| 1 | // tPixelUtil.cpp  |
| 2 | //  |
| 3 | // Helper functions for manipulating and parsing pixel-data in packed and compressed block formats.  |
| 4 | //  |
| 5 | // Copyright (c) 2022-2024 Tristan Grimmer.  |
| 6 | // Permission to use, copy, modify, and/or distribute this software for any purpose with or without fee is hereby  |
| 7 | // granted, provided that the above copyright notice and this permission notice appear in all copies.  |
| 8 | //  |
| 9 | // THE SOFTWARE IS PROVIDED "AS IS" AND THE AUTHOR DISCLAIMS ALL WARRANTIES WITH REGARD TO THIS SOFTWARE INCLUDING ALL  |
| 10 | // IMPLIED WARRANTIES OF MERCHANTABILITY AND FITNESS. IN NO EVENT SHALL THE AUTHOR BE LIABLE FOR ANY SPECIAL, DIRECT,  |
| 11 | // INDIRECT, OR CONSEQUENTIAL DAMAGES OR ANY DAMAGES WHATSOEVER RESULTING FROM LOSS OF USE, DATA OR PROFITS, WHETHER IN  |
| 12 | // AN ACTION OF CONTRACT, NEGLIGENCE OR OTHER TORTIOUS ACTION, ARISING OUT OF OR IN CONNECTION WITH THE USE OR  |
| 13 | // PERFORMANCE OF THIS SOFTWARE.  |
| 14 |   |
| 15 | #include <Foundation/tAssert.h>  |
| 16 | #include <Foundation/tStandard.h>  |
| 17 | #include <Foundation/tSmallFloat.h>  |
| 18 | #include <System/tMachine.h>  |
| 19 | #include "Image/tPixelUtil.h"  |
| 20 | #include "PVRTDecompress/PVRTDecompress.h"  |
| 21 | #define BCDEC_IMPLEMENTATION  |
| 22 | #include "bcdec/bcdec.h"  |
| 23 | #define ETCDEC_IMPLEMENTATION  |
| 24 | #include "etcdec/etcdec.h"  |
| 25 | #include "astcenc.h"  |
| 26 |   |
| 27 |   |
| 28 | namespace tImage  |
| 29 | {  |
| 30 | bool CanReverseRowData_Packed (tPixelFormat);  |
| 31 | bool CanReverseRowData_BC (tPixelFormat, int height);  |
| 32 |   |
| 33 | uint8* CreateReversedRowData_Packed (const uint8* pixelData, tPixelFormat pixelDataFormat, int width, int height);  |
| 34 | uint8* CreateReversedRowData_BC (const uint8* pixelData, tPixelFormat pixelDataFormat, int numBlocksW, int numBlocksH);  |
| 35 |   |
| 36 | // The BC2 block is the same for DXT2 and DXT3, although we don't support 2 (premultiplied alpha). Size is 128 bits.  |
| 37 | #pragma pack(push, 1)  |
| 38 | struct BC2Block  |
| 39 | {  |
| 40 | uint16 AlphaTableRows[4]; // Each alpha is 4 bits.  |
| 41 | BC1Block ColourBlock;  |
| 42 | };  |
| 43 |   |
| 44 | // The BC3 block is the same for DXT4 and 5, although we don't support 4 (premultiplied alpha). Size is 128 bits.  |
| 45 | struct BC3Block  |
| 46 | {  |
| 47 | uint8 Alpha0;  |
| 48 | uint8 Alpha1;  |
| 49 | uint8 AlphaTable[6]; // Each of the 4x4 pixel entries is 3 bits.  |
| 50 | BC1Block ColourBlock;  |
| 51 |   |
| 52 | // These accessors are needed because of the unusual alignment of the 3bit alpha indexes. They each return or set a  |
| 53 | // value in [0, 2^12) which represents a single row. The row variable should be in [0, 3]  |
| 54 | uint16 GetAlphaRow(int row);  |
| 55 | void SetAlphaRow(int row, uint16 val);  |
| 56 |   |
| 57 | };  |
| 58 | #pragma pack(pop)  |
| 59 | }  |
| 60 |   |
| 61 |   |
| 62 | tImage::DecodeResult tImage::DecodePixelData(tPixelFormat fmt, const uint8* src, int srcSize, int w, int h, tColour4b*& decoded4b, tColour4f*& decoded4f, tColourProfile profile, float RGBM_RGBD_MaxRange)  |
| 63 | {  |
| 64 | if (decoded4b || decoded4f)  |
| 65 | return DecodeResult::BuffersNotClear;  |
| 66 |   |
| 67 | if (!tIsPackedFormat(format: fmt) && !tIsBCFormat(format: fmt) && !tIsASTCFormat(format: fmt) && !tIsPVRFormat(format: fmt))  |
| 68 | return DecodeResult::UnsupportedFormat;  |
| 69 |   |
| 70 | if ((w <= 0) || (h <= 0) || !src)  |
| 71 | return DecodeResult::InvalidInput;  |
| 72 |   |
| 73 | if (tImage::tIsPackedFormat(format: fmt))  |
| 74 | {  |
| 75 | return DecodePixelData_Packed(fmt, data: src, dataSize: srcSize, w, h, decoded4b, decoded4f, RGBM_RGBD_MaxRange);  |
| 76 | }  |
| 77 | else if (tImage::tIsBCFormat(format: fmt))  |
| 78 | {  |
| 79 | return DecodePixelData_Block(fmt, data: src, dataSize: srcSize, w, h, decoded4b, decoded4f);  |
| 80 | }  |
| 81 | else if (tImage::tIsASTCFormat(format: fmt))  |
| 82 | {  |
| 83 | return DecodePixelData_ASTC(fmt, data: src, dataSize: srcSize, w, h, decoded4f, profile);  |
| 84 | }  |
| 85 | else if (tImage::tIsPVRFormat(format: fmt))  |
| 86 | {  |
| 87 | return DecodePixelData_PVR(fmt, data: src, dataSize: srcSize, w, h, decoded4b, decoded4f);  |
| 88 | }  |
| 89 | else // Unsupported PixelFormat  |
| 90 | {  |
| 91 | return DecodeResult::UnsupportedFormat;  |
| 92 | }  |
| 93 |   |
| 94 | return DecodeResult::Success;  |
| 95 | }  |
| 96 |   |
| 97 |   |
| 98 | tImage::DecodeResult tImage::DecodePixelData_Packed(tPixelFormat fmt, const uint8* src, int srcSize, int w, int h, tColour4b*& decoded4b, tColour4f*& decoded4f, float RGBM_RGBD_MaxRange)  |
| 99 | {  |
| 100 | if (decoded4b || decoded4f)  |
| 101 | return DecodeResult::BuffersNotClear;  |
| 102 |   |
| 103 | if (!tIsPackedFormat(format: fmt))  |
| 104 | return DecodeResult::UnsupportedFormat;  |
| 105 |   |
| 106 | if ((w <= 0) || (h <= 0) || !src)  |
| 107 | return DecodeResult::InvalidInput;  |
| 108 |   |
| 109 | switch (fmt)  |
| 110 | {  |
| 111 | case tPixelFormat::A8:  |
| 112 | // Convert to 32-bit RGBA with alpha in A and 0s for RGB.  |
| 113 | decoded4b = new tColour4b[w*h];  |
| 114 | for (int ij = 0; ij < w*h; ij++)  |
| 115 | {  |
| 116 | tColour4b col(0u, 0u, 0u, src[ij]);  |
| 117 | decoded4b[ij].Set(col);  |
| 118 | }  |
| 119 | break;  |
| 120 |   |
| 121 | case tPixelFormat::L8:  |
| 122 | case tPixelFormat::R8:  |
| 123 | {  |
| 124 | // Convert to 32-bit RGBA with red or luminance in R and 255 for A. If SpreadLuminance flag set,  |
| 125 | // also set luminance or red in the GB channels, if not then GB get 0s.  |
| 126 | decoded4b = new tColour4b[w*h];  |
| 127 | for (int ij = 0; ij < w*h; ij++)  |
| 128 | {  |
| 129 | tColour4b col(src[ij], 0u, 0u, 255u);  |
| 130 | decoded4b[ij].Set(col);  |
| 131 | }  |
| 132 | break;  |
| 133 | }  |
| 134 |   |
| 135 | case tPixelFormat::R8G8:  |
| 136 | decoded4b = new tColour4b[w*h];  |
| 137 | for (int ij = 0; ij < w*h; ij++)  |
| 138 | {  |
| 139 | tColour4b col(src[ij*2+0], src[ij*2+1], 0u, 255u);  |
| 140 | decoded4b[ij].Set(col);  |
| 141 | }  |
| 142 | break;  |
| 143 |   |
| 144 | case tPixelFormat::R8G8B8:  |
| 145 | decoded4b = new tColour4b[w*h];  |
| 146 | for (int ij = 0; ij < w*h; ij++)  |
| 147 | {  |
| 148 | tColour4b col(src[ij*3+0], src[ij*3+1], src[ij*3+2], 255u);  |
| 149 | decoded4b[ij].Set(col);  |
| 150 | }  |
| 151 | break;  |
| 152 |   |
| 153 | case tPixelFormat::R8G8B8A8:  |
| 154 | decoded4b = new tColour4b[w*h];  |
| 155 | for (int ij = 0; ij < w*h; ij++)  |
| 156 | {  |
| 157 | tColour4b col(src[ij*4+0], src[ij*4+1], src[ij*4+2], src[ij*4+3]);  |
| 158 | decoded4b[ij].Set(col);  |
| 159 | }  |
| 160 | break;  |
| 161 |   |
| 162 | case tPixelFormat::B8G8R8:  |
| 163 | decoded4b = new tColour4b[w*h];  |
| 164 | for (int ij = 0; ij < w*h; ij++)  |
| 165 | {  |
| 166 | tColour4b col(src[ij*3+2], src[ij*3+1], src[ij*3+0], 255u);  |
| 167 | decoded4b[ij].Set(col);  |
| 168 | }  |
| 169 | break;  |
| 170 |   |
| 171 | case tPixelFormat::B8G8R8A8:  |
| 172 | decoded4b = new tColour4b[w*h];  |
| 173 | for (int ij = 0; ij < w*h; ij++)  |
| 174 | {  |
| 175 | tColour4b col(src[ij*4+2], src[ij*4+1], src[ij*4+0], src[ij*4+3]);  |
| 176 | decoded4b[ij].Set(col);  |
| 177 | }  |
| 178 | break;  |
| 179 |   |
| 180 | case tPixelFormat::G3B5R5G3:  |
| 181 | decoded4b = new tColour4b[w*h];  |
| 182 | for (int ij = 0; ij < w*h; ij++)  |
| 183 | {  |
| 184 | // On an LE machine casting to a uint16 effectively swaps the bytes when doing bit ops.  |
| 185 | // This means red will be in the most significant bits -- that's why it looks backwards.  |
| 186 | // GGGBBBBB RRRRRGGG in memory is RRRRRGGG GGGBBBBB as a uint16.  |
| 187 | uint16 u = *((uint16*)(src+ij*2));  |
| 188 |   |
| 189 | uint8 r = (u ) >> 11; // 1111 1000 0000 0000 >> 11.  |
| 190 | uint8 g = (u & 0x07E0) >> 5; // 0000 0111 1110 0000 >> 5.  |
| 191 | uint8 b = (u & 0x001F) ; // 0000 0000 0001 1111 >> 0.  |
| 192 |   |
| 193 | // Normalize to range.  |
| 194 | // Careful here, you can't just do bit ops to get the components into position.  |
| 195 | // For example, a full red (11111) has to go to 255 (1.0f), and a zero red (00000) to 0(0.0f).  |
| 196 | // That is, the normalize has to divide by the range. At first I just masked and shifted the bits  |
| 197 | // to the right spot in an 8-bit type, but you don't know what to put in the LSBits. Putting 0s  |
| 198 | // would be bad (an 4 bit alpha of 1111 would go to 11110000... suddenly image not fully opaque)  |
| 199 | // and putting all 1s would add red (or alpha or whatever) when there was none. Half way won't  |
| 200 | // work either. You need the endpoints to work.  |
| 201 | float rf = (float(r) / 31.0f); // Max is 2^5 - 1.  |
| 202 | float gf = (float(g) / 63.0f); // Max is 2^6 - 1.  |
| 203 | float bf = (float(b) / 31.0f); // Max is 2^5 - 1.  |
| 204 | tColour4b col(rf, gf, bf, 1.0f);  |
| 205 | decoded4b[ij].Set(col);  |
| 206 | }  |
| 207 | break;  |
| 208 |   |
| 209 | case tPixelFormat::G4B4A4R4:  |
| 210 | decoded4b = new tColour4b[w*h];  |
| 211 | for (int ij = 0; ij < w*h; ij++)  |
| 212 | {  |
| 213 | // GGGGBBBB AAAARRRR in memory is AAAARRRR GGGGBBBB as a uint16.  |
| 214 | uint16 u = *((uint16*)(src+ij*2));  |
| 215 | uint8 a = (u ) >> 12; // 1111 0000 0000 0000 >> 12.  |
| 216 | uint8 r = (u & 0x0F00) >> 8; // 0000 1111 0000 0000 >> 8.  |
| 217 | uint8 g = (u & 0x00F0) >> 4; // 0000 0000 1111 0000 >> 4.  |
| 218 | uint8 b = (u & 0x000F) ; // 0000 0000 0000 1111 >> 0.  |
| 219 |   |
| 220 | // Normalize to range. See note above.  |
| 221 | float af = float(a) / 15.0f; // Max is 2^4 - 1.  |
| 222 | float rf = float(r) / 15.0f;  |
| 223 | float gf = float(g) / 15.0f;  |
| 224 | float bf = float(b) / 15.0f;  |
| 225 |   |
| 226 | tColour4b col(rf, gf, bf, af);  |
| 227 | decoded4b[ij].Set(col);  |
| 228 | }  |
| 229 | break;  |
| 230 |   |
| 231 | case tPixelFormat::B4A4R4G4:  |
| 232 | decoded4b = new tColour4b[w*h];  |
| 233 | for (int ij = 0; ij < w*h; ij++)  |
| 234 | {  |
| 235 | // BBBBAAAA RRRRGGGG in memory is RRRRGGGG BBBBAAAA as a uint16.  |
| 236 | uint16 u = *((uint16*)(src+ij*2));  |
| 237 | uint8 r = (u ) >> 12; // 1111 0000 0000 0000 >> 12.  |
| 238 | uint8 g = (u & 0x0F00) >> 8; // 0000 1111 0000 0000 >> 8.  |
| 239 | uint8 b = (u & 0x00F0) >> 4; // 0000 0000 1111 0000 >> 4.  |
| 240 | uint8 a = (u & 0x000F) ; // 0000 0000 0000 1111 >> 0.  |
| 241 |   |
| 242 | // Normalize to range. See note above.  |
| 243 | float af = float(a) / 15.0f; // Max is 2^4 - 1.  |
| 244 | float rf = float(r) / 15.0f;  |
| 245 | float gf = float(g) / 15.0f;  |
| 246 | float bf = float(b) / 15.0f;  |
| 247 |   |
| 248 | tColour4b col(rf, gf, bf, af);  |
| 249 | decoded4b[ij].Set(col);  |
| 250 | }  |
| 251 | break;  |
| 252 |   |
| 253 | case tPixelFormat::G3B5A1R5G2:  |
| 254 | decoded4b = new tColour4b[w*h];  |
| 255 | for (int ij = 0; ij < w*h; ij++)  |
| 256 | {  |
| 257 | // GGGBBBBB ARRRRRGG in memory is ARRRRRGG GGGBBBBB as a uint16.  |
| 258 | uint16 u = *((uint16*)(src+ij*2));  |
| 259 | bool a = (u & 0x8000); // 1000 0000 0000 0000.  |
| 260 | uint8 r = (u & 0x7C00) >> 10; // 0111 1100 0000 0000 >> 10.  |
| 261 | uint8 g = (u & 0x03E0) >> 5; // 0000 0011 1110 0000 >> 5.  |
| 262 | uint8 b = (u & 0x001F) ; // 0000 0000 0001 1111 >> 0.  |
| 263 |   |
| 264 | // Normalize to range. See note above.  |
| 265 | float rf = float(r) / 31.0f; // Max is 2^5 - 1.  |
| 266 | float gf = float(g) / 31.0f;  |
| 267 | float bf = float(b) / 31.0f;  |
| 268 |   |
| 269 | tColour4b col(rf, gf, bf, a ? 1.0f : 0.0f);  |
| 270 | decoded4b[ij].Set(col);  |
| 271 | }  |
| 272 | break;  |
| 273 |   |
| 274 | case tPixelFormat::G2B5A1R5G3:  |
| 275 | decoded4b = new tColour4b[w*h];  |
| 276 | for (int ij = 0; ij < w*h; ij++)  |
| 277 | {  |
| 278 | // GGBBBBBA RRRRRGGG in memory is RRRRRGGG GGBBBBBA as a uint16.  |
| 279 | uint16 u = *((uint16*)(src+ij*2));  |
| 280 | uint8 r = (u & 0xF800) >> 11; // 1111 1000 0000 0000 >> 11.  |
| 281 | uint8 g = (u & 0x07C0) >> 6; // 0000 0111 1100 0000 >> 6.  |
| 282 | uint8 b = (u & 0x003E) >> 1; // 0000 0000 0011 1110 >> 1.  |
| 283 | bool a = (u & 0x0001); // 0000 0000 0000 0001.  |
| 284 |   |
| 285 | // Normalize to range. See note above.  |
| 286 | float rf = float(r) / 31.0f; // Max is 2^5 - 1.  |
| 287 | float gf = float(g) / 31.0f;  |
| 288 | float bf = float(b) / 31.0f;  |
| 289 |   |
| 290 | tColour4b col(rf, gf, bf, a ? 1.0f : 0.0f);  |
| 291 | decoded4b[ij].Set(col);  |
| 292 | }  |
| 293 | break;  |
| 294 |   |
| 295 | case tPixelFormat::R16:  |
| 296 | {  |
| 297 | decoded4b = new tColour4b[w*h];  |
| 298 | uint16* udata = (uint16*)src;  |
| 299 | for (int ij = 0; ij < w*h; ij++)  |
| 300 | {  |
| 301 | uint8 r = udata[ij*1 + 0] >> 8;  |
| 302 | tColour4b col(r, 0u, 0u, 255u);  |
| 303 | decoded4b[ij].Set(col);  |
| 304 | }  |
| 305 | break;  |
| 306 | }  |
| 307 |   |
| 308 | case tPixelFormat::R16G16:  |
| 309 | {  |
| 310 | decoded4b = new tColour4b[w*h];  |
| 311 | uint16* udata = (uint16*)src;  |
| 312 | for (int ij = 0; ij < w*h; ij++)  |
| 313 | {  |
| 314 | uint8 r = udata[ij*2 + 0] >> 8;  |
| 315 | uint8 g = udata[ij*2 + 1] >> 8;  |
| 316 | tColour4b col(r, g, 0u, 255u);  |
| 317 | decoded4b[ij].Set(col);  |
| 318 | }  |
| 319 | break;  |
| 320 | }  |
| 321 |   |
| 322 | case tPixelFormat::R16G16B16:  |
| 323 | {  |
| 324 | decoded4b = new tColour4b[w*h];  |
| 325 | uint16* udata = (uint16*)src;  |
| 326 | for (int ij = 0; ij < w*h; ij++)  |
| 327 | {  |
| 328 | uint8 r = udata[ij*3 + 0] >> 8;  |
| 329 | uint8 g = udata[ij*3 + 1] >> 8;  |
| 330 | uint8 b = udata[ij*3 + 2] >> 8;  |
| 331 | tColour4b col(r, g, b, 255u);  |
| 332 | decoded4b[ij].Set(col);  |
| 333 | }  |
| 334 | break;  |
| 335 | }  |
| 336 |   |
| 337 | case tPixelFormat::R16G16B16A16:  |
| 338 | {  |
| 339 | decoded4b = new tColour4b[w*h];  |
| 340 | uint16* udata = (uint16*)src;  |
| 341 | for (int ij = 0; ij < w*h; ij++)  |
| 342 | {  |
| 343 | uint8 r = udata[ij*4 + 0] >> 8;  |
| 344 | uint8 g = udata[ij*4 + 1] >> 8;  |
| 345 | uint8 b = udata[ij*4 + 2] >> 8;  |
| 346 | uint8 a = udata[ij*4 + 3] >> 8;  |
| 347 | tColour4b col(r, g, b, a);  |
| 348 | decoded4b[ij].Set(col);  |
| 349 | }  |
| 350 | break;  |
| 351 | }  |
| 352 |   |
| 353 | case tPixelFormat::R32:  |
| 354 | {  |
| 355 | decoded4b = new tColour4b[w*h];  |
| 356 | uint32* udata = (uint32*)src;  |
| 357 | for (int ij = 0; ij < w*h; ij++)  |
| 358 | {  |
| 359 | uint8 r = udata[ij*1 + 0] >> 24;  |
| 360 | tColour4b col(r, 0u, 0u, 255u);  |
| 361 | decoded4b[ij].Set(col);  |
| 362 | }  |
| 363 | break;  |
| 364 | }  |
| 365 |   |
| 366 | case tPixelFormat::R32G32:  |
| 367 | {  |
| 368 | decoded4b = new tColour4b[w*h];  |
| 369 | uint32* udata = (uint32*)src;  |
| 370 | for (int ij = 0; ij < w*h; ij++)  |
| 371 | {  |
| 372 | uint8 r = udata[ij*2 + 0] >> 24;  |
| 373 | uint8 g = udata[ij*2 + 1] >> 24;  |
| 374 | tColour4b col(r, g, 0u, 255u);  |
| 375 | decoded4b[ij].Set(col);  |
| 376 | }  |
| 377 | break;  |
| 378 | }  |
| 379 |   |
| 380 | case tPixelFormat::R32G32B32:  |
| 381 | {  |
| 382 | decoded4b = new tColour4b[w*h];  |
| 383 | uint32* udata = (uint32*)src;  |
| 384 | for (int ij = 0; ij < w*h; ij++)  |
| 385 | {  |
| 386 | uint8 r = udata[ij*3 + 0] >> 24;  |
| 387 | uint8 g = udata[ij*3 + 1] >> 24;  |
| 388 | uint8 b = udata[ij*3 + 2] >> 24;  |
| 389 | tColour4b col(r, g, b, 255u);  |
| 390 | decoded4b[ij].Set(col);  |
| 391 | }  |
| 392 | break;  |
| 393 | }  |
| 394 |   |
| 395 | case tPixelFormat::R32G32B32A32:  |
| 396 | {  |
| 397 | decoded4b = new tColour4b[w*h];  |
| 398 | uint32* udata = (uint32*)src;  |
| 399 | for (int ij = 0; ij < w*h; ij++)  |
| 400 | {  |
| 401 | uint8 r = udata[ij*4 + 0] >> 24;  |
| 402 | uint8 g = udata[ij*4 + 1] >> 24;  |
| 403 | uint8 b = udata[ij*4 + 2] >> 24;  |
| 404 | uint8 a = udata[ij*4 + 3] >> 24;  |
| 405 | tColour4b col(r, g, b, a);  |
| 406 | decoded4b[ij].Set(col);  |
| 407 | }  |
| 408 | break;  |
| 409 | }  |
| 410 |   |
| 411 | case tPixelFormat::R16f:  |
| 412 | {  |
| 413 | // This HDR format has 1 red half-float channel.  |
| 414 | decoded4f = new tColour4f[w*h];  |
| 415 | tHalf* hdata = (tHalf*)src;  |
| 416 | for (int ij = 0; ij < w*h; ij++)  |
| 417 | {  |
| 418 | float r = hdata[ij*1 + 0];  |
| 419 | tColour4f col(r, 0.0f, 0.0f, 1.0f);  |
| 420 | decoded4f[ij].Set(col);  |
| 421 | }  |
| 422 | break;  |
| 423 | }  |
| 424 |   |
| 425 | case tPixelFormat::R16G16f:  |
| 426 | {  |
| 427 | // This HDR format has 2 half-float channels. Red and green.  |
| 428 | decoded4f = new tColour4f[w*h];  |
| 429 | tHalf* hdata = (tHalf*)src;  |
| 430 | for (int ij = 0; ij < w*h; ij++)  |
| 431 | {  |
| 432 | float r = hdata[ij*2 + 0];  |
| 433 | float g = hdata[ij*2 + 1];  |
| 434 | tColour4f col(r, g, 0.0f, 1.0f);  |
| 435 | decoded4f[ij].Set(col);  |
| 436 | }  |
| 437 | break;  |
| 438 | }  |
| 439 |   |
| 440 | case tPixelFormat::R16G16B16f:  |
| 441 | {  |
| 442 | // This HDR format has 3 half-float channels. RGB.  |
| 443 | decoded4f = new tColour4f[w*h];  |
| 444 | tHalf* hdata = (tHalf*)src;  |
| 445 | for (int ij = 0; ij < w*h; ij++)  |
| 446 | {  |
| 447 | float r = hdata[ij*3 + 0];  |
| 448 | float g = hdata[ij*3 + 1];  |
| 449 | float b = hdata[ij*3 + 2];  |
| 450 | tColour4f col(r, g, b, 1.0f);  |
| 451 | decoded4f[ij].Set(col);  |
| 452 | }  |
| 453 | break;  |
| 454 | }  |
| 455 |   |
| 456 | case tPixelFormat::R16G16B16A16f:  |
| 457 | {  |
| 458 | // This HDR format has 4 half-float channels. RGBA.  |
| 459 | decoded4f = new tColour4f[w*h];  |
| 460 | tHalf* hdata = (tHalf*)src;  |
| 461 | for (int ij = 0; ij < w*h; ij++)  |
| 462 | {  |
| 463 | float r = hdata[ij*4 + 0];  |
| 464 | float g = hdata[ij*4 + 1];  |
| 465 | float b = hdata[ij*4 + 2];  |
| 466 | float a = hdata[ij*4 + 3];  |
| 467 | tColour4f col(r, g, b, a);  |
| 468 | decoded4f[ij].Set(col);  |
| 469 | }  |
| 470 | break;  |
| 471 | }  |
| 472 |   |
| 473 | case tPixelFormat::R32f:  |
| 474 | {  |
| 475 | // This HDR format has 1 red float channel.  |
| 476 | decoded4f = new tColour4f[w*h];  |
| 477 | float* fdata = (float*)src;  |
| 478 | for (int ij = 0; ij < w*h; ij++)  |
| 479 | {  |
| 480 | float r = fdata[ij*1 + 0];  |
| 481 | tColour4f col(r, 0.0f, 0.0f, 1.0f);  |
| 482 | decoded4f[ij].Set(col);  |
| 483 | }  |
| 484 | break;  |
| 485 | }  |
| 486 |   |
| 487 | case tPixelFormat::R32G32f:  |
| 488 | {  |
| 489 | // This HDR format has 2 float channels. Red and green.  |
| 490 | decoded4f = new tColour4f[w*h];  |
| 491 | float* fdata = (float*)src;  |
| 492 | for (int ij = 0; ij < w*h; ij++)  |
| 493 | {  |
| 494 | float r = fdata[ij*2 + 0];  |
| 495 | float g = fdata[ij*2 + 1];  |
| 496 | tColour4f col(r, g, 0.0f, 1.0f);  |
| 497 | decoded4f[ij].Set(col);  |
| 498 | }  |
| 499 | break;  |
| 500 | }  |
| 501 |   |
| 502 | case tPixelFormat::R32G32B32f:  |
| 503 | {  |
| 504 | // This HDR format has 3 float channels. RGB.  |
| 505 | decoded4f = new tColour4f[w*h];  |
| 506 | float* fdata = (float*)src;  |
| 507 | for (int ij = 0; ij < w*h; ij++)  |
| 508 | {  |
| 509 | float r = fdata[ij*3 + 0];  |
| 510 | float g = fdata[ij*3 + 1];  |
| 511 | float b = fdata[ij*3 + 2];  |
| 512 | tColour4f col(r, g, b, 1.0f);  |
| 513 | decoded4f[ij].Set(col);  |
| 514 | }  |
| 515 | break;  |
| 516 | }  |
| 517 |   |
| 518 | case tPixelFormat::R32G32B32A32f:  |
| 519 | {  |
| 520 | // This HDR format has 4 RGBA floats.  |
| 521 | decoded4f = new tColour4f[w*h];  |
| 522 | float* fdata = (float*)src;  |
| 523 | for (int ij = 0; ij < w*h; ij++)  |
| 524 | {  |
| 525 | float r = fdata[ij*4 + 0];  |
| 526 | float g = fdata[ij*4 + 1];  |
| 527 | float b = fdata[ij*4 + 2];  |
| 528 | float a = fdata[ij*4 + 3];  |
| 529 | tColour4f col(r, g, b, a);  |
| 530 | decoded4f[ij].Set(col);  |
| 531 | }  |
| 532 | break;  |
| 533 | }  |
| 534 |   |
| 535 | case tPixelFormat::R11G11B10uf:  |
| 536 | {  |
| 537 | // This HDR format has 3 RGB floats packed into 32-bits.  |
| 538 | decoded4f = new tColour4f[w*h];  |
| 539 | uint32* fdata = (uint32*)src;  |
| 540 | for (int ij = 0; ij < w*h; ij++)  |
| 541 | {  |
| 542 | tPackedF11F11F10 packed(fdata[ij]);  |
| 543 | float r, g, b;  |
| 544 | packed.Get(x&: r, y&: g, z&: b);  |
| 545 | tColour4f col(r, g, b, 1.0f);  |
| 546 | decoded4f[ij].Set(col);  |
| 547 | }  |
| 548 | break;  |
| 549 | }  |
| 550 |   |
| 551 | case tPixelFormat::B10G11R11uf:  |
| 552 | {  |
| 553 | // This HDR format has 3 RGB floats packed into 32-bits.  |
| 554 | decoded4f = new tColour4f[w*h];  |
| 555 | uint32* fdata = (uint32*)src;  |
| 556 | for (int ij = 0; ij < w*h; ij++)  |
| 557 | {  |
| 558 | tPackedF10F11F11 packed(fdata[ij]);  |
| 559 | float r, g, b;  |
| 560 | packed.Get(x&: b, y&: g, z&: r);  |
| 561 | tColour4f col(r, g, b, 1.0f);  |
| 562 | decoded4f[ij].Set(col);  |
| 563 | }  |
| 564 | break;  |
| 565 | }  |
| 566 |   |
| 567 | case tPixelFormat::R9G9B9E5uf:  |
| 568 | {  |
| 569 | // This HDR format has 3 RGB floats packed into 32-bits.  |
| 570 | decoded4f = new tColour4f[w*h];  |
| 571 | uint32* fdata = (uint32*)src;  |
| 572 | for (int ij = 0; ij < w*h; ij++)  |
| 573 | {  |
| 574 | tPackedM9M9M9E5 packed(fdata[ij]);  |
| 575 | float r, g, b;  |
| 576 | packed.Get(x&: r, y&: g, z&: b);  |
| 577 | tColour4f col(r, g, b, 1.0f);  |
| 578 | decoded4f[ij].Set(col);  |
| 579 | }  |
| 580 | break;  |
| 581 | }  |
| 582 |   |
| 583 | case tPixelFormat::E5B9G9R9uf:  |
| 584 | {  |
| 585 | // This HDR format has 3 RGB floats packed into 32-bits.  |
| 586 | decoded4f = new tColour4f[w*h];  |
| 587 | uint32* fdata = (uint32*)src;  |
| 588 | for (int ij = 0; ij < w*h; ij++)  |
| 589 | {  |
| 590 | tPackedE5M9M9M9 packed(fdata[ij]);  |
| 591 | float r, g, b;  |
| 592 | packed.Get(x&: b, y&: g, z&: r);  |
| 593 | tColour4f col(r, g, b, 1.0f);  |
| 594 | decoded4f[ij].Set(col);  |
| 595 | }  |
| 596 | break;  |
| 597 | }  |
| 598 |   |
| 599 | case tPixelFormat::R8G8B8M8:  |
| 600 | {  |
| 601 | // This HDR format has 8-bit RGB components and a shared 8-bit multiplier.  |
| 602 | decoded4f = new tColour4f[w*h];  |
| 603 | uint8* udata = (uint8*)src;  |
| 604 | for (int ij = 0; ij < w*h; ij++)  |
| 605 | {  |
| 606 | float r = float(src[ij*4+0]) / 255.0f;  |
| 607 | float g = float(src[ij*4+1]) / 255.0f;  |
| 608 | float b = float(src[ij*4+2]) / 255.0f;  |
| 609 | float m = float(src[ij*4+3]) / 255.0f;  |
| 610 | r *= m * RGBM_RGBD_MaxRange;  |
| 611 | g *= m * RGBM_RGBD_MaxRange;  |
| 612 | b *= m * RGBM_RGBD_MaxRange;  |
| 613 | tColour4f col(r, g, b, 1.0f);  |
| 614 | decoded4f[ij].Set(col);  |
| 615 | }  |
| 616 | break;  |
| 617 | }  |
| 618 |   |
| 619 | case tPixelFormat::R8G8B8D8:  |
| 620 | {  |
| 621 | // This HDR format has 8-bit RGB components and a shared 8-bit divisor.  |
| 622 | decoded4f = new tColour4f[w*h];  |
| 623 | uint8* udata = (uint8*)src;  |
| 624 | for (int ij = 0; ij < w*h; ij++)  |
| 625 | {  |
| 626 | float r = float(src[ij*4+0]) / 255.0f;  |
| 627 | float g = float(src[ij*4+1]) / 255.0f;  |
| 628 | float b = float(src[ij*4+2]) / 255.0f;  |
| 629 | float d = float(src[ij*4+3]) / 255.0f;  |
| 630 | if (d == 0.0f)  |
| 631 | {  |
| 632 | r = 0.0f;  |
| 633 | g = 0.0f;  |
| 634 | b = 0.0f;  |
| 635 | d = 1.0f;  |
| 636 | }  |
| 637 | r *= (RGBM_RGBD_MaxRange/255.0f) / d;  |
| 638 | g *= (RGBM_RGBD_MaxRange/255.0f) / d;  |
| 639 | b *= (RGBM_RGBD_MaxRange/255.0f) / d;  |
| 640 | tColour4f col(r, g, b, 1.0f);  |
| 641 | decoded4f[ij].Set(col);  |
| 642 | }  |
| 643 | break;  |
| 644 | }  |
| 645 |   |
| 646 | default:  |
| 647 | return DecodeResult::PackedDecodeError;  |
| 648 | }  |
| 649 |   |
| 650 | return DecodeResult::Success;  |
| 651 | }  |
| 652 |   |
| 653 |   |
| 654 | tImage::DecodeResult tImage::DecodePixelData_Block(tPixelFormat fmt, const uint8* src, int srcSize, int w, int h, tColour4b*& decoded4b, tColour4f*& decoded4f)  |
| 655 | {  |
| 656 | if (decoded4b || decoded4f)  |
| 657 | return DecodeResult::BuffersNotClear;  |
| 658 |   |
| 659 | if (!tIsBCFormat(format: fmt))  |
| 660 | return DecodeResult::UnsupportedFormat;  |
| 661 |   |
| 662 | if ((w <= 0) || (h <= 0) || !src)  |
| 663 | return DecodeResult::InvalidInput;  |
| 664 |   |
| 665 | // We need extra room because the decompressor (bcdec) does not take an input for  |
| 666 | // the width and height, only the pitch (bytes per row). This means a texture that is 5  |
| 667 | // high will actually have row 6, 7, 8 written to.  |
| 668 | int wfull = 4 * tGetNumBlocks(blockWH: 4, imageWH: w);  |
| 669 | int hfull = 4 * tGetNumBlocks(blockWH: 4, imageWH: h);  |
| 670 | tColour4b* decodedFull4i = nullptr;  |
| 671 | tColour4f* decodedFull4f = nullptr;  |
| 672 | switch (fmt)  |
| 673 | {  |
| 674 | case tPixelFormat::BC1DXT1:  |
| 675 | case tPixelFormat::BC1DXT1A:  |
| 676 | {  |
| 677 | decodedFull4i = new tColour4b[wfull*hfull];  |
| 678 | for (int y = 0; y < hfull; y += 4)  |
| 679 | for (int x = 0; x < wfull; x += 4)  |
| 680 | {  |
| 681 | uint8* dst = (uint8*)decodedFull4i + (y*wfull + x) * 4;  |
| 682 |   |
| 683 | // At first didn't understand the pitch (3rd) argument. It's cuz the block needs to be written into  |
| 684 | // multiple rows of the destination and we need to know how far to increment to the next row of 4.  |
| 685 | bcdec_bc1(compressedBlock: src, decompressedBlock: dst, destinationPitch: wfull * 4);  |
| 686 | src += BCDEC_BC1_BLOCK_SIZE;  |
| 687 | }  |
| 688 |   |
| 689 | break;  |
| 690 | }  |
| 691 |   |
| 692 | case tPixelFormat::BC2DXT2DXT3:  |
| 693 | {  |
| 694 | decodedFull4i = new tColour4b[wfull*hfull];  |
| 695 | for (int y = 0; y < hfull; y += 4)  |
| 696 | for (int x = 0; x < wfull; x += 4)  |
| 697 | {  |
| 698 | uint8* dst = (uint8*)decodedFull4i + (y*wfull + x) * 4;  |
| 699 | bcdec_bc2(compressedBlock: src, decompressedBlock: dst, destinationPitch: wfull * 4);  |
| 700 | src += BCDEC_BC2_BLOCK_SIZE;  |
| 701 | }  |
| 702 | break;  |
| 703 | }  |
| 704 |   |
| 705 | case tPixelFormat::BC3DXT4DXT5:  |
| 706 | {  |
| 707 | decodedFull4i = new tColour4b[wfull*hfull];  |
| 708 | for (int y = 0; y < hfull; y += 4)  |
| 709 | for (int x = 0; x < wfull; x += 4)  |
| 710 | {  |
| 711 | uint8* dst = (uint8*)decodedFull4i + (y*wfull + x) * 4;  |
| 712 | bcdec_bc3(compressedBlock: src, decompressedBlock: dst, destinationPitch: wfull * 4);  |
| 713 | src += BCDEC_BC3_BLOCK_SIZE;  |
| 714 | }  |
| 715 | break;  |
| 716 | }  |
| 717 |   |
| 718 | case tPixelFormat::BC4ATI1U:  |
| 719 | {  |
| 720 | // This HDR format decompresses to R uint8s.  |
| 721 | uint8* rdata = new uint8[wfull*hfull];  |
| 722 | for (int y = 0; y < hfull; y += 4)  |
| 723 | for (int x = 0; x < wfull; x += 4)  |
| 724 | {  |
| 725 | uint8* dst = (rdata + (y*wfull + x) * 1);  |
| 726 | bcdec_bc4(compressedBlock: src, decompressedBlock: dst, destinationPitch: wfull * 1);  |
| 727 | src += BCDEC_BC4_BLOCK_SIZE;  |
| 728 | }  |
| 729 |   |
| 730 | // Now convert to 32-bit RGBA.  |
| 731 | decodedFull4i = new tColour4b[wfull*hfull];  |
| 732 | for (int xy = 0; xy < wfull*hfull; xy++)  |
| 733 | {  |
| 734 | uint8 v = rdata[xy];  |
| 735 | tColour4b col(v, 0u, 0u, 255u);  |
| 736 | decodedFull4i[xy].Set(col);  |
| 737 | }  |
| 738 | delete[] rdata;  |
| 739 | break;  |
| 740 | }  |
| 741 |   |
| 742 | case tPixelFormat::BC4ATI1S:  |
| 743 | {  |
| 744 | // This HDR format decompresses to R uint8s.  |
| 745 | uint8* rdata = new uint8[wfull*hfull];  |
| 746 | for (int y = 0; y < hfull; y += 4)  |
| 747 | for (int x = 0; x < wfull; x += 4)  |
| 748 | {  |
| 749 | uint8* dst = (rdata + (y*wfull + x) * 1);  |
| 750 | bcdec_bc4(compressedBlock: src, decompressedBlock: dst, destinationPitch: wfull * 1);  |
| 751 | src += BCDEC_BC4_BLOCK_SIZE;  |
| 752 | }  |
| 753 |   |
| 754 | // Now convert to 32-bit RGBA.  |
| 755 | decodedFull4i = new tColour4b[wfull*hfull];  |
| 756 | for (int xy = 0; xy < wfull*hfull; xy++)  |
| 757 | {  |
| 758 | int vi = *((int8*)&rdata[xy]);  |
| 759 | vi += 128;  |
| 760 | uint8 v = uint8(vi);  |
| 761 | tColour4b col(v, 0u, 0u, 255u);  |
| 762 | decodedFull4i[xy].Set(col);  |
| 763 | }  |
| 764 | delete[] rdata;  |
| 765 | break;  |
| 766 | }  |
| 767 |   |
| 768 | case tPixelFormat::BC5ATI2U:  |
| 769 | {  |
| 770 | // This HDR format decompresses to RG uint8s.  |
| 771 | struct RG { uint8 R; uint8 G; };  |
| 772 | RG* rgData = new RG[wfull*hfull];  |
| 773 |   |
| 774 | for (int y = 0; y < hfull; y += 4)  |
| 775 | for (int x = 0; x < wfull; x += 4)  |
| 776 | {  |
| 777 | uint8* dst = (uint8*)rgData + (y*wfull + x) * 2;  |
| 778 | bcdec_bc5(compressedBlock: src, decompressedBlock: dst, destinationPitch: wfull * 2);  |
| 779 | src += BCDEC_BC5_BLOCK_SIZE;  |
| 780 | }  |
| 781 |   |
| 782 | // Now convert to 32-bit RGBA with 0,255 for B,A.  |
| 783 | decodedFull4i = new tColour4b[wfull*hfull];  |
| 784 | for (int xy = 0; xy < wfull*hfull; xy++)  |
| 785 | {  |
| 786 | tColour4b col(rgData[xy].R, rgData[xy].G, 0u, 255u);  |
| 787 | decodedFull4i[xy].Set(col);  |
| 788 | }  |
| 789 | delete[] rgData;  |
| 790 | break;  |
| 791 | }  |
| 792 |   |
| 793 | case tPixelFormat::BC5ATI2S:  |
| 794 | {  |
| 795 | // This HDR format decompresses to RG uint8s.  |
| 796 | struct RG { uint8 R; uint8 G; };  |
| 797 | RG* rgData = new RG[wfull*hfull];  |
| 798 |   |
| 799 | for (int y = 0; y < hfull; y += 4)  |
| 800 | for (int x = 0; x < wfull; x += 4)  |
| 801 | {  |
| 802 | uint8* dst = (uint8*)rgData + (y*wfull + x) * 2;  |
| 803 | bcdec_bc5(compressedBlock: src, decompressedBlock: dst, destinationPitch: wfull * 2);  |
| 804 | src += BCDEC_BC5_BLOCK_SIZE;  |
| 805 | }  |
| 806 |   |
| 807 | // Now convert to 32-bit RGBA with 0,255 for B,A.  |
| 808 | decodedFull4i = new tColour4b[wfull*hfull];  |
| 809 | for (int xy = 0; xy < wfull*hfull; xy++)  |
| 810 | {  |
| 811 | int iR = *((int8*)&rgData[xy].R);  |
| 812 | int iG = *((int8*)&rgData[xy].G);  |
| 813 | iR += 128;  |
| 814 | iG += 128;  |
| 815 | tColour4b col(uint8(iR), uint8(iG), 0u, 255u);  |
| 816 | decodedFull4i[xy].Set(col);  |
| 817 | }  |
| 818 | delete[] rgData;  |
| 819 | break;  |
| 820 | }  |
| 821 |   |
| 822 | case tPixelFormat::BC6U:  |
| 823 | case tPixelFormat::BC6S:  |
| 824 | {  |
| 825 | // This HDR format decompresses to RGB floats.  |
| 826 | tColour3f* rgbData = new tColour3f[wfull*hfull];  |
| 827 |   |
| 828 | for (int y = 0; y < hfull; y += 4)  |
| 829 | for (int x = 0; x < wfull; x += 4)  |
| 830 | {  |
| 831 | uint8* dst = (uint8*)((float*)rgbData + (y*wfull + x) * 3);  |
| 832 | bool signedData = fmt == tPixelFormat::BC6S;  |
| 833 | bcdec_bc6h_float(compressedBlock: src, decompressedBlock: dst, destinationPitch: wfull * 3, isSigned: signedData);  |
| 834 | src += BCDEC_BC6H_BLOCK_SIZE;  |
| 835 | }  |
| 836 |   |
| 837 | // Now convert to 4-float (128-bit) RGBA with 1.0f alpha.  |
| 838 | decodedFull4f = new tColour4f[wfull*hfull];  |
| 839 | for (int xy = 0; xy < wfull*hfull; xy++)  |
| 840 | {  |
| 841 | tColour4f col(rgbData[xy], 1.0f);  |
| 842 | decodedFull4f[xy].Set(col);  |
| 843 | }  |
| 844 | delete[] rgbData;  |
| 845 | break;  |
| 846 | }  |
| 847 |   |
| 848 | case tPixelFormat::BC7:  |
| 849 | {  |
| 850 | decodedFull4i = new tColour4b[wfull*hfull];  |
| 851 | for (int y = 0; y < hfull; y += 4)  |
| 852 | for (int x = 0; x < wfull; x += 4)  |
| 853 | {  |
| 854 | uint8* dst = (uint8*)decodedFull4i + (y*wfull + x) * 4;  |
| 855 | bcdec_bc7(compressedBlock: src, decompressedBlock: dst, destinationPitch: wfull * 4);  |
| 856 | src += BCDEC_BC7_BLOCK_SIZE;  |
| 857 | }  |
| 858 | break;  |
| 859 | }  |
| 860 |   |
| 861 | case tPixelFormat::ETC1:  |
| 862 | case tPixelFormat::ETC2RGB: // Same decoder. Backwards compatible.  |
| 863 | {  |
| 864 | decodedFull4i = new tColour4b[wfull*hfull];  |
| 865 | for (int y = 0; y < hfull; y += 4)  |
| 866 | for (int x = 0; x < wfull; x += 4)  |
| 867 | {  |
| 868 | uint8* dst = (uint8*)decodedFull4i + (y*wfull + x) * 4;  |
| 869 | etcdec_etc_rgb(compressedBlock: src, decompressedBlock: dst, destinationPitch: wfull * 4);  |
| 870 | src += ETCDEC_ETC_RGB_BLOCK_SIZE;  |
| 871 | }  |
| 872 | break;  |
| 873 | }  |
| 874 |   |
| 875 | case tPixelFormat::ETC2RGBA:  |
| 876 | {  |
| 877 | decodedFull4i = new tColour4b[wfull*hfull];  |
| 878 | for (int y = 0; y < hfull; y += 4)  |
| 879 | for (int x = 0; x < wfull; x += 4)  |
| 880 | {  |
| 881 | uint8* dst = (uint8*)decodedFull4i + (y*wfull + x) * 4;  |
| 882 | etcdec_eac_rgba(compressedBlock: src, decompressedBlock: dst, destinationPitch: wfull * 4);  |
| 883 | src += ETCDEC_EAC_RGBA_BLOCK_SIZE;  |
| 884 | }  |
| 885 | break;  |
| 886 | }  |
| 887 |   |
| 888 | case tPixelFormat::ETC2RGBA1:  |
| 889 | {  |
| 890 | decodedFull4i = new tColour4b[wfull*hfull];  |
| 891 | for (int y = 0; y < hfull; y += 4)  |
| 892 | for (int x = 0; x < wfull; x += 4)  |
| 893 | {  |
| 894 | uint8* dst = (uint8*)decodedFull4i + (y*wfull + x) * 4;  |
| 895 | etcdec_etc_rgb_a1(compressedBlock: src, decompressedBlock: dst, destinationPitch: wfull * 4);  |
| 896 | src += ETCDEC_ETC_RGB_A1_BLOCK_SIZE;  |
| 897 | }  |
| 898 | break;  |
| 899 | }  |
| 900 |   |
| 901 | case tPixelFormat::EACR11U:  |
| 902 | {  |
| 903 | // This format decompresses to R uint16.  |
| 904 | uint16* rdata = new uint16[wfull*hfull];  |
| 905 |   |
| 906 | for (int y = 0; y < hfull; y += 4)  |
| 907 | for (int x = 0; x < wfull; x += 4)  |
| 908 | {  |
| 909 | uint16* dst = (rdata + (y*wfull + x) * 1);  |
| 910 | etcdec_eac_r11_u16(compressedBlock: src, decompressedBlock: dst, destinationPitch: wfull * sizeof(uint16));  |
| 911 | src += ETCDEC_EAC_R11_BLOCK_SIZE;  |
| 912 | }  |
| 913 |   |
| 914 | // Now convert to 32-bit RGBA.  |
| 915 | decodedFull4i = new tColour4b[wfull*hfull];  |
| 916 | for (int xy = 0; xy < wfull*hfull; xy++)  |
| 917 | {  |
| 918 | uint8 v = uint8( (255*rdata[xy]) / 65535 );  |
| 919 | tColour4b col(v, 0u, 0u, 255u);  |
| 920 | decodedFull4i[xy].Set(col);  |
| 921 | }  |
| 922 | delete[] rdata;  |
| 923 | break;  |
| 924 | }  |
| 925 |   |
| 926 | case tPixelFormat::EACR11S:  |
| 927 | {  |
| 928 | // This format decompresses to R float.  |
| 929 | float* rdata = new float[wfull*hfull];  |
| 930 |   |
| 931 | for (int y = 0; y < hfull; y += 4)  |
| 932 | for (int x = 0; x < wfull; x += 4)  |
| 933 | {  |
| 934 | float* dst = (rdata + (y*wfull + x) * 1);  |
| 935 | etcdec_eac_r11_float(compressedBlock: src, decompressedBlock: dst, destinationPitch: wfull * sizeof(float), isSigned: 1);  |
| 936 | src += ETCDEC_EAC_R11_BLOCK_SIZE;  |
| 937 | }  |
| 938 |   |
| 939 | // Now convert to 32-bit RGBA.  |
| 940 | decodedFull4i = new tColour4b[wfull*hfull];  |
| 941 | for (int xy = 0; xy < wfull*hfull; xy++)  |
| 942 | {  |
| 943 | float vf = tMath::tSaturate(val: (rdata[xy]+1.0f) / 2.0f);  |
| 944 | uint8 v = uint8( 255.0f * vf );  |
| 945 | tColour4b col(v, 0u, 0u, 255u);  |
| 946 | decodedFull4i[xy].Set(col);  |
| 947 | }  |
| 948 | delete[] rdata;  |
| 949 | break;  |
| 950 | }  |
| 951 |   |
| 952 | case tPixelFormat::EACRG11U:  |
| 953 | {  |
| 954 | struct RG { uint16 R; uint16 G; };  |
| 955 | // This format decompresses to RG uint16s.  |
| 956 | RG* rdata = new RG[wfull*hfull];  |
| 957 |   |
| 958 | for (int y = 0; y < hfull; y += 4)  |
| 959 | for (int x = 0; x < wfull; x += 4)  |
| 960 | {  |
| 961 | uint16* dst = (uint16*)rdata + (y*wfull + x) * 2;  |
| 962 | etcdec_eac_rg11_u16(compressedBlock: src, decompressedBlock: dst, destinationPitch: wfull * sizeof(RG));  |
| 963 | src += ETCDEC_EAC_RG11_BLOCK_SIZE;  |
| 964 | }  |
| 965 |   |
| 966 | // Now convert to 32-bit RGBA.  |
| 967 | decodedFull4i = new tColour4b[wfull*hfull];  |
| 968 | for (int xy = 0; xy < wfull*hfull; xy++)  |
| 969 | {  |
| 970 | uint8 r = uint8( (255*rdata[xy].R) / 65535 );  |
| 971 | uint8 g = uint8( (255*rdata[xy].G) / 65535 );  |
| 972 | tColour4b col(r, g, 0u, 255u);  |
| 973 | decodedFull4i[xy].Set(col);  |
| 974 | }  |
| 975 | delete[] rdata;  |
| 976 | break;  |
| 977 | }  |
| 978 |   |
| 979 | case tPixelFormat::EACRG11S:  |
| 980 | {  |
| 981 | struct RG { float R; float G; };  |
| 982 | // This format decompresses to RG floats.  |
| 983 | RG* rdata = new RG[wfull*hfull];  |
| 984 |   |
| 985 | for (int y = 0; y < hfull; y += 4)  |
| 986 | for (int x = 0; x < wfull; x += 4)  |
| 987 | {  |
| 988 | float* dst = (float*)rdata + (y*wfull + x) * 2;  |
| 989 | etcdec_eac_rg11_float(compressedBlock: src, decompressedBlock: dst, destinationPitch: wfull * sizeof(RG), isSigned: 1);  |
| 990 | src += ETCDEC_EAC_RG11_BLOCK_SIZE;  |
| 991 | }  |
| 992 |   |
| 993 | // Now convert to 32-bit RGBA.  |
| 994 | decodedFull4i = new tColour4b[wfull*hfull];  |
| 995 | for (int xy = 0; xy < wfull*hfull; xy++)  |
| 996 | {  |
| 997 | float rf = tMath::tSaturate(val: (rdata[xy].R+1.0f) / 2.0f);  |
| 998 | float gf = tMath::tSaturate(val: (rdata[xy].G+1.0f) / 2.0f);  |
| 999 | uint8 r = uint8( 255.0f * rf );  |
| 1000 | uint8 g = uint8( 255.0f * gf );  |
| 1001 | tColour4b col(r, g, 0u, 255u);  |
| 1002 | decodedFull4i[xy].Set(col);  |
| 1003 | }  |
| 1004 | delete[] rdata;  |
| 1005 | break;  |
| 1006 | }  |
| 1007 |   |
| 1008 | default:  |
| 1009 | return DecodeResult::BlockDecodeError;  |
| 1010 | }  |
| 1011 |   |
| 1012 | // Decode worked. We are now in RGBA 32-bit or float 128-bit. Width and height are already correct.  |
| 1013 | // This isn't the most efficient because we don't have a stride in a tLayer, but correctness first.  |
| 1014 | // Basically the decodedFull data may be too big if we needed extra room for w and h to do the decompression.  |
| 1015 | // This happens when the image dimensions where not multiples of the block size. We deal with that here.  |
| 1016 | // This is only inefficient if the dimensions were not a mult of 4, otherwise we can use the buffer directly.  |
| 1017 | tAssert((decodedFull4i || decodedFull4f) && !(decodedFull4i && decodedFull4f));  |
| 1018 |   |
| 1019 | // At this point the job is to get decoded4b or decoded4f to be valid. First check if sizes match exactly.  |
| 1020 | if ((wfull == w) && (hfull == h))  |
| 1021 | {  |
| 1022 | decoded4b = decodedFull4i;  |
| 1023 | decoded4f = decodedFull4f;  |
| 1024 | }  |
| 1025 | else  |
| 1026 | {  |
| 1027 | if (decodedFull4i)  |
| 1028 | {  |
| 1029 | decoded4b = new tColour4b[w*h];  |
| 1030 | tColour4b* src = decodedFull4i;  |
| 1031 | tColour4b* dst = decoded4b;  |
| 1032 | for (int r = 0; r < h; r++)  |
| 1033 | {  |
| 1034 | tStd::tMemcpy(dest: dst, src, numBytes: w*sizeof(tColour4b));  |
| 1035 | src += wfull;  |
| 1036 | dst += w;  |
| 1037 | }  |
| 1038 | delete[] decodedFull4i;  |
| 1039 | }  |
| 1040 | else if (decodedFull4f)  |
| 1041 | {  |
| 1042 | decoded4f = new tColour4f[w*h];  |
| 1043 | tColour4f* src = decodedFull4f;  |
| 1044 | tColour4f* dst = decoded4f;  |
| 1045 | for (int r = 0; r < h; r++)  |
| 1046 | {  |
| 1047 | tStd::tMemcpy(dest: dst, src, numBytes: w*sizeof(tColour4f));  |
| 1048 | src += wfull;  |
| 1049 | dst += w;  |
| 1050 | }  |
| 1051 | delete[] decodedFull4f;  |
| 1052 | }  |
| 1053 | }  |
| 1054 |   |
| 1055 | return DecodeResult::Success;  |
| 1056 | }  |
| 1057 |   |
| 1058 |   |
| 1059 | tImage::DecodeResult tImage::DecodePixelData_ASTC(tPixelFormat fmt, const uint8* src, int srcSize, int w, int h, tColour4f*& decoded4f, tColourProfile profile)  |
| 1060 | {  |
| 1061 | if (decoded4f)  |
| 1062 | return DecodeResult::BuffersNotClear;  |
| 1063 |   |
| 1064 | if (!tIsASTCFormat(format: fmt))  |
| 1065 | return DecodeResult::UnsupportedFormat;  |
| 1066 |   |
| 1067 | if ((w <= 0) || (h <= 0) || !src)  |
| 1068 | return DecodeResult::InvalidInput;  |
| 1069 |   |
| 1070 | int blockW = 0;  |
| 1071 | int blockH = 0;  |
| 1072 | int blockD = 1;  |
| 1073 |   |
| 1074 | // Convert source colour profile to astc colour profile.  |
| 1075 | astcenc_profile profileastc = ASTCENC_PRF_LDR_SRGB;  |
| 1076 | switch (profile)  |
| 1077 | {  |
| 1078 | case tColourProfile::Auto: profileastc = ASTCENC_PRF_HDR_RGB_LDR_A;break; // Works for LDR also.  |
| 1079 | case tColourProfile::LDRsRGB_LDRlA: profileastc = ASTCENC_PRF_LDR_SRGB; break;  |
| 1080 | case tColourProfile::LDRgRGB_LDRlA: profileastc = ASTCENC_PRF_LDR_SRGB; break; // Best approximation.  |
| 1081 | case tColourProfile::LDRlRGBA: profileastc = ASTCENC_PRF_LDR; break;  |
| 1082 | case tColourProfile::HDRlRGB_LDRlA: profileastc = ASTCENC_PRF_HDR_RGB_LDR_A;break;  |
| 1083 | case tColourProfile::HDRlRGBA: profileastc = ASTCENC_PRF_HDR; break;  |
| 1084 | }  |
| 1085 |   |
| 1086 | switch (fmt)  |
| 1087 | {  |
| 1088 | case tPixelFormat::ASTC4X4: blockW = 4; blockH = 4; break;  |
| 1089 | case tPixelFormat::ASTC5X4: blockW = 5; blockH = 4; break;  |
| 1090 | case tPixelFormat::ASTC5X5: blockW = 5; blockH = 5; break;  |
| 1091 | case tPixelFormat::ASTC6X5: blockW = 6; blockH = 5; break;  |
| 1092 | case tPixelFormat::ASTC6X6: blockW = 6; blockH = 6; break;  |
| 1093 | case tPixelFormat::ASTC8X5: blockW = 8; blockH = 5; break;  |
| 1094 | case tPixelFormat::ASTC8X6: blockW = 8; blockH = 6; break;  |
| 1095 | case tPixelFormat::ASTC8X8: blockW = 8; blockH = 8; break;  |
| 1096 | case tPixelFormat::ASTC10X5: blockW = 10; blockH = 5; break;  |
| 1097 | case tPixelFormat::ASTC10X6: blockW = 10; blockH = 6; break;  |
| 1098 | case tPixelFormat::ASTC10X8: blockW = 10; blockH = 8; break;  |
| 1099 | case tPixelFormat::ASTC10X10: blockW = 10; blockH = 10; break;  |
| 1100 | case tPixelFormat::ASTC12X10: blockW = 12; blockH = 10; break;  |
| 1101 | case tPixelFormat::ASTC12X12: blockW = 12; blockH = 12; break;  |
| 1102 | default: break;  |
| 1103 | }  |
| 1104 |   |
| 1105 | if (!blockW || !blockH)  |
| 1106 | return DecodeResult::ASTCDecodeError;  |
| 1107 |   |
| 1108 | float quality = ASTCENC_PRE_MEDIUM; // Only need for compression.  |
| 1109 | astcenc_error result = ASTCENC_SUCCESS;  |
| 1110 | astcenc_config config;  |
| 1111 | astcenc_config_init(profile: profileastc, block_x: blockW, block_y: blockH, block_z: blockD, quality, flags: ASTCENC_FLG_DECOMPRESS_ONLY, config: &config);  |
| 1112 |   |
| 1113 | // astcenc_get_error_string(status) can be called for details.  |
| 1114 | if (result != ASTCENC_SUCCESS)  |
| 1115 | return DecodeResult::ASTCDecodeError;  |
| 1116 |   |
| 1117 | astcenc_context* context = nullptr;  |
| 1118 | int numThreads = tMath::tMax(a: tSystem::tGetNumCores(), b: 2);  |
| 1119 | result = astcenc_context_alloc(config: &config, thread_count: numThreads, context: &context);  |
| 1120 | if (result != ASTCENC_SUCCESS)  |
| 1121 | return DecodeResult::ASTCDecodeError;  |
| 1122 |   |
| 1123 | decoded4f = new tColour4f[w*h];  |
| 1124 | astcenc_image image;  |
| 1125 | image.dim_x = w;  |
| 1126 | image.dim_y = h;  |
| 1127 | image.dim_z = 1;  |
| 1128 | image.data_type = ASTCENC_TYPE_F32;  |
| 1129 |   |
| 1130 | tColour4f* slices = decoded4f;  |
| 1131 | image.data = reinterpret_cast<void**>(&slices);  |
| 1132 | astcenc_swizzle swizzle { .r: ASTCENC_SWZ_R, .g: ASTCENC_SWZ_G, .b: ASTCENC_SWZ_B, .a: ASTCENC_SWZ_A };  |
| 1133 |   |
| 1134 | result = astcenc_decompress_image(context, data: src, data_len: srcSize, image_out: &image, swizzle: &swizzle, thread_index: 0);  |
| 1135 | if (result != ASTCENC_SUCCESS)  |
| 1136 | {  |
| 1137 | astcenc_context_free(context);  |
| 1138 | delete[] decoded4f;  |
| 1139 | decoded4f = nullptr;  |
| 1140 | return DecodeResult::ASTCDecodeError;  |
| 1141 | }  |
| 1142 | astcenc_context_free(context);  |
| 1143 |   |
| 1144 | return DecodeResult::Success;  |
| 1145 | }  |
| 1146 |   |
| 1147 |   |
| 1148 | tImage::DecodeResult tImage::DecodePixelData_PVR(tPixelFormat fmt, const uint8* src, int srcSize, int w, int h, tColour4b*& decoded4b, tColour4f*& decoded4f)  |
| 1149 | {  |
| 1150 | if (decoded4b || decoded4f)  |
| 1151 | return DecodeResult::BuffersNotClear;  |
| 1152 |   |
| 1153 | if (!tIsPVRFormat(format: fmt))  |
| 1154 | return DecodeResult::UnsupportedFormat;  |
| 1155 |   |
| 1156 | if ((w <= 0) || (h <= 0) || !src)  |
| 1157 | return DecodeResult::InvalidInput;  |
| 1158 |   |
| 1159 | // The PVRTDecompress calls expect the decoded destination array to be bug enough to handle w*h tColour4b pixels.  |
| 1160 | // The function handles cases where the min width and height are too small, so even a 1x1 image can be handed off.  |
| 1161 | switch (fmt)  |
| 1162 | {  |
| 1163 | case tPixelFormat::PVRBPP4:  |
| 1164 | {  |
| 1165 | decoded4b = new tColour4b[w*h];  |
| 1166 | uint32_t do2bitMode = 0;  |
| 1167 | uint32_t numSrcBytesDecompressed = pvr::PVRTDecompressPVRTC(compressedData: src, do2bitMode, xDim: w, yDim: h, outResultImage: (uint8_t*)decoded4b);  |
| 1168 | if (numSrcBytesDecompressed == 0)  |
| 1169 | {  |
| 1170 | delete[] decoded4b;  |
| 1171 | decoded4b = nullptr;  |
| 1172 | return DecodeResult::PVRDecodeError;  |
| 1173 | }  |
| 1174 | break;  |
| 1175 | }  |
| 1176 |   |
| 1177 | case tPixelFormat::PVRBPP2:  |
| 1178 | {  |
| 1179 | decoded4b = new tColour4b[w*h];  |
| 1180 | uint32_t do2bitMode = 1;  |
| 1181 | uint32_t numSrcBytesDecompressed = pvr::PVRTDecompressPVRTC(compressedData: src, do2bitMode, xDim: w, yDim: h, outResultImage: (uint8_t*)decoded4b);  |
| 1182 | if (numSrcBytesDecompressed == 0)  |
| 1183 | {  |
| 1184 | delete[] decoded4b;  |
| 1185 | decoded4b = nullptr;  |
| 1186 | return DecodeResult::PVRDecodeError;  |
| 1187 | }  |
| 1188 | break;  |
| 1189 | }  |
| 1190 |   |
| 1191 | #ifdef PIXEL_FORMAT_INCLUDE_NOT_IMPLEMENTED  |
| 1192 | case tPixelFormat::PVR2BPP4:  |
| 1193 | case tPixelFormat::PVR2BPP2:  |
| 1194 | return DecodeResult::UnsupportedFormat;  |
| 1195 |   |
| 1196 | case tPixelFormat::PVRHDRBPP8:  |
| 1197 | case tPixelFormat::PVRHDRBPP6:  |
| 1198 | return DecodeResult::UnsupportedFormat;  |
| 1199 |   |
| 1200 | case tPixelFormat::PVR2HDRBPP8:  |
| 1201 | case tPixelFormat::PVR2HDRBPP6:  |
| 1202 | return DecodeResult::UnsupportedFormat;  |
| 1203 | #endif  |
| 1204 |   |
| 1205 | default:  |
| 1206 | return DecodeResult::PVRDecodeError;  |
| 1207 | }  |
| 1208 |   |
| 1209 | return DecodeResult::Success;  |
| 1210 | }  |
| 1211 |   |
| 1212 |   |
| 1213 | uint16 tImage::BC3Block::GetAlphaRow(int row)  |
| 1214 | {  |
| 1215 | tAssert(row < 4);  |
| 1216 | switch (row)  |
| 1217 | {  |
| 1218 | case 1:  |
| 1219 | return (AlphaTable[2] << 4) | (0x0F & (AlphaTable[1] >> 4));  |
| 1220 |   |
| 1221 | case 0:  |
| 1222 | return ((AlphaTable[1] & 0x0F) << 8) | AlphaTable[0];  |
| 1223 |   |
| 1224 | case 3:  |
| 1225 | return (AlphaTable[5] << 4) | (0x0F & (AlphaTable[4] >> 4));  |
| 1226 |   |
| 1227 | case 2:  |
| 1228 | return ((AlphaTable[4] & 0x0F) << 8) | AlphaTable[3];  |
| 1229 | }  |
| 1230 | return 0;  |
| 1231 | }  |
| 1232 |   |
| 1233 |   |
| 1234 | void tImage::BC3Block::SetAlphaRow(int row, uint16 val)  |
| 1235 | {  |
| 1236 | tAssert(row < 4);  |
| 1237 | tAssert(val < 4096);  |
| 1238 | switch (row)  |
| 1239 | {  |
| 1240 | case 1:  |
| 1241 | AlphaTable[2] = val >> 4;  |
| 1242 | AlphaTable[1] = (AlphaTable[1] & 0x0F) | ((val & 0x000F) << 4);  |
| 1243 | break;  |
| 1244 |   |
| 1245 | case 0:  |
| 1246 | AlphaTable[1] = (AlphaTable[1] & 0xF0) | (val >> 8);  |
| 1247 | AlphaTable[0] = val & 0x00FF;  |
| 1248 | break;  |
| 1249 |   |
| 1250 | case 3:  |
| 1251 | AlphaTable[5] = val >> 4;  |
| 1252 | AlphaTable[4] = (AlphaTable[4] & 0x0F) | ((val & 0x000F) << 4);  |
| 1253 | break;  |
| 1254 |   |
| 1255 | case 2:  |
| 1256 | AlphaTable[4] = (AlphaTable[4] & 0xF0) | (val >> 8);  |
| 1257 | AlphaTable[3] = val & 0x00FF;  |
| 1258 | break;  |
| 1259 | }  |
| 1260 | }  |
| 1261 |   |
| 1262 |   |
| 1263 | bool tImage::DoBC1BlocksHaveBinaryAlpha(tImage::BC1Block* block, int numBlocks)  |
| 1264 | {  |
| 1265 | // The only way to check if the DXT1 format has alpha is by checking each block individually. If the block uses  |
| 1266 | // alpha, the min and max colours are ordered in a particular order.  |
| 1267 | for (int b = 0; b < numBlocks; b++)  |
| 1268 | {  |
| 1269 | if (block->Colour0 <= block->Colour1)  |
| 1270 | {  |
| 1271 | // It seems that at least the nVidia DXT compressor can generate an opaque DXT1 block with the colours in the order for a transparent one.  |
| 1272 | // This forces us to check all the indexes to see if the alpha index (11 in binary) is used -- if not then it's still an opaque block.  |
| 1273 | for (int row = 0; row < 4; row++)  |
| 1274 | {  |
| 1275 | uint8 bits = block->LookupTableRows[row];  |
| 1276 | if  |
| 1277 | (  |
| 1278 | ((bits & 0x03) == 0x03) ||  |
| 1279 | ((bits & 0x0C) == 0x0C) ||  |
| 1280 | ((bits & 0x30) == 0x30) ||  |
| 1281 | ((bits & 0xC0) == 0xC0)  |
| 1282 | )  |
| 1283 | {  |
| 1284 | return true;  |
| 1285 | }  |
| 1286 | }  |
| 1287 | }  |
| 1288 |   |
| 1289 | block++;  |
| 1290 | }  |
| 1291 |   |
| 1292 | return false;  |
| 1293 | }  |
| 1294 |   |
| 1295 |   |
| 1296 | bool tImage::CanReverseRowData(tPixelFormat format, int height)  |
| 1297 | {  |
| 1298 | if (tIsPackedFormat(format))  |
| 1299 | return CanReverseRowData_Packed(format);  |
| 1300 |   |
| 1301 | if (tIsBCFormat(format))  |
| 1302 | return CanReverseRowData_BC(format, height);  |
| 1303 |   |
| 1304 | return false;  |
| 1305 | }  |
| 1306 |   |
| 1307 |   |
| 1308 | uint8* tImage::CreateReversedRowData(const uint8* pixelData, tPixelFormat pixelDataFormat, int numBlocksW, int numBlocksH)  |
| 1309 | {  |
| 1310 | if (tIsPackedFormat(format: pixelDataFormat))  |
| 1311 | return CreateReversedRowData_Packed(pixelData, pixelDataFormat, width: numBlocksW, height: numBlocksH);  |
| 1312 |   |
| 1313 | if (tIsBCFormat(format: pixelDataFormat))  |
| 1314 | return CreateReversedRowData_BC(pixelData, pixelDataFormat, numBlocksW, numBlocksH);  |
| 1315 |   |
| 1316 | return nullptr;  |
| 1317 | }  |
| 1318 |   |
| 1319 |   |
| 1320 | bool tImage::CanReverseRowData_Packed(tPixelFormat format)  |
| 1321 | {  |
| 1322 | int bitsPerPixel = tImage::tGetBitsPerPixel(format);  |
| 1323 | if ((bitsPerPixel % 8) == 0)  |
| 1324 | return true;  |
| 1325 |   |
| 1326 | return false;  |
| 1327 | }  |
| 1328 |   |
| 1329 |   |
| 1330 | uint8* tImage::CreateReversedRowData_Packed(const uint8* pixelData, tPixelFormat pixelDataFormat, int width, int height)  |
| 1331 | {  |
| 1332 | // We only support pixel formats that contain a whole number of bytes per pixel.  |
| 1333 | // That will cover all reasonable RGB and RGBA formats, but not ASTC formats.  |
| 1334 | if (!CanReverseRowData_Packed(format: pixelDataFormat))  |
| 1335 | return nullptr;  |
| 1336 |   |
| 1337 | int bitsPerPixel = tImage::tGetBitsPerPixel(pixelDataFormat);  |
| 1338 | int bytesPerPixel = bitsPerPixel/8;  |
| 1339 | int numBytes = width*height*bytesPerPixel;  |
| 1340 |   |
| 1341 | uint8* reversedPixelData = new uint8[numBytes];  |
| 1342 | uint8* dstData = reversedPixelData;  |
| 1343 | for (int row = height-1; row >= 0; row--)  |
| 1344 | {  |
| 1345 | for (int col = 0; col < width; col++)  |
| 1346 | {  |
| 1347 | const uint8* srcData = pixelData + row*bytesPerPixel*width + col*bytesPerPixel;  |
| 1348 | for (int byte = 0; byte < bytesPerPixel; byte++, dstData++, srcData++)  |
| 1349 | *dstData = *srcData;  |
| 1350 | }  |
| 1351 | }  |
| 1352 | return reversedPixelData;  |
| 1353 | }  |
| 1354 |   |
| 1355 |   |
| 1356 | bool tImage::CanReverseRowData_BC(tPixelFormat format, int height)  |
| 1357 | {  |
| 1358 | switch (format)  |
| 1359 | {  |
| 1360 | case tPixelFormat::BC1DXT1A:  |
| 1361 | case tPixelFormat::BC1DXT1:  |
| 1362 | case tPixelFormat::BC2DXT2DXT3:  |
| 1363 | case tPixelFormat::BC3DXT4DXT5:  |
| 1364 | if ((height % tGetBlockHeight(format)) == 0)  |
| 1365 | return true;  |
| 1366 | break;  |
| 1367 | }  |
| 1368 |   |
| 1369 | return false;  |
| 1370 | }  |
| 1371 |   |
| 1372 |   |
| 1373 | uint8* tImage::CreateReversedRowData_BC(const uint8* pixelData, tPixelFormat pixelDataFormat, int numBlocksW, int numBlocksH)  |
| 1374 | {  |
| 1375 | // We do not support all BC formats for this..  |
| 1376 | if (!CanReverseRowData_BC(format: pixelDataFormat, height: numBlocksH*tGetBlockHeight(pixelDataFormat)))  |
| 1377 | return nullptr;  |
| 1378 |   |
| 1379 | int bcBlockSize = tImage::tGetBytesPerBlock(pixelDataFormat);  |
| 1380 | int numBlocks = numBlocksW*numBlocksH;  |
| 1381 | int numBytes = numBlocks * bcBlockSize;  |
| 1382 |   |
| 1383 | uint8* reversedPixelData = new uint8[numBytes];  |
| 1384 | uint8* dstData = reversedPixelData;  |
| 1385 | for (int row = numBlocksH-1; row >= 0; row--)  |
| 1386 | {  |
| 1387 | for (int col = 0; col < numBlocksW; col++)  |
| 1388 | {  |
| 1389 | const uint8* srcData = pixelData + row*bcBlockSize*numBlocksW + col*bcBlockSize;  |
| 1390 | for (int byte = 0; byte < bcBlockSize; byte++, dstData++, srcData++)  |
| 1391 | *dstData = *srcData;  |
| 1392 | }  |
| 1393 | }  |
| 1394 |   |
| 1395 | // Now we flip the inter-block rows by messing with the block's lookup-table. We handle three types of  |
| 1396 | // blocks: BC1, BC2, and BC3. BC4/5 probably could be handled, and BC6/7 are too complex.  |
| 1397 | switch (pixelDataFormat)  |
| 1398 | {  |
| 1399 | case tPixelFormat::BC1DXT1A:  |
| 1400 | case tPixelFormat::BC1DXT1:  |
| 1401 | {  |
| 1402 | tImage::BC1Block* block = (tImage::BC1Block*)reversedPixelData;  |
| 1403 | for (int b = 0; b < numBlocks; b++, block++)  |
| 1404 | {  |
| 1405 | // Reorder each row's colour indexes.  |
| 1406 | tStd::tSwap(a&: block->LookupTableRows[0], b&: block->LookupTableRows[3]);  |
| 1407 | tStd::tSwap(a&: block->LookupTableRows[1], b&: block->LookupTableRows[2]);  |
| 1408 | }  |
| 1409 | break;  |
| 1410 | }  |
| 1411 |   |
| 1412 | case tPixelFormat::BC2DXT2DXT3:  |
| 1413 | {  |
| 1414 | tImage::BC2Block* block = (tImage::BC2Block*)reversedPixelData;  |
| 1415 | for (int b = 0; b < numBlocks; b++, block++)  |
| 1416 | {  |
| 1417 | // Reorder the explicit alphas AND the colour indexes.  |
| 1418 | tStd::tSwap(a&: block->AlphaTableRows[0], b&: block->AlphaTableRows[3]);  |
| 1419 | tStd::tSwap(a&: block->AlphaTableRows[1], b&: block->AlphaTableRows[2]);  |
| 1420 | tStd::tSwap(a&: block->ColourBlock.LookupTableRows[0], b&: block->ColourBlock.LookupTableRows[3]);  |
| 1421 | tStd::tSwap(a&: block->ColourBlock.LookupTableRows[1], b&: block->ColourBlock.LookupTableRows[2]);  |
| 1422 | }  |
| 1423 | break;  |
| 1424 | }  |
| 1425 |   |
| 1426 | case tPixelFormat::BC3DXT4DXT5:  |
| 1427 | {  |
| 1428 | tImage::BC3Block* block = (tImage::BC3Block*)reversedPixelData;  |
| 1429 | for (int b = 0; b < numBlocks; b++, block++)  |
| 1430 | {  |
| 1431 | // Reorder the alpha indexes AND the colour indexes.  |
| 1432 | uint16 orig0 = block->GetAlphaRow(row: 0);  |
| 1433 | block->SetAlphaRow(row: 0, val: block->GetAlphaRow(row: 3));  |
| 1434 | block->SetAlphaRow(row: 3, val: orig0);  |
| 1435 |   |
| 1436 | uint16 orig1 = block->GetAlphaRow(row: 1);  |
| 1437 | block->SetAlphaRow(row: 1, val: block->GetAlphaRow(row: 2));  |
| 1438 | block->SetAlphaRow(row: 2, val: orig1);  |
| 1439 |   |
| 1440 | tStd::tSwap(a&: block->ColourBlock.LookupTableRows[0], b&: block->ColourBlock.LookupTableRows[3]);  |
| 1441 | tStd::tSwap(a&: block->ColourBlock.LookupTableRows[1], b&: block->ColourBlock.LookupTableRows[2]);  |
| 1442 | }  |
| 1443 | break;  |
| 1444 | }  |
| 1445 |   |
| 1446 | default:  |
| 1447 | // We should not get here. Should have early returned already.  |
| 1448 | tAssert(!"Should be unreachable." );  |
| 1449 | delete[] reversedPixelData;  |
| 1450 | return nullptr;  |
| 1451 | }  |
| 1452 |   |
| 1453 | return reversedPixelData;  |
| 1454 | }  |
| 1455 | |