8* >( block );
73 		CompressAlphaDxt5( alpha, mask, rBlock );
74 
75 		if ( ( flags & ( kBc5 ) ) != 0 )
76 		{
77 			for( int i = 0; i < 16; ++i )
78 			{
79 				alpha[i*4 + 3] = rgba[i*4 + 1]; // copy G to A
80 			}
81 
82 			u8* gBlock = reinterpret_cast< u8* >( block ) + 8;
83 			CompressAlphaDxt5( alpha, mask, gBlock );
84 		}
85 
86 		return;
87 	}
88 
89 	// get the block locations
90 	void* colourBlock = block;
91 	void* alphaBlock = block;
92 	if( ( flags & ( kDxt3 | kDxt5 ) ) != 0 )
93 		colourBlock = reinterpret_cast< u8* >( block ) + 8;
94 
95 	// create the minimal point set
96 	ColourSet colours( rgba, mask, flags );
97 
98 	// check the compression type and compress colour
99 	if( colours.GetCount() == 1 )
100 	{
101 		// always do a single colour fit
102 		SingleColourFit fit( &colours, flags );
103 		fit.Compress( colourBlock );
104 	}
105 	else if( ( flags & kColourRangeFit ) != 0 || colours.GetCount() == 0 )
106 	{
107 		// do a range fit
108 		RangeFit fit( &colours, flags, metric );
109 		fit.Compress( colourBlock );
110 	}
111 	else
112 	{
113 		// default to a cluster fit (could be iterative or not)
114 		ClusterFit fit( &colours, flags, metric );
115 		fit.Compress( colourBlock );
116 	}
117 
118 	// compress alpha separately if necessary
119 	if( ( flags & kDxt3 ) != 0 )
120 		CompressAlphaDxt3( rgba, mask, alphaBlock );
121 	else if( ( flags & kDxt5 ) != 0 )
122 		CompressAlphaDxt5( rgba, mask, alphaBlock );
123 }
124 
Decompress(u8 * rgba,void const * block,int flags)125 void Decompress( u8* rgba, void const* block, int flags )
126 {
127 	// fix any bad flags
128 	flags = FixFlags( flags );
129 
130 	// get the block locations
131 	void const* colourBlock = block;
132 	void const* alphaBock = block;
133 	if( ( flags & ( kDxt3 | kDxt5 ) ) != 0 )
134 		colourBlock = reinterpret_cast< u8 const* >( block ) + 8;
135 
136 	// decompress colour
137 	DecompressColour( rgba, colourBlock, ( flags & kDxt1 ) != 0 );
138 
139 	// decompress alpha separately if necessary
140 	if( ( flags & kDxt3 ) != 0 )
141 		DecompressAlphaDxt3( rgba, alphaBock );
142 	else if( ( flags & kDxt5 ) != 0 )
143 		DecompressAlphaDxt5( rgba, alphaBock );
144 }
145 
GetStorageRequirements(int width,int height,int flags)146 int GetStorageRequirements( int width, int height, int flags )
147 {
148 	// fix any bad flags
149 	flags = FixFlags( flags );
150 
151 	// compute the storage requirements
152 	int blockcount = ( ( width + 3 )/4 ) * ( ( height + 3 )/4 );
153 	int blocksize = ( ( flags & ( kDxt1 | kBc4 ) ) != 0 ) ? 8 : 16;
154 	return blockcount*blocksize;
155 }
156 
CompressImage(u8 const * rgba,int width,int height,void * blocks,int flags,float * metric)157 void CompressImage( u8 const* rgba, int width, int height, void* blocks, int flags, float* metric )
158 {
159 	// fix any bad flags
160 	flags = FixFlags( flags );
161 
162 	// initialise the block output
163 	u8* targetBlock = reinterpret_cast< u8* >( blocks );
164 	int bytesPerBlock = ( ( flags & ( kDxt1 | kBc4 ) ) != 0 ) ? 8 : 16;
165 
166 	// loop over blocks
167 	for( int y = 0; y < height; y += 4 )
168 	{
169 		for( int x = 0; x < width; x += 4 )
170 		{
171 			// build the 4x4 block of pixels
172 			u8 sourceRgba[16*4];
173 			u8* targetPixel = sourceRgba;
174 			int mask = 0;
175 			for( int py = 0; py < 4; ++py )
176 			{
177 				for( int px = 0; px < 4; ++px )
178 				{
179 					// get the source pixel in the image
180 					int sx = x + px;
181 					int sy = y + py;
182 
183 					// enable if we're in the image
184 					if( sx < width && sy < height )
185 					{
186 						// copy the rgba value
187 						u8 const* sourcePixel = rgba + 4*( width*sy + sx );
188 						for( int i = 0; i < 4; ++i )
189 							*targetPixel++ = *sourcePixel++;
190 
191 						// enable this pixel
192 						mask |= ( 1 << ( 4*py + px ) );
193 					}
194 					else
195 					{
196 						// skip this pixel as its outside the image
197 						targetPixel += 4;
198 					}
199 				}
200 			}
201 
202 			// compress it into the output
203 			CompressMasked( sourceRgba, mask, targetBlock, flags, metric );
204 
205 			// advance
206 			targetBlock += bytesPerBlock;
207 		}
208 	}
209 }
210 
DecompressImage(u8 * rgba,int width,int height,void const * blocks,int flags)211 void DecompressImage( u8* rgba, int width, int height, void const* blocks, int flags )
212 {
213 	// fix any bad flags
214 	flags = FixFlags( flags );
215 
216 	// initialise the block input
217 	u8 const* sourceBlock = reinterpret_cast< u8 const* >( blocks );
218 	int bytesPerBlock = ( ( flags & ( kDxt1 | kBc4 ) ) != 0 ) ? 8 : 16;
219 
220 	// loop over blocks
221 	for( int y = 0; y < height; y += 4 )
222 	{
223 		for( int x = 0; x < width; x += 4 )
224 		{
225 			// decompress the block
226 			u8 targetRgba[4*16];
227 			Decompress( targetRgba, sourceBlock, flags );
228 
229 			// write the decompressed pixels to the correct image locations
230 			u8 const* sourcePixel = targetRgba;
231 			for( int py = 0; py < 4; ++py )
232 			{
233 				for( int px = 0; px < 4; ++px )
234 				{
235 					// get the target location
236 					int sx = x + px;
237 					int sy = y + py;
238 					if( sx < width && sy < height )
239 					{
240 						u8* targetPixel = rgba + 4*( width*sy + sx );
241 
242 						// copy the rgba value
243 						for( int i = 0; i < 4; ++i )
244 							*targetPixel++ = *sourcePixel++;
245 					}
246 					else
247 					{
248 						// skip this pixel as its outside the image
249 						sourcePixel += 4;
250 					}
251 				}
252 			}
253 
254 			// advance
255 			sourceBlock += bytesPerBlock;
256 		}
257 	}
258 }
259 
260 } // namespace squish
261