John Bauman | 8940182 | 2014-05-06 15:04:28 -0400 | [diff] [blame] | 1 | // SwiftShader Software Renderer |
| 2 | // |
John Bauman | 66b8ab2 | 2014-05-06 15:57:45 -0400 | [diff] [blame^] | 3 | // Copyright(c) 2005-2013 TransGaming Inc. |
John Bauman | 8940182 | 2014-05-06 15:04:28 -0400 | [diff] [blame] | 4 | // |
| 5 | // All rights reserved. No part of this software may be copied, distributed, transmitted, |
| 6 | // transcribed, stored in a retrieval system, translated into any human or computer |
| 7 | // language by any means, or disclosed to third parties without the explicit written |
| 8 | // agreement of TransGaming Inc. Without such an agreement, no rights or licenses, express |
| 9 | // or implied, including but not limited to any patent rights, are granted to you. |
| 10 | // |
| 11 | |
| 12 | #include "Surface.hpp" |
| 13 | |
| 14 | #include "Color.hpp" |
| 15 | #include "Context.hpp" |
John Bauman | 19bac1e | 2014-05-06 15:23:49 -0400 | [diff] [blame] | 16 | #include "Renderer.hpp" |
John Bauman | 8940182 | 2014-05-06 15:04:28 -0400 | [diff] [blame] | 17 | #include "Common/Half.hpp" |
| 18 | #include "Common/Memory.hpp" |
| 19 | #include "Common/CPUID.hpp" |
| 20 | #include "Common/Resource.hpp" |
| 21 | #include "Common/Debug.hpp" |
John Bauman | 19bac1e | 2014-05-06 15:23:49 -0400 | [diff] [blame] | 22 | #include "Reactor/Reactor.hpp" |
John Bauman | 8940182 | 2014-05-06 15:04:28 -0400 | [diff] [blame] | 23 | |
| 24 | #include <xmmintrin.h> |
| 25 | #include <emmintrin.h> |
| 26 | |
| 27 | #undef min |
| 28 | #undef max |
| 29 | |
| 30 | namespace sw |
| 31 | { |
| 32 | extern bool quadLayoutEnabled; |
| 33 | extern bool complementaryDepthBuffer; |
| 34 | extern TranscendentalPrecision logPrecision; |
| 35 | |
| 36 | unsigned int *Surface::palette = 0; |
| 37 | unsigned int Surface::paletteID = 0; |
| 38 | |
John Bauman | 19bac1e | 2014-05-06 15:23:49 -0400 | [diff] [blame] | 39 | void Rect::clip(int minX, int minY, int maxX, int maxY) |
| 40 | { |
| 41 | x0 = sw::clamp(x0, minX, maxX); |
| 42 | y0 = sw::clamp(y0, minY, maxY); |
| 43 | x1 = sw::clamp(x1, minX, maxX); |
| 44 | y1 = sw::clamp(y1, minY, maxY); |
| 45 | } |
| 46 | |
John Bauman | 8940182 | 2014-05-06 15:04:28 -0400 | [diff] [blame] | 47 | void Surface::Buffer::write(int x, int y, int z, const Color<float> &color) |
| 48 | { |
| 49 | void *element = (unsigned char*)buffer + x * bytes + y * pitchB + z * sliceB; |
| 50 | |
| 51 | write(element, color); |
| 52 | } |
| 53 | |
| 54 | void Surface::Buffer::write(int x, int y, const Color<float> &color) |
| 55 | { |
| 56 | void *element = (unsigned char*)buffer + x * bytes + y * pitchB; |
| 57 | |
| 58 | write(element, color); |
| 59 | } |
| 60 | |
| 61 | inline void Surface::Buffer::write(void *element, const Color<float> &color) |
| 62 | { |
| 63 | switch(format) |
| 64 | { |
| 65 | case FORMAT_A8: |
| 66 | *(unsigned char*)element = unorm<8>(color.a); |
| 67 | break; |
| 68 | case FORMAT_R8: |
| 69 | *(unsigned char*)element = unorm<8>(color.r); |
| 70 | break; |
| 71 | case FORMAT_R3G3B2: |
| 72 | *(unsigned char*)element = (unorm<3>(color.r) << 5) | (unorm<3>(color.g) << 2) | (unorm<2>(color.b) << 0); |
| 73 | break; |
| 74 | case FORMAT_A8R3G3B2: |
| 75 | *(unsigned short*)element = (unorm<8>(color.a) << 8) | (unorm<3>(color.r) << 5) | (unorm<3>(color.g) << 2) | (unorm<2>(color.b) << 0); |
| 76 | break; |
| 77 | case FORMAT_X4R4G4B4: |
| 78 | *(unsigned short*)element = 0xF000 | (unorm<4>(color.r) << 8) | (unorm<4>(color.g) << 4) | (unorm<4>(color.b) << 0); |
| 79 | break; |
| 80 | case FORMAT_A4R4G4B4: |
| 81 | *(unsigned short*)element = (unorm<4>(color.a) << 12) | (unorm<4>(color.r) << 8) | (unorm<4>(color.g) << 4) | (unorm<4>(color.b) << 0); |
| 82 | break; |
| 83 | case FORMAT_R5G6B5: |
| 84 | *(unsigned short*)element = (unorm<5>(color.r) << 11) | (unorm<6>(color.g) << 5) | (unorm<5>(color.b) << 0); |
| 85 | break; |
| 86 | case FORMAT_A1R5G5B5: |
| 87 | *(unsigned short*)element = (unorm<1>(color.a) << 15) | (unorm<5>(color.r) << 10) | (unorm<5>(color.g) << 5) | (unorm<5>(color.b) << 0); |
| 88 | break; |
| 89 | case FORMAT_X1R5G5B5: |
| 90 | *(unsigned short*)element = 0x8000 | (unorm<5>(color.r) << 10) | (unorm<5>(color.g) << 5) | (unorm<5>(color.b) << 0); |
| 91 | break; |
| 92 | case FORMAT_A8R8G8B8: |
| 93 | *(unsigned int*)element = (unorm<8>(color.a) << 24) | (unorm<8>(color.r) << 16) | (unorm<8>(color.g) << 8) | (unorm<8>(color.b) << 0); |
| 94 | break; |
| 95 | case FORMAT_X8R8G8B8: |
| 96 | *(unsigned int*)element = 0xFF000000 | (unorm<8>(color.r) << 16) | (unorm<8>(color.g) << 8) | (unorm<8>(color.b) << 0); |
| 97 | break; |
| 98 | case FORMAT_A8B8G8R8: |
| 99 | *(unsigned int*)element = (unorm<8>(color.a) << 24) | (unorm<8>(color.b) << 16) | (unorm<8>(color.g) << 8) | (unorm<8>(color.r) << 0); |
| 100 | break; |
| 101 | case FORMAT_X8B8G8R8: |
| 102 | *(unsigned int*)element = 0xFF000000 | (unorm<8>(color.b) << 16) | (unorm<8>(color.g) << 8) | (unorm<8>(color.r) << 0); |
| 103 | break; |
| 104 | case FORMAT_A2R10G10B10: |
| 105 | *(unsigned int*)element = (unorm<2>(color.a) << 30) | (unorm<10>(color.r) << 20) | (unorm<10>(color.g) << 10) | (unorm<10>(color.b) << 0); |
| 106 | break; |
| 107 | case FORMAT_A2B10G10R10: |
| 108 | *(unsigned int*)element = (unorm<2>(color.a) << 30) | (unorm<10>(color.b) << 20) | (unorm<10>(color.g) << 10) | (unorm<10>(color.r) << 0); |
| 109 | break; |
| 110 | case FORMAT_G8R8: |
| 111 | *(unsigned int*)element = (unorm<8>(color.g) << 8) | (unorm<8>(color.r) << 0); |
| 112 | break; |
| 113 | case FORMAT_G16R16: |
| 114 | *(unsigned int*)element = (unorm<16>(color.g) << 16) | (unorm<16>(color.r) << 0); |
| 115 | break; |
| 116 | case FORMAT_A16B16G16R16: |
| 117 | ((unsigned short*)element)[0] = unorm<16>(color.r); |
| 118 | ((unsigned short*)element)[1] = unorm<16>(color.g); |
| 119 | ((unsigned short*)element)[2] = unorm<16>(color.b); |
| 120 | ((unsigned short*)element)[3] = unorm<16>(color.a); |
| 121 | break; |
| 122 | case FORMAT_V8U8: |
| 123 | *(unsigned short*)element = (snorm<8>(color.g) << 8) | (snorm<8>(color.r) << 0); |
| 124 | break; |
| 125 | case FORMAT_L6V5U5: |
| 126 | *(unsigned short*)element = (unorm<6>(color.b) << 10) | (snorm<5>(color.g) << 5) | (snorm<5>(color.r) << 0); |
| 127 | break; |
| 128 | case FORMAT_Q8W8V8U8: |
| 129 | *(unsigned int*)element = (snorm<8>(color.a) << 24) | (snorm<8>(color.b) << 16) | (snorm<8>(color.g) << 8) | (snorm<8>(color.r) << 0); |
| 130 | break; |
| 131 | case FORMAT_X8L8V8U8: |
| 132 | *(unsigned int*)element = 0xFF000000 | (unorm<8>(color.b) << 16) | (snorm<8>(color.g) << 8) | (snorm<8>(color.r) << 0); |
| 133 | break; |
| 134 | case FORMAT_V16U16: |
| 135 | *(unsigned int*)element = (snorm<16>(color.g) << 16) | (snorm<16>(color.r) << 0); |
| 136 | break; |
| 137 | case FORMAT_A2W10V10U10: |
| 138 | *(unsigned int*)element = (unorm<2>(color.a) << 30) | (snorm<10>(color.b) << 20) | (snorm<10>(color.g) << 10) | (snorm<10>(color.r) << 0); |
| 139 | break; |
| 140 | case FORMAT_A16W16V16U16: |
| 141 | ((unsigned short*)element)[0] = snorm<16>(color.r); |
| 142 | ((unsigned short*)element)[1] = snorm<16>(color.g); |
| 143 | ((unsigned short*)element)[2] = snorm<16>(color.b); |
| 144 | ((unsigned short*)element)[3] = unorm<16>(color.a); |
| 145 | break; |
| 146 | case FORMAT_Q16W16V16U16: |
| 147 | ((unsigned short*)element)[0] = snorm<16>(color.r); |
| 148 | ((unsigned short*)element)[1] = snorm<16>(color.g); |
| 149 | ((unsigned short*)element)[2] = snorm<16>(color.b); |
| 150 | ((unsigned short*)element)[3] = snorm<16>(color.a); |
| 151 | break; |
| 152 | case FORMAT_R8G8B8: |
| 153 | ((unsigned char*)element)[0] = unorm<8>(color.b); |
| 154 | ((unsigned char*)element)[1] = unorm<8>(color.g); |
| 155 | ((unsigned char*)element)[2] = unorm<8>(color.r); |
| 156 | break; |
| 157 | case FORMAT_R16F: |
| 158 | *(half*)element = (half)color.r; |
| 159 | break; |
| 160 | case FORMAT_G16R16F: |
| 161 | ((half*)element)[0] = (half)color.r; |
| 162 | ((half*)element)[1] = (half)color.g; |
| 163 | break; |
| 164 | case FORMAT_A16B16G16R16F: |
| 165 | ((half*)element)[0] = (half)color.r; |
| 166 | ((half*)element)[1] = (half)color.g; |
| 167 | ((half*)element)[2] = (half)color.b; |
| 168 | ((half*)element)[3] = (half)color.a; |
| 169 | break; |
| 170 | case FORMAT_R32F: |
| 171 | *(float*)element = color.r; |
| 172 | break; |
| 173 | case FORMAT_G32R32F: |
| 174 | ((float*)element)[0] = color.r; |
| 175 | ((float*)element)[1] = color.g; |
| 176 | break; |
| 177 | case FORMAT_A32B32G32R32F: |
| 178 | ((float*)element)[0] = color.r; |
| 179 | ((float*)element)[1] = color.g; |
| 180 | ((float*)element)[2] = color.b; |
| 181 | ((float*)element)[3] = color.a; |
| 182 | break; |
| 183 | case FORMAT_D32F: |
| 184 | case FORMAT_D32F_LOCKABLE: |
John Bauman | 66b8ab2 | 2014-05-06 15:57:45 -0400 | [diff] [blame^] | 185 | case FORMAT_D32FS8_TEXTURE: |
| 186 | case FORMAT_D32FS8_SHADOW: |
John Bauman | 8940182 | 2014-05-06 15:04:28 -0400 | [diff] [blame] | 187 | *((float*)element) = color.r; |
| 188 | break; |
| 189 | case FORMAT_D32F_COMPLEMENTARY: |
| 190 | *((float*)element) = 1 - color.r; |
| 191 | break; |
| 192 | case FORMAT_S8: |
| 193 | *((unsigned char*)element) = unorm<8>(color.r); |
| 194 | break; |
| 195 | case FORMAT_L8: |
| 196 | *(unsigned char*)element = unorm<8>(color.r); |
| 197 | break; |
| 198 | case FORMAT_A4L4: |
| 199 | *(unsigned char*)element = (unorm<4>(color.a) << 4) | (unorm<4>(color.r) << 0); |
| 200 | break; |
| 201 | case FORMAT_L16: |
| 202 | *(unsigned short*)element = unorm<16>(color.r); |
| 203 | break; |
| 204 | case FORMAT_A8L8: |
| 205 | *(unsigned short*)element = (unorm<8>(color.a) << 8) | (unorm<8>(color.r) << 0); |
| 206 | break; |
| 207 | default: |
| 208 | ASSERT(false); |
| 209 | } |
| 210 | } |
| 211 | |
| 212 | Color<float> Surface::Buffer::read(int x, int y, int z) const |
| 213 | { |
| 214 | void *element = (unsigned char*)buffer + x * bytes + y * pitchB + z * sliceB; |
| 215 | |
| 216 | return read(element); |
| 217 | } |
| 218 | |
| 219 | Color<float> Surface::Buffer::read(int x, int y) const |
| 220 | { |
| 221 | void *element = (unsigned char*)buffer + x * bytes + y * pitchB; |
| 222 | |
| 223 | return read(element); |
| 224 | } |
| 225 | |
| 226 | inline Color<float> Surface::Buffer::read(void *element) const |
| 227 | { |
| 228 | float r = 1; |
| 229 | float g = 1; |
| 230 | float b = 1; |
| 231 | float a = 1; |
| 232 | |
| 233 | switch(format) |
| 234 | { |
| 235 | case FORMAT_P8: |
| 236 | { |
| 237 | ASSERT(palette); |
| 238 | |
| 239 | unsigned int abgr = palette[*(unsigned char*)element]; |
| 240 | |
| 241 | r = (abgr & 0x000000FF) * (1.0f / 0x000000FF); |
| 242 | g = (abgr & 0x0000FF00) * (1.0f / 0x0000FF00); |
| 243 | b = (abgr & 0x00FF0000) * (1.0f / 0x00FF0000); |
| 244 | a = (abgr & 0xFF000000) * (1.0f / 0xFF000000); |
| 245 | } |
| 246 | break; |
| 247 | case FORMAT_A8P8: |
| 248 | { |
| 249 | ASSERT(palette); |
| 250 | |
| 251 | unsigned int bgr = palette[((unsigned char*)element)[0]]; |
| 252 | |
| 253 | r = (bgr & 0x000000FF) * (1.0f / 0x000000FF); |
| 254 | g = (bgr & 0x0000FF00) * (1.0f / 0x0000FF00); |
| 255 | b = (bgr & 0x00FF0000) * (1.0f / 0x00FF0000); |
| 256 | a = ((unsigned char*)element)[1] * (1.0f / 0xFF); |
| 257 | } |
| 258 | break; |
| 259 | case FORMAT_A8: |
| 260 | r = 0; |
| 261 | g = 0; |
| 262 | b = 0; |
| 263 | a = *(unsigned char*)element * (1.0f / 0xFF); |
| 264 | break; |
| 265 | case FORMAT_R8: |
| 266 | r = *(unsigned char*)element * (1.0f / 0xFF); |
| 267 | break; |
| 268 | case FORMAT_R3G3B2: |
| 269 | { |
| 270 | unsigned char rgb = *(unsigned char*)element; |
| 271 | |
| 272 | r = (rgb & 0xE0) * (1.0f / 0xE0); |
| 273 | g = (rgb & 0x1C) * (1.0f / 0x1C); |
| 274 | b = (rgb & 0x03) * (1.0f / 0x03); |
| 275 | } |
| 276 | break; |
| 277 | case FORMAT_A8R3G3B2: |
| 278 | { |
| 279 | unsigned short argb = *(unsigned short*)element; |
| 280 | |
| 281 | a = (argb & 0xFF00) * (1.0f / 0xFF00); |
| 282 | r = (argb & 0x00E0) * (1.0f / 0x00E0); |
| 283 | g = (argb & 0x001C) * (1.0f / 0x001C); |
| 284 | b = (argb & 0x0003) * (1.0f / 0x0003); |
| 285 | } |
| 286 | break; |
| 287 | case FORMAT_X4R4G4B4: |
| 288 | { |
| 289 | unsigned short rgb = *(unsigned short*)element; |
| 290 | |
| 291 | r = (rgb & 0x0F00) * (1.0f / 0x0F00); |
| 292 | g = (rgb & 0x00F0) * (1.0f / 0x00F0); |
| 293 | b = (rgb & 0x000F) * (1.0f / 0x000F); |
| 294 | } |
| 295 | break; |
| 296 | case FORMAT_A4R4G4B4: |
| 297 | { |
| 298 | unsigned short argb = *(unsigned short*)element; |
| 299 | |
| 300 | a = (argb & 0xF000) * (1.0f / 0xF000); |
| 301 | r = (argb & 0x0F00) * (1.0f / 0x0F00); |
| 302 | g = (argb & 0x00F0) * (1.0f / 0x00F0); |
| 303 | b = (argb & 0x000F) * (1.0f / 0x000F); |
| 304 | } |
| 305 | break; |
| 306 | case FORMAT_R5G6B5: |
| 307 | { |
| 308 | unsigned short rgb = *(unsigned short*)element; |
| 309 | |
| 310 | r = (rgb & 0xF800) * (1.0f / 0xF800); |
| 311 | g = (rgb & 0x07E0) * (1.0f / 0x07E0); |
| 312 | b = (rgb & 0x001F) * (1.0f / 0x001F); |
| 313 | } |
| 314 | break; |
| 315 | case FORMAT_A1R5G5B5: |
| 316 | { |
| 317 | unsigned short argb = *(unsigned short*)element; |
| 318 | |
| 319 | a = (argb & 0x8000) * (1.0f / 0x8000); |
| 320 | r = (argb & 0x7C00) * (1.0f / 0x7C00); |
| 321 | g = (argb & 0x03E0) * (1.0f / 0x03E0); |
| 322 | b = (argb & 0x001F) * (1.0f / 0x001F); |
| 323 | } |
| 324 | break; |
| 325 | case FORMAT_X1R5G5B5: |
| 326 | { |
| 327 | unsigned short xrgb = *(unsigned short*)element; |
| 328 | |
| 329 | r = (xrgb & 0x7C00) * (1.0f / 0x7C00); |
| 330 | g = (xrgb & 0x03E0) * (1.0f / 0x03E0); |
| 331 | b = (xrgb & 0x001F) * (1.0f / 0x001F); |
| 332 | } |
| 333 | break; |
| 334 | case FORMAT_A8R8G8B8: |
| 335 | { |
| 336 | unsigned int argb = *(unsigned int*)element; |
| 337 | |
| 338 | a = (argb & 0xFF000000) * (1.0f / 0xFF000000); |
| 339 | r = (argb & 0x00FF0000) * (1.0f / 0x00FF0000); |
| 340 | g = (argb & 0x0000FF00) * (1.0f / 0x0000FF00); |
| 341 | b = (argb & 0x000000FF) * (1.0f / 0x000000FF); |
| 342 | } |
| 343 | break; |
| 344 | case FORMAT_X8R8G8B8: |
| 345 | { |
| 346 | unsigned int xrgb = *(unsigned int*)element; |
| 347 | |
| 348 | r = (xrgb & 0x00FF0000) * (1.0f / 0x00FF0000); |
| 349 | g = (xrgb & 0x0000FF00) * (1.0f / 0x0000FF00); |
| 350 | b = (xrgb & 0x000000FF) * (1.0f / 0x000000FF); |
| 351 | } |
| 352 | break; |
| 353 | case FORMAT_A8B8G8R8: |
| 354 | { |
| 355 | unsigned int abgr = *(unsigned int*)element; |
| 356 | |
| 357 | a = (abgr & 0xFF000000) * (1.0f / 0xFF000000); |
| 358 | b = (abgr & 0x00FF0000) * (1.0f / 0x00FF0000); |
| 359 | g = (abgr & 0x0000FF00) * (1.0f / 0x0000FF00); |
| 360 | r = (abgr & 0x000000FF) * (1.0f / 0x000000FF); |
| 361 | } |
| 362 | break; |
| 363 | case FORMAT_X8B8G8R8: |
| 364 | { |
| 365 | unsigned int xbgr = *(unsigned int*)element; |
| 366 | |
| 367 | b = (xbgr & 0x00FF0000) * (1.0f / 0x00FF0000); |
| 368 | g = (xbgr & 0x0000FF00) * (1.0f / 0x0000FF00); |
| 369 | r = (xbgr & 0x000000FF) * (1.0f / 0x000000FF); |
| 370 | } |
| 371 | break; |
| 372 | case FORMAT_G8R8: |
| 373 | { |
| 374 | unsigned short gr = *(unsigned short*)element; |
| 375 | |
| 376 | g = (gr & 0xFF00) * (1.0f / 0xFF00); |
| 377 | r = (gr & 0x00FF) * (1.0f / 0x00FF); |
| 378 | } |
| 379 | break; |
| 380 | case FORMAT_G16R16: |
| 381 | { |
| 382 | unsigned int gr = *(unsigned int*)element; |
| 383 | |
| 384 | g = (gr & 0xFFFF0000) * (1.0f / 0xFFFF0000); |
| 385 | r = (gr & 0x0000FFFF) * (1.0f / 0x0000FFFF); |
| 386 | } |
| 387 | break; |
| 388 | case FORMAT_A2R10G10B10: |
| 389 | { |
| 390 | unsigned int argb = *(unsigned int*)element; |
| 391 | |
| 392 | a = (argb & 0xC0000000) * (1.0f / 0xC0000000); |
| 393 | r = (argb & 0x3FF00000) * (1.0f / 0x3FF00000); |
| 394 | g = (argb & 0x000FFC00) * (1.0f / 0x000FFC00); |
| 395 | b = (argb & 0x000003FF) * (1.0f / 0x000003FF); |
| 396 | } |
| 397 | break; |
| 398 | case FORMAT_A2B10G10R10: |
| 399 | { |
| 400 | unsigned int abgr = *(unsigned int*)element; |
| 401 | |
| 402 | a = (abgr & 0xC0000000) * (1.0f / 0xC0000000); |
| 403 | b = (abgr & 0x3FF00000) * (1.0f / 0x3FF00000); |
| 404 | g = (abgr & 0x000FFC00) * (1.0f / 0x000FFC00); |
| 405 | r = (abgr & 0x000003FF) * (1.0f / 0x000003FF); |
| 406 | } |
| 407 | break; |
| 408 | case FORMAT_A16B16G16R16: |
| 409 | r = ((unsigned short*)element)[0] * (1.0f / 0xFFFF); |
| 410 | g = ((unsigned short*)element)[1] * (1.0f / 0xFFFF); |
| 411 | b = ((unsigned short*)element)[2] * (1.0f / 0xFFFF); |
| 412 | a = ((unsigned short*)element)[3] * (1.0f / 0xFFFF); |
| 413 | break; |
| 414 | case FORMAT_V8U8: |
| 415 | { |
| 416 | unsigned short vu = *(unsigned short*)element; |
| 417 | |
| 418 | r = ((int)(vu & 0x00FF) << 24) * (1.0f / 0x7F000000); |
| 419 | g = ((int)(vu & 0xFF00) << 16) * (1.0f / 0x7F000000); |
| 420 | } |
| 421 | break; |
| 422 | case FORMAT_L6V5U5: |
| 423 | { |
| 424 | unsigned short lvu = *(unsigned short*)element; |
| 425 | |
| 426 | r = ((int)(lvu & 0x001F) << 27) * (1.0f / 0x78000000); |
| 427 | g = ((int)(lvu & 0x03E0) << 22) * (1.0f / 0x78000000); |
| 428 | b = (lvu & 0xFC00) * (1.0f / 0xFC00); |
| 429 | } |
| 430 | break; |
| 431 | case FORMAT_Q8W8V8U8: |
| 432 | { |
| 433 | unsigned int qwvu = *(unsigned int*)element; |
| 434 | |
| 435 | r = ((int)(qwvu & 0x000000FF) << 24) * (1.0f / 0x7F000000); |
| 436 | g = ((int)(qwvu & 0x0000FF00) << 16) * (1.0f / 0x7F000000); |
| 437 | b = ((int)(qwvu & 0x00FF0000) << 8) * (1.0f / 0x7F000000); |
| 438 | a = ((int)(qwvu & 0xFF000000) << 0) * (1.0f / 0x7F000000); |
| 439 | } |
| 440 | break; |
| 441 | case FORMAT_X8L8V8U8: |
| 442 | { |
| 443 | unsigned int xlvu = *(unsigned int*)element; |
| 444 | |
| 445 | r = ((int)(xlvu & 0x000000FF) << 24) * (1.0f / 0x7F000000); |
| 446 | g = ((int)(xlvu & 0x0000FF00) << 16) * (1.0f / 0x7F000000); |
| 447 | b = (xlvu & 0x00FF0000) * (1.0f / 0x00FF0000); |
| 448 | } |
| 449 | break; |
| 450 | case FORMAT_R8G8B8: |
| 451 | r = ((unsigned char*)element)[2] * (1.0f / 0xFF); |
| 452 | g = ((unsigned char*)element)[1] * (1.0f / 0xFF); |
| 453 | b = ((unsigned char*)element)[0] * (1.0f / 0xFF); |
| 454 | break; |
| 455 | case FORMAT_V16U16: |
| 456 | { |
| 457 | unsigned int vu = *(unsigned int*)element; |
| 458 | |
| 459 | r = ((int)(vu & 0x0000FFFF) << 16) * (1.0f / 0x7FFF0000); |
| 460 | g = ((int)(vu & 0xFFFF0000) << 0) * (1.0f / 0x7FFF0000); |
| 461 | } |
| 462 | break; |
| 463 | case FORMAT_A2W10V10U10: |
| 464 | { |
| 465 | unsigned int awvu = *(unsigned int*)element; |
| 466 | |
| 467 | r = ((int)(awvu & 0x000003FF) << 22) * (1.0f / 0x7FC00000); |
| 468 | g = ((int)(awvu & 0x000FFC00) << 12) * (1.0f / 0x7FC00000); |
| 469 | b = ((int)(awvu & 0x3FF00000) << 2) * (1.0f / 0x7FC00000); |
| 470 | a = (awvu & 0xC0000000) * (1.0f / 0xC0000000); |
| 471 | } |
| 472 | break; |
| 473 | case FORMAT_A16W16V16U16: |
| 474 | r = ((signed short*)element)[0] * (1.0f / 0x7FFF); |
| 475 | g = ((signed short*)element)[1] * (1.0f / 0x7FFF); |
| 476 | b = ((signed short*)element)[2] * (1.0f / 0x7FFF); |
| 477 | a = ((unsigned short*)element)[3] * (1.0f / 0xFFFF); |
| 478 | break; |
| 479 | case FORMAT_Q16W16V16U16: |
| 480 | r = ((signed short*)element)[0] * (1.0f / 0x7FFF); |
| 481 | g = ((signed short*)element)[1] * (1.0f / 0x7FFF); |
| 482 | b = ((signed short*)element)[2] * (1.0f / 0x7FFF); |
| 483 | a = ((signed short*)element)[3] * (1.0f / 0x7FFF); |
| 484 | break; |
| 485 | case FORMAT_L8: |
| 486 | r = |
| 487 | g = |
| 488 | b = *(unsigned char*)element * (1.0f / 0xFF); |
| 489 | break; |
| 490 | case FORMAT_A4L4: |
| 491 | { |
| 492 | unsigned char al = *(unsigned char*)element; |
| 493 | |
| 494 | r = |
| 495 | g = |
| 496 | b = (al & 0x0F) * (1.0f / 0x0F); |
| 497 | a = (al & 0xF0) * (1.0f / 0xF0); |
| 498 | } |
| 499 | break; |
| 500 | case FORMAT_L16: |
| 501 | r = |
| 502 | g = |
| 503 | b = *(unsigned short*)element * (1.0f / 0xFFFF); |
| 504 | break; |
| 505 | case FORMAT_A8L8: |
| 506 | r = |
| 507 | g = |
| 508 | b = ((unsigned char*)element)[0] * (1.0f / 0xFF); |
| 509 | a = ((unsigned char*)element)[1] * (1.0f / 0xFF); |
| 510 | break; |
| 511 | case FORMAT_R16F: |
| 512 | r = *(half*)element; |
| 513 | break; |
| 514 | case FORMAT_G16R16F: |
| 515 | r = ((half*)element)[0]; |
| 516 | g = ((half*)element)[1]; |
| 517 | break; |
| 518 | case FORMAT_A16B16G16R16F: |
| 519 | r = ((half*)element)[0]; |
| 520 | g = ((half*)element)[1]; |
| 521 | b = ((half*)element)[2]; |
| 522 | a = ((half*)element)[3]; |
| 523 | break; |
| 524 | case FORMAT_R32F: |
| 525 | r = *(float*)element; |
| 526 | break; |
| 527 | case FORMAT_G32R32F: |
| 528 | r = ((float*)element)[0]; |
| 529 | g = ((float*)element)[1]; |
| 530 | break; |
| 531 | case FORMAT_A32B32G32R32F: |
| 532 | r = ((float*)element)[0]; |
| 533 | g = ((float*)element)[1]; |
| 534 | b = ((float*)element)[2]; |
| 535 | a = ((float*)element)[3]; |
| 536 | break; |
| 537 | case FORMAT_D32F: |
| 538 | case FORMAT_D32F_LOCKABLE: |
John Bauman | 66b8ab2 | 2014-05-06 15:57:45 -0400 | [diff] [blame^] | 539 | case FORMAT_D32FS8_TEXTURE: |
| 540 | case FORMAT_D32FS8_SHADOW: |
John Bauman | 8940182 | 2014-05-06 15:04:28 -0400 | [diff] [blame] | 541 | r = *(float*)element; |
| 542 | g = r; |
| 543 | b = r; |
| 544 | a = r; |
| 545 | break; |
| 546 | case FORMAT_D32F_COMPLEMENTARY: |
John Bauman | 66b8ab2 | 2014-05-06 15:57:45 -0400 | [diff] [blame^] | 547 | r = 1.0f - *(float*)element; |
John Bauman | 8940182 | 2014-05-06 15:04:28 -0400 | [diff] [blame] | 548 | g = r; |
| 549 | b = r; |
| 550 | a = r; |
| 551 | break; |
| 552 | case FORMAT_S8: |
| 553 | r = *(unsigned char*)element * (1.0f / 0xFF); |
| 554 | break; |
| 555 | default: |
| 556 | ASSERT(false); |
| 557 | } |
| 558 | |
| 559 | // if(sRGB) |
| 560 | // { |
| 561 | // r = sRGBtoLinear(r); |
| 562 | // g = sRGBtoLinear(g); |
| 563 | // b = sRGBtoLinear(b); |
| 564 | // } |
| 565 | |
| 566 | return Color<float>(r, g, b, a); |
| 567 | } |
| 568 | |
| 569 | Color<float> Surface::Buffer::sample(float x, float y, float z) const |
| 570 | { |
| 571 | x -= 0.5f; |
| 572 | y -= 0.5f; |
| 573 | z -= 0.5f; |
| 574 | |
| 575 | int x0 = clamp((int)x, 0, width - 1); |
| 576 | int x1 = (x0 + 1 >= width) ? x0 : x0 + 1; |
| 577 | |
| 578 | int y0 = clamp((int)y, 0, height - 1); |
| 579 | int y1 = (y0 + 1 >= height) ? y0 : y0 + 1; |
| 580 | |
| 581 | int z0 = clamp((int)z, 0, depth - 1); |
| 582 | int z1 = (z0 + 1 >= depth) ? z0 : z0 + 1; |
| 583 | |
| 584 | Color<float> c000 = read(x0, y0, z0); |
| 585 | Color<float> c100 = read(x1, y0, z0); |
| 586 | Color<float> c010 = read(x0, y1, z0); |
| 587 | Color<float> c110 = read(x1, y1, z0); |
| 588 | Color<float> c001 = read(x0, y0, z1); |
| 589 | Color<float> c101 = read(x1, y0, z1); |
| 590 | Color<float> c011 = read(x0, y1, z1); |
| 591 | Color<float> c111 = read(x1, y1, z1); |
| 592 | |
| 593 | float fx = x - x0; |
| 594 | float fy = y - y0; |
| 595 | float fz = z - z0; |
| 596 | |
| 597 | c000 *= (1 - fx) * (1 - fy) * (1 - fz); |
| 598 | c100 *= fx * (1 - fy) * (1 - fz); |
| 599 | c010 *= (1 - fx) * fy * (1 - fz); |
| 600 | c110 *= fx * fy * (1 - fz); |
| 601 | c001 *= (1 - fx) * (1 - fy) * fz; |
| 602 | c101 *= fx * (1 - fy) * fz; |
| 603 | c011 *= (1 - fx) * fy * fz; |
| 604 | c111 *= fx * fy * fz; |
| 605 | |
| 606 | return c000 + c100 + c010 + c110 + c001 + c101 + c011 + c111; |
| 607 | } |
| 608 | |
| 609 | Color<float> Surface::Buffer::sample(float x, float y) const |
| 610 | { |
| 611 | x -= 0.5f; |
| 612 | y -= 0.5f; |
| 613 | |
| 614 | int x0 = clamp((int)x, 0, width - 1); |
| 615 | int x1 = (x0 + 1 >= width) ? x0 : x0 + 1; |
| 616 | |
| 617 | int y0 = clamp((int)y, 0, height - 1); |
| 618 | int y1 = (y0 + 1 >= height) ? y0 : y0 + 1; |
| 619 | |
| 620 | Color<float> c00 = read(x0, y0); |
| 621 | Color<float> c10 = read(x1, y0); |
| 622 | Color<float> c01 = read(x0, y1); |
| 623 | Color<float> c11 = read(x1, y1); |
| 624 | |
| 625 | float fx = x - x0; |
| 626 | float fy = y - y0; |
| 627 | |
| 628 | c00 *= (1 - fx) * (1 - fy); |
| 629 | c10 *= fx * (1 - fy); |
| 630 | c01 *= (1 - fx) * fy; |
| 631 | c11 *= fx * fy; |
| 632 | |
| 633 | return c00 + c10 + c01 + c11; |
| 634 | } |
| 635 | |
John Bauman | 19bac1e | 2014-05-06 15:23:49 -0400 | [diff] [blame] | 636 | void *Surface::Buffer::lockRect(int x, int y, int z, Lock lock) |
John Bauman | 8940182 | 2014-05-06 15:04:28 -0400 | [diff] [blame] | 637 | { |
| 638 | this->lock = lock; |
| 639 | |
| 640 | switch(lock) |
| 641 | { |
| 642 | case LOCK_UNLOCKED: |
| 643 | case LOCK_READONLY: |
| 644 | break; |
| 645 | case LOCK_WRITEONLY: |
| 646 | case LOCK_READWRITE: |
| 647 | case LOCK_DISCARD: |
| 648 | dirty = true; |
| 649 | break; |
| 650 | default: |
| 651 | ASSERT(false); |
| 652 | } |
| 653 | |
| 654 | switch(format) |
| 655 | { |
| 656 | #if S3TC_SUPPORT |
| 657 | case FORMAT_DXT1: |
John Bauman | 66b8ab2 | 2014-05-06 15:57:45 -0400 | [diff] [blame^] | 658 | #endif |
John Bauman | 8940182 | 2014-05-06 15:04:28 -0400 | [diff] [blame] | 659 | case FORMAT_ATI1: |
John Bauman | 19bac1e | 2014-05-06 15:23:49 -0400 | [diff] [blame] | 660 | return (unsigned char*)buffer + 8 * (x / 4) + (y / 4) * pitchB + z * sliceB; |
John Bauman | 66b8ab2 | 2014-05-06 15:57:45 -0400 | [diff] [blame^] | 661 | #if S3TC_SUPPORT |
John Bauman | 8940182 | 2014-05-06 15:04:28 -0400 | [diff] [blame] | 662 | case FORMAT_DXT3: |
| 663 | case FORMAT_DXT5: |
John Bauman | 66b8ab2 | 2014-05-06 15:57:45 -0400 | [diff] [blame^] | 664 | #endif |
John Bauman | 8940182 | 2014-05-06 15:04:28 -0400 | [diff] [blame] | 665 | case FORMAT_ATI2: |
John Bauman | 19bac1e | 2014-05-06 15:23:49 -0400 | [diff] [blame] | 666 | return (unsigned char*)buffer + 16 * (x / 4) + (y / 4) * pitchB + z * sliceB; |
John Bauman | 8940182 | 2014-05-06 15:04:28 -0400 | [diff] [blame] | 667 | default: |
John Bauman | 19bac1e | 2014-05-06 15:23:49 -0400 | [diff] [blame] | 668 | return (unsigned char*)buffer + x * bytes + y * pitchB + z * sliceB; |
John Bauman | 8940182 | 2014-05-06 15:04:28 -0400 | [diff] [blame] | 669 | } |
| 670 | |
| 671 | return 0; |
| 672 | } |
| 673 | |
| 674 | void Surface::Buffer::unlockRect() |
| 675 | { |
| 676 | lock = LOCK_UNLOCKED; |
| 677 | } |
| 678 | |
| 679 | Surface::Surface(Resource *texture, int width, int height, int depth, Format format, bool lockable, bool renderTarget) : lockable(lockable), renderTarget(renderTarget) |
| 680 | { |
| 681 | resource = texture ? texture : new Resource(0); |
John Bauman | 19bac1e | 2014-05-06 15:23:49 -0400 | [diff] [blame] | 682 | hasParent = texture != 0; |
John Bauman | 8940182 | 2014-05-06 15:04:28 -0400 | [diff] [blame] | 683 | depth = max(1, depth); |
| 684 | |
| 685 | external.buffer = 0; |
| 686 | external.width = width; |
| 687 | external.height = height; |
| 688 | external.depth = depth; |
| 689 | external.format = format; |
| 690 | external.bytes = bytes(external.format); |
| 691 | external.pitchB = pitchB(external.width, external.format, renderTarget && !texture); |
| 692 | external.pitchP = pitchP(external.width, external.format, renderTarget && !texture); |
| 693 | external.sliceB = sliceB(external.width, external.height, external.format, renderTarget && !texture); |
| 694 | external.sliceP = sliceP(external.width, external.height, external.format, renderTarget && !texture); |
| 695 | external.lock = LOCK_UNLOCKED; |
| 696 | external.dirty = false; |
John Bauman | 8940182 | 2014-05-06 15:04:28 -0400 | [diff] [blame] | 697 | |
| 698 | internal.buffer = 0; |
| 699 | internal.width = width; |
| 700 | internal.height = height; |
| 701 | internal.depth = depth; |
| 702 | internal.format = selectInternalFormat(format); |
| 703 | internal.bytes = bytes(internal.format); |
| 704 | internal.pitchB = pitchB(internal.width, internal.format, renderTarget); |
| 705 | internal.pitchP = pitchP(internal.width, internal.format, renderTarget); |
| 706 | internal.sliceB = sliceB(internal.width, internal.height, internal.format, renderTarget); |
| 707 | internal.sliceP = sliceP(internal.width, internal.height, internal.format, renderTarget); |
| 708 | internal.lock = LOCK_UNLOCKED; |
| 709 | internal.dirty = false; |
John Bauman | 8940182 | 2014-05-06 15:04:28 -0400 | [diff] [blame] | 710 | |
| 711 | stencil.buffer = 0; |
| 712 | stencil.width = width; |
| 713 | stencil.height = height; |
| 714 | stencil.depth = depth; |
| 715 | stencil.format = FORMAT_S8; |
| 716 | stencil.bytes = bytes(stencil.format); |
| 717 | stencil.pitchB = pitchB(stencil.width, stencil.format, renderTarget); |
| 718 | stencil.pitchP = pitchP(stencil.width, stencil.format, renderTarget); |
| 719 | stencil.sliceB = sliceB(stencil.width, stencil.height, stencil.format, renderTarget); |
| 720 | stencil.sliceP = sliceP(stencil.width, stencil.height, stencil.format, renderTarget); |
| 721 | stencil.lock = LOCK_UNLOCKED; |
| 722 | stencil.dirty = false; |
John Bauman | 8940182 | 2014-05-06 15:04:28 -0400 | [diff] [blame] | 723 | |
| 724 | dirtyMipmaps = true; |
John Bauman | 66b8ab2 | 2014-05-06 15:57:45 -0400 | [diff] [blame^] | 725 | paletteUsed = 0; |
John Bauman | 8940182 | 2014-05-06 15:04:28 -0400 | [diff] [blame] | 726 | } |
| 727 | |
| 728 | Surface::~Surface() |
| 729 | { |
John Bauman | 8a4f6fc | 2014-05-06 15:26:18 -0400 | [diff] [blame] | 730 | // Synchronize so we can deallocate the buffers below |
| 731 | resource->lock(DESTRUCT); |
| 732 | resource->unlock(); |
| 733 | |
John Bauman | 8940182 | 2014-05-06 15:04:28 -0400 | [diff] [blame] | 734 | if(!hasParent) |
| 735 | { |
| 736 | resource->destruct(); |
| 737 | } |
| 738 | |
| 739 | deallocate(external.buffer); |
| 740 | |
| 741 | if(internal.buffer != external.buffer) |
| 742 | { |
| 743 | deallocate(internal.buffer); |
| 744 | } |
| 745 | |
| 746 | deallocate(stencil.buffer); |
| 747 | |
| 748 | external.buffer = 0; |
| 749 | internal.buffer = 0; |
| 750 | stencil.buffer = 0; |
| 751 | } |
| 752 | |
John Bauman | 19bac1e | 2014-05-06 15:23:49 -0400 | [diff] [blame] | 753 | void *Surface::lockExternal(int x, int y, int z, Lock lock, Accessor client) |
John Bauman | 8940182 | 2014-05-06 15:04:28 -0400 | [diff] [blame] | 754 | { |
| 755 | resource->lock(client); |
| 756 | |
| 757 | if(!external.buffer) |
| 758 | { |
| 759 | if(internal.buffer && identicalFormats()) |
| 760 | { |
| 761 | external.buffer = internal.buffer; |
| 762 | } |
| 763 | else |
| 764 | { |
| 765 | external.buffer = allocateBuffer(external.width, external.height, external.depth, external.format); |
| 766 | } |
| 767 | } |
| 768 | |
| 769 | if(internal.dirty) |
| 770 | { |
| 771 | if(lock != LOCK_DISCARD) |
| 772 | { |
| 773 | update(external, internal); |
| 774 | } |
John Bauman | 66b8ab2 | 2014-05-06 15:57:45 -0400 | [diff] [blame^] | 775 | |
| 776 | internal.dirty = false; |
John Bauman | 8940182 | 2014-05-06 15:04:28 -0400 | [diff] [blame] | 777 | } |
| 778 | |
| 779 | switch(lock) |
| 780 | { |
| 781 | case LOCK_READONLY: |
| 782 | break; |
| 783 | case LOCK_WRITEONLY: |
| 784 | case LOCK_READWRITE: |
| 785 | case LOCK_DISCARD: |
| 786 | dirtyMipmaps = true; |
| 787 | break; |
| 788 | default: |
| 789 | ASSERT(false); |
| 790 | } |
| 791 | |
John Bauman | 19bac1e | 2014-05-06 15:23:49 -0400 | [diff] [blame] | 792 | return external.lockRect(x, y, z, lock); |
John Bauman | 8940182 | 2014-05-06 15:04:28 -0400 | [diff] [blame] | 793 | } |
| 794 | |
| 795 | void Surface::unlockExternal() |
| 796 | { |
| 797 | resource->unlock(); |
| 798 | |
| 799 | external.unlockRect(); |
| 800 | } |
| 801 | |
John Bauman | 19bac1e | 2014-05-06 15:23:49 -0400 | [diff] [blame] | 802 | void *Surface::lockInternal(int x, int y, int z, Lock lock, Accessor client) |
John Bauman | 8940182 | 2014-05-06 15:04:28 -0400 | [diff] [blame] | 803 | { |
| 804 | if(lock != LOCK_UNLOCKED) |
| 805 | { |
| 806 | resource->lock(client); |
| 807 | } |
| 808 | |
| 809 | if(!internal.buffer) |
| 810 | { |
| 811 | if(external.buffer && identicalFormats()) |
| 812 | { |
| 813 | internal.buffer = external.buffer; |
| 814 | } |
| 815 | else |
| 816 | { |
| 817 | internal.buffer = allocateBuffer(internal.width, internal.height, internal.depth, internal.format); |
| 818 | } |
| 819 | } |
| 820 | |
| 821 | // FIXME: WHQL requires conversion to lower external precision and back |
| 822 | if(logPrecision >= WHQL) |
| 823 | { |
| 824 | if(internal.dirty && renderTarget && internal.format != external.format) |
| 825 | { |
| 826 | if(lock != LOCK_DISCARD) |
| 827 | { |
| 828 | switch(external.format) |
| 829 | { |
| 830 | case FORMAT_R3G3B2: |
| 831 | case FORMAT_A8R3G3B2: |
| 832 | case FORMAT_A1R5G5B5: |
| 833 | case FORMAT_A2R10G10B10: |
| 834 | case FORMAT_A2B10G10R10: |
| 835 | lockExternal(0, 0, 0, LOCK_READWRITE, client); |
| 836 | unlockExternal(); |
| 837 | break; |
| 838 | default: |
| 839 | // Difference passes WHQL |
| 840 | break; |
| 841 | } |
| 842 | } |
| 843 | } |
| 844 | } |
| 845 | |
John Bauman | 66b8ab2 | 2014-05-06 15:57:45 -0400 | [diff] [blame^] | 846 | if(external.dirty || (isPalette(external.format) && paletteUsed != Surface::paletteID)) |
John Bauman | 8940182 | 2014-05-06 15:04:28 -0400 | [diff] [blame] | 847 | { |
| 848 | if(lock != LOCK_DISCARD) |
| 849 | { |
| 850 | update(internal, external); |
| 851 | } |
John Bauman | 8940182 | 2014-05-06 15:04:28 -0400 | [diff] [blame] | 852 | |
John Bauman | 66b8ab2 | 2014-05-06 15:57:45 -0400 | [diff] [blame^] | 853 | external.dirty = false; |
| 854 | paletteUsed = Surface::paletteID; |
John Bauman | 8940182 | 2014-05-06 15:04:28 -0400 | [diff] [blame] | 855 | } |
| 856 | |
| 857 | switch(lock) |
| 858 | { |
| 859 | case LOCK_UNLOCKED: |
| 860 | case LOCK_READONLY: |
| 861 | break; |
| 862 | case LOCK_WRITEONLY: |
| 863 | case LOCK_READWRITE: |
| 864 | case LOCK_DISCARD: |
| 865 | dirtyMipmaps = true; |
| 866 | break; |
| 867 | default: |
| 868 | ASSERT(false); |
| 869 | } |
| 870 | |
| 871 | if(lock == LOCK_READONLY && client == PUBLIC) |
| 872 | { |
| 873 | resolve(); |
| 874 | } |
| 875 | |
John Bauman | 19bac1e | 2014-05-06 15:23:49 -0400 | [diff] [blame] | 876 | return internal.lockRect(x, y, z, lock); |
John Bauman | 8940182 | 2014-05-06 15:04:28 -0400 | [diff] [blame] | 877 | } |
| 878 | |
| 879 | void Surface::unlockInternal() |
| 880 | { |
| 881 | resource->unlock(); |
| 882 | |
| 883 | internal.unlockRect(); |
| 884 | } |
| 885 | |
| 886 | void *Surface::lockStencil(int front, Accessor client) |
| 887 | { |
| 888 | resource->lock(client); |
| 889 | |
| 890 | if(!stencil.buffer) |
| 891 | { |
| 892 | stencil.buffer = allocateBuffer(stencil.width, stencil.height, stencil.depth, stencil.format); |
| 893 | } |
| 894 | |
John Bauman | 8940182 | 2014-05-06 15:04:28 -0400 | [diff] [blame] | 895 | return stencil.lockRect(0, 0, front, LOCK_READWRITE); // FIXME |
| 896 | } |
| 897 | |
| 898 | void Surface::unlockStencil() |
| 899 | { |
| 900 | resource->unlock(); |
| 901 | |
| 902 | stencil.unlockRect(); |
| 903 | } |
| 904 | |
| 905 | int Surface::bytes(Format format) |
| 906 | { |
| 907 | switch(format) |
| 908 | { |
| 909 | case FORMAT_NULL: return 0; |
| 910 | case FORMAT_P8: return 1; |
| 911 | case FORMAT_A8P8: return 2; |
| 912 | case FORMAT_A8: return 1; |
| 913 | case FORMAT_R8: return 1; |
| 914 | case FORMAT_R3G3B2: return 1; |
| 915 | case FORMAT_A8R3G3B2: return 2; |
| 916 | case FORMAT_R5G6B5: return 2; |
| 917 | case FORMAT_A1R5G5B5: return 2; |
| 918 | case FORMAT_X1R5G5B5: return 2; |
| 919 | case FORMAT_X4R4G4B4: return 2; |
| 920 | case FORMAT_A4R4G4B4: return 2; |
| 921 | case FORMAT_R8G8B8: return 3; |
| 922 | case FORMAT_X8R8G8B8: return 4; |
| 923 | // case FORMAT_X8G8R8B8Q: return 4; |
| 924 | case FORMAT_A8R8G8B8: return 4; |
| 925 | // case FORMAT_A8G8R8B8Q: return 4; |
| 926 | case FORMAT_X8B8G8R8: return 4; |
| 927 | case FORMAT_A8B8G8R8: return 4; |
| 928 | case FORMAT_A2R10G10B10: return 4; |
| 929 | case FORMAT_A2B10G10R10: return 4; |
| 930 | case FORMAT_G8R8: return 2; |
| 931 | case FORMAT_G16R16: return 4; |
| 932 | case FORMAT_A16B16G16R16: return 8; |
| 933 | // Compressed formats |
| 934 | #if S3TC_SUPPORT |
| 935 | case FORMAT_DXT1: return 2; // Column of four pixels |
| 936 | case FORMAT_DXT3: return 4; // Column of four pixels |
| 937 | case FORMAT_DXT5: return 4; // Column of four pixels |
John Bauman | 66b8ab2 | 2014-05-06 15:57:45 -0400 | [diff] [blame^] | 938 | #endif |
John Bauman | 8940182 | 2014-05-06 15:04:28 -0400 | [diff] [blame] | 939 | case FORMAT_ATI1: return 2; // Column of four pixels |
| 940 | case FORMAT_ATI2: return 4; // Column of four pixels |
John Bauman | 8940182 | 2014-05-06 15:04:28 -0400 | [diff] [blame] | 941 | // Bumpmap formats |
| 942 | case FORMAT_V8U8: return 2; |
| 943 | case FORMAT_L6V5U5: return 2; |
| 944 | case FORMAT_Q8W8V8U8: return 4; |
| 945 | case FORMAT_X8L8V8U8: return 4; |
| 946 | case FORMAT_A2W10V10U10: return 4; |
| 947 | case FORMAT_V16U16: return 4; |
| 948 | case FORMAT_A16W16V16U16: return 8; |
| 949 | case FORMAT_Q16W16V16U16: return 8; |
| 950 | // Luminance formats |
| 951 | case FORMAT_L8: return 1; |
| 952 | case FORMAT_A4L4: return 1; |
| 953 | case FORMAT_L16: return 2; |
| 954 | case FORMAT_A8L8: return 2; |
| 955 | // Floating-point formats |
| 956 | case FORMAT_R16F: return 2; |
| 957 | case FORMAT_G16R16F: return 4; |
| 958 | case FORMAT_A16B16G16R16F: return 8; |
| 959 | case FORMAT_R32F: return 4; |
| 960 | case FORMAT_G32R32F: return 8; |
| 961 | case FORMAT_A32B32G32R32F: return 16; |
| 962 | // Depth/stencil formats |
| 963 | case FORMAT_D16: return 2; |
| 964 | case FORMAT_D32: return 4; |
| 965 | case FORMAT_D24X8: return 4; |
| 966 | case FORMAT_D24S8: return 4; |
| 967 | case FORMAT_D24FS8: return 4; |
| 968 | case FORMAT_D32F: return 4; |
| 969 | case FORMAT_D32F_COMPLEMENTARY: return 4; |
| 970 | case FORMAT_D32F_LOCKABLE: return 4; |
John Bauman | 66b8ab2 | 2014-05-06 15:57:45 -0400 | [diff] [blame^] | 971 | case FORMAT_D32FS8_TEXTURE: return 4; |
| 972 | case FORMAT_D32FS8_SHADOW: return 4; |
| 973 | case FORMAT_DF24S8: return 4; |
| 974 | case FORMAT_DF16S8: return 2; |
John Bauman | 8940182 | 2014-05-06 15:04:28 -0400 | [diff] [blame] | 975 | case FORMAT_INTZ: return 4; |
| 976 | case FORMAT_S8: return 1; |
| 977 | default: |
| 978 | ASSERT(false); |
| 979 | } |
| 980 | |
| 981 | return 0; |
| 982 | } |
| 983 | |
| 984 | int Surface::pitchB(int width, Format format, bool target) |
| 985 | { |
| 986 | if(target || isDepth(format) || isStencil(format)) |
| 987 | { |
| 988 | width = ((width + 1) & ~1); |
| 989 | } |
| 990 | |
| 991 | switch(format) |
| 992 | { |
| 993 | #if S3TC_SUPPORT |
| 994 | case FORMAT_DXT1: |
| 995 | return 8 * ((width + 3) / 4); // 64 bit per 4x4 block, computed per 4 rows |
| 996 | case FORMAT_DXT3: |
| 997 | case FORMAT_DXT5: |
| 998 | return 16 * ((width + 3) / 4); // 128 bit per 4x4 block, computed per 4 rows |
John Bauman | 66b8ab2 | 2014-05-06 15:57:45 -0400 | [diff] [blame^] | 999 | #endif |
John Bauman | 8940182 | 2014-05-06 15:04:28 -0400 | [diff] [blame] | 1000 | case FORMAT_ATI1: |
| 1001 | return 2 * ((width + 3) / 4); // 64 bit per 4x4 block, computed per row |
| 1002 | case FORMAT_ATI2: |
| 1003 | return 4 * ((width + 3) / 4); // 128 bit per 4x4 block, computed per row |
John Bauman | 8940182 | 2014-05-06 15:04:28 -0400 | [diff] [blame] | 1004 | default: |
| 1005 | return bytes(format) * width; |
| 1006 | } |
| 1007 | } |
| 1008 | |
| 1009 | int Surface::pitchP(int width, Format format, bool target) |
| 1010 | { |
| 1011 | int B = bytes(format); |
| 1012 | |
| 1013 | return B > 0 ? pitchB(width, format, target) / B : 0; |
| 1014 | } |
| 1015 | |
| 1016 | int Surface::sliceB(int width, int height, Format format, bool target) |
| 1017 | { |
| 1018 | if(target || isDepth(format) || isStencil(format)) |
| 1019 | { |
| 1020 | height = ((height + 1) & ~1); |
| 1021 | } |
| 1022 | |
| 1023 | switch(format) |
| 1024 | { |
| 1025 | #if S3TC_SUPPORT |
| 1026 | case FORMAT_DXT1: |
| 1027 | case FORMAT_DXT3: |
| 1028 | case FORMAT_DXT5: |
| 1029 | return pitchB(width, format, target) * ((height + 3) / 4); // Pitch computed per 4 rows |
John Bauman | 66b8ab2 | 2014-05-06 15:57:45 -0400 | [diff] [blame^] | 1030 | #endif |
John Bauman | 8940182 | 2014-05-06 15:04:28 -0400 | [diff] [blame] | 1031 | case FORMAT_ATI1: // Pitch computed per row |
| 1032 | case FORMAT_ATI2: // Pitch computed per row |
John Bauman | 8940182 | 2014-05-06 15:04:28 -0400 | [diff] [blame] | 1033 | default: |
| 1034 | return pitchB(width, format, target) * height; |
| 1035 | } |
| 1036 | } |
| 1037 | |
| 1038 | int Surface::sliceP(int width, int height, Format format, bool target) |
| 1039 | { |
| 1040 | int B = bytes(format); |
| 1041 | |
| 1042 | return B > 0 ? sliceB(width, height, format, target) / B : 0; |
| 1043 | } |
| 1044 | |
| 1045 | void Surface::update(Buffer &destination, Buffer &source) |
| 1046 | { |
| 1047 | // ASSERT(source.lock != LOCK_UNLOCKED); |
| 1048 | // ASSERT(destination.lock != LOCK_UNLOCKED); |
| 1049 | |
| 1050 | if(destination.buffer != source.buffer) |
| 1051 | { |
| 1052 | ASSERT(source.dirty && !destination.dirty); |
| 1053 | |
| 1054 | switch(source.format) |
| 1055 | { |
| 1056 | case FORMAT_R8G8B8: decodeR8G8B8(destination, source); break; // FIXME: Check destination format |
| 1057 | case FORMAT_X8B8G8R8: decodeX8B8G8R8(destination, source); break; // FIXME: Check destination format |
| 1058 | case FORMAT_A8B8G8R8: decodeA8B8G8R8(destination, source); break; // FIXME: Check destination format |
| 1059 | case FORMAT_R5G6B5: decodeR5G6B5(destination, source); break; // FIXME: Check destination format |
| 1060 | case FORMAT_X1R5G5B5: decodeX1R5G5B5(destination, source); break; // FIXME: Check destination format |
| 1061 | case FORMAT_A1R5G5B5: decodeA1R5G5B5(destination, source); break; // FIXME: Check destination format |
| 1062 | case FORMAT_X4R4G4B4: decodeX4R4G4B4(destination, source); break; // FIXME: Check destination format |
| 1063 | case FORMAT_A4R4G4B4: decodeA4R4G4B4(destination, source); break; // FIXME: Check destination format |
| 1064 | case FORMAT_P8: decodeP8(destination, source); break; // FIXME: Check destination format |
| 1065 | #if S3TC_SUPPORT |
| 1066 | case FORMAT_DXT1: decodeDXT1(destination, source); break; // FIXME: Check destination format |
| 1067 | case FORMAT_DXT3: decodeDXT3(destination, source); break; // FIXME: Check destination format |
| 1068 | case FORMAT_DXT5: decodeDXT5(destination, source); break; // FIXME: Check destination format |
| 1069 | case FORMAT_ATI1: decodeATI1(destination, source); break; // FIXME: Check destination format |
| 1070 | case FORMAT_ATI2: decodeATI2(destination, source); break; // FIXME: Check destination format |
| 1071 | #endif |
| 1072 | default: genericUpdate(destination, source); break; |
| 1073 | } |
| 1074 | } |
John Bauman | 8940182 | 2014-05-06 15:04:28 -0400 | [diff] [blame] | 1075 | } |
| 1076 | |
| 1077 | void Surface::genericUpdate(Buffer &destination, Buffer &source) |
| 1078 | { |
| 1079 | unsigned char *sourceSlice = (unsigned char*)source.buffer; |
| 1080 | unsigned char *destinationSlice = (unsigned char*)destination.buffer; |
| 1081 | |
| 1082 | int depth = min(destination.depth, source.depth); |
| 1083 | int height = min(destination.height, source.height); |
| 1084 | int width = min(destination.width, source.width); |
| 1085 | int rowBytes = width * source.bytes; |
| 1086 | |
| 1087 | for(int z = 0; z < depth; z++) |
| 1088 | { |
| 1089 | unsigned char *sourceRow = sourceSlice; |
| 1090 | unsigned char *destinationRow = destinationSlice; |
| 1091 | |
| 1092 | for(int y = 0; y < height; y++) |
| 1093 | { |
| 1094 | if(source.format == destination.format) |
| 1095 | { |
| 1096 | memcpy(destinationRow, sourceRow, rowBytes); |
| 1097 | } |
| 1098 | else |
| 1099 | { |
| 1100 | unsigned char *sourceElement = sourceRow; |
| 1101 | unsigned char *destinationElement = destinationRow; |
| 1102 | |
| 1103 | for(int x = 0; x < width; x++) |
| 1104 | { |
| 1105 | Color<float> color = source.read(sourceElement); |
| 1106 | destination.write(destinationElement, color); |
| 1107 | |
| 1108 | sourceElement += source.bytes; |
| 1109 | destinationElement += destination.bytes; |
| 1110 | } |
| 1111 | } |
| 1112 | |
| 1113 | sourceRow += source.pitchB; |
| 1114 | destinationRow += destination.pitchB; |
| 1115 | } |
| 1116 | |
| 1117 | sourceSlice += source.sliceB; |
| 1118 | destinationSlice += destination.sliceB; |
| 1119 | } |
| 1120 | } |
| 1121 | |
| 1122 | void Surface::decodeR8G8B8(Buffer &destination, const Buffer &source) |
| 1123 | { |
| 1124 | unsigned char *sourceSlice = (unsigned char*)source.buffer; |
| 1125 | unsigned char *destinationSlice = (unsigned char*)destination.buffer; |
| 1126 | |
| 1127 | for(int z = 0; z < destination.depth && z < source.depth; z++) |
| 1128 | { |
| 1129 | unsigned char *sourceRow = sourceSlice; |
| 1130 | unsigned char *destinationRow = destinationSlice; |
| 1131 | |
| 1132 | for(int y = 0; y < destination.height && y < source.height; y++) |
| 1133 | { |
| 1134 | unsigned char *sourceElement = sourceRow; |
| 1135 | unsigned char *destinationElement = destinationRow; |
| 1136 | |
| 1137 | for(int x = 0; x < destination.width && x < source.width; x++) |
| 1138 | { |
| 1139 | unsigned int b = sourceElement[0]; |
| 1140 | unsigned int g = sourceElement[1]; |
| 1141 | unsigned int r = sourceElement[2]; |
| 1142 | |
| 1143 | *(unsigned int*)destinationElement = 0xFF000000 | (r << 16) | (g << 8) | (b << 0); |
| 1144 | |
| 1145 | sourceElement += source.bytes; |
| 1146 | destinationElement += destination.bytes; |
| 1147 | } |
| 1148 | |
| 1149 | sourceRow += source.pitchB; |
| 1150 | destinationRow += destination.pitchB; |
| 1151 | } |
| 1152 | |
| 1153 | sourceSlice += source.sliceB; |
| 1154 | destinationSlice += destination.sliceB; |
| 1155 | } |
| 1156 | } |
| 1157 | |
| 1158 | void Surface::decodeX8B8G8R8(Buffer &destination, const Buffer &source) |
| 1159 | { |
| 1160 | unsigned char *sourceSlice = (unsigned char*)source.buffer; |
| 1161 | unsigned char *destinationSlice = (unsigned char*)destination.buffer; |
| 1162 | |
| 1163 | for(int z = 0; z < destination.depth && z < source.depth; z++) |
| 1164 | { |
| 1165 | unsigned char *sourceRow = sourceSlice; |
| 1166 | unsigned char *destinationRow = destinationSlice; |
| 1167 | |
| 1168 | for(int y = 0; y < destination.height && y < source.height; y++) |
| 1169 | { |
| 1170 | unsigned char *sourceElement = sourceRow; |
| 1171 | unsigned char *destinationElement = destinationRow; |
| 1172 | |
| 1173 | for(int x = 0; x < destination.width && x < source.width; x++) |
| 1174 | { |
| 1175 | unsigned int r = sourceElement[0]; |
| 1176 | unsigned int g = sourceElement[1]; |
| 1177 | unsigned int b = sourceElement[2]; |
| 1178 | |
| 1179 | *(unsigned int*)destinationElement = 0xFF000000 | (r << 16) | (g << 8) | (b << 0); |
| 1180 | |
| 1181 | sourceElement += source.bytes; |
| 1182 | destinationElement += destination.bytes; |
| 1183 | } |
| 1184 | |
| 1185 | sourceRow += source.pitchB; |
| 1186 | destinationRow += destination.pitchB; |
| 1187 | } |
| 1188 | |
| 1189 | sourceSlice += source.sliceB; |
| 1190 | destinationSlice += destination.sliceB; |
| 1191 | } |
| 1192 | } |
| 1193 | |
| 1194 | void Surface::decodeA8B8G8R8(Buffer &destination, const Buffer &source) |
| 1195 | { |
| 1196 | unsigned char *sourceSlice = (unsigned char*)source.buffer; |
| 1197 | unsigned char *destinationSlice = (unsigned char*)destination.buffer; |
| 1198 | |
| 1199 | for(int z = 0; z < destination.depth && z < source.depth; z++) |
| 1200 | { |
| 1201 | unsigned char *sourceRow = sourceSlice; |
| 1202 | unsigned char *destinationRow = destinationSlice; |
| 1203 | |
| 1204 | for(int y = 0; y < destination.height && y < source.height; y++) |
| 1205 | { |
| 1206 | unsigned char *sourceElement = sourceRow; |
| 1207 | unsigned char *destinationElement = destinationRow; |
| 1208 | |
| 1209 | for(int x = 0; x < destination.width && x < source.width; x++) |
| 1210 | { |
| 1211 | unsigned int r = sourceElement[0]; |
| 1212 | unsigned int g = sourceElement[1]; |
| 1213 | unsigned int b = sourceElement[2]; |
| 1214 | unsigned int a = sourceElement[3]; |
| 1215 | |
| 1216 | *(unsigned int*)destinationElement = (a << 24) | (r << 16) | (g << 8) | (b << 0); |
| 1217 | |
| 1218 | sourceElement += source.bytes; |
| 1219 | destinationElement += destination.bytes; |
| 1220 | } |
| 1221 | |
| 1222 | sourceRow += source.pitchB; |
| 1223 | destinationRow += destination.pitchB; |
| 1224 | } |
| 1225 | |
| 1226 | sourceSlice += source.sliceB; |
| 1227 | destinationSlice += destination.sliceB; |
| 1228 | } |
| 1229 | } |
| 1230 | |
| 1231 | void Surface::decodeR5G6B5(Buffer &destination, const Buffer &source) |
| 1232 | { |
| 1233 | unsigned char *sourceSlice = (unsigned char*)source.buffer; |
| 1234 | unsigned char *destinationSlice = (unsigned char*)destination.buffer; |
| 1235 | |
| 1236 | for(int z = 0; z < destination.depth && z < source.depth; z++) |
| 1237 | { |
| 1238 | unsigned char *sourceRow = sourceSlice; |
| 1239 | unsigned char *destinationRow = destinationSlice; |
| 1240 | |
| 1241 | for(int y = 0; y < destination.height && y < source.height; y++) |
| 1242 | { |
| 1243 | unsigned char *sourceElement = sourceRow; |
| 1244 | unsigned char *destinationElement = destinationRow; |
| 1245 | |
| 1246 | for(int x = 0; x < destination.width && x < source.width; x++) |
| 1247 | { |
| 1248 | unsigned int rgb = *(unsigned short*)sourceElement; |
| 1249 | |
| 1250 | unsigned int r = (((rgb & 0xF800) * 67385 + 0x800000) >> 8) & 0x00FF0000; |
| 1251 | unsigned int g = (((rgb & 0x07E0) * 8289 + 0x8000) >> 8) & 0x0000FF00; |
| 1252 | unsigned int b = (((rgb & 0x001F) * 2106 + 0x80) >> 8); |
| 1253 | |
| 1254 | *(unsigned int*)destinationElement = 0xFF000000 | r | g | b; |
| 1255 | |
| 1256 | sourceElement += source.bytes; |
| 1257 | destinationElement += destination.bytes; |
| 1258 | } |
| 1259 | |
| 1260 | sourceRow += source.pitchB; |
| 1261 | destinationRow += destination.pitchB; |
| 1262 | } |
| 1263 | |
| 1264 | sourceSlice += source.sliceB; |
| 1265 | destinationSlice += destination.sliceB; |
| 1266 | } |
| 1267 | } |
| 1268 | |
| 1269 | void Surface::decodeX1R5G5B5(Buffer &destination, const Buffer &source) |
| 1270 | { |
| 1271 | unsigned char *sourceSlice = (unsigned char*)source.buffer; |
| 1272 | unsigned char *destinationSlice = (unsigned char*)destination.buffer; |
| 1273 | |
| 1274 | for(int z = 0; z < destination.depth && z < source.depth; z++) |
| 1275 | { |
| 1276 | unsigned char *sourceRow = sourceSlice; |
| 1277 | unsigned char *destinationRow = destinationSlice; |
| 1278 | |
| 1279 | for(int y = 0; y < destination.height && y < source.height; y++) |
| 1280 | { |
| 1281 | unsigned char *sourceElement = sourceRow; |
| 1282 | unsigned char *destinationElement = destinationRow; |
| 1283 | |
| 1284 | for(int x = 0; x < destination.width && x < source.width; x++) |
| 1285 | { |
| 1286 | unsigned int xrgb = *(unsigned short*)sourceElement; |
| 1287 | |
| 1288 | unsigned int r = (((xrgb & 0x7C00) * 134771 + 0x800000) >> 8) & 0x00FF0000; |
| 1289 | unsigned int g = (((xrgb & 0x03E0) * 16846 + 0x8000) >> 8) & 0x0000FF00; |
| 1290 | unsigned int b = (((xrgb & 0x001F) * 2106 + 0x80) >> 8); |
| 1291 | |
| 1292 | *(unsigned int*)destinationElement = 0xFF000000 | r | g | b; |
| 1293 | |
| 1294 | sourceElement += source.bytes; |
| 1295 | destinationElement += destination.bytes; |
| 1296 | } |
| 1297 | |
| 1298 | sourceRow += source.pitchB; |
| 1299 | destinationRow += destination.pitchB; |
| 1300 | } |
| 1301 | |
| 1302 | sourceSlice += source.sliceB; |
| 1303 | destinationSlice += destination.sliceB; |
| 1304 | } |
| 1305 | } |
| 1306 | |
| 1307 | void Surface::decodeA1R5G5B5(Buffer &destination, const Buffer &source) |
| 1308 | { |
| 1309 | unsigned char *sourceSlice = (unsigned char*)source.buffer; |
| 1310 | unsigned char *destinationSlice = (unsigned char*)destination.buffer; |
| 1311 | |
| 1312 | for(int z = 0; z < destination.depth && z < source.depth; z++) |
| 1313 | { |
| 1314 | unsigned char *sourceRow = sourceSlice; |
| 1315 | unsigned char *destinationRow = destinationSlice; |
| 1316 | |
| 1317 | for(int y = 0; y < destination.height && y < source.height; y++) |
| 1318 | { |
| 1319 | unsigned char *sourceElement = sourceRow; |
| 1320 | unsigned char *destinationElement = destinationRow; |
| 1321 | |
| 1322 | for(int x = 0; x < destination.width && x < source.width; x++) |
| 1323 | { |
| 1324 | unsigned int argb = *(unsigned short*)sourceElement; |
| 1325 | |
| 1326 | unsigned int a = (argb & 0x8000) * 130560; |
| 1327 | unsigned int r = (((argb & 0x7C00) * 134771 + 0x800000) >> 8) & 0x00FF0000; |
| 1328 | unsigned int g = (((argb & 0x03E0) * 16846 + 0x8000) >> 8) & 0x0000FF00; |
| 1329 | unsigned int b = (((argb & 0x001F) * 2106 + 0x80) >> 8); |
| 1330 | |
| 1331 | *(unsigned int*)destinationElement = a | r | g | b; |
| 1332 | |
| 1333 | sourceElement += source.bytes; |
| 1334 | destinationElement += destination.bytes; |
| 1335 | } |
| 1336 | |
| 1337 | sourceRow += source.pitchB; |
| 1338 | destinationRow += destination.pitchB; |
| 1339 | } |
| 1340 | |
| 1341 | sourceSlice += source.sliceB; |
| 1342 | destinationSlice += destination.sliceB; |
| 1343 | } |
| 1344 | } |
| 1345 | |
| 1346 | void Surface::decodeX4R4G4B4(Buffer &destination, const Buffer &source) |
| 1347 | { |
| 1348 | unsigned char *sourceSlice = (unsigned char*)source.buffer; |
| 1349 | unsigned char *destinationSlice = (unsigned char*)destination.buffer; |
| 1350 | |
| 1351 | for(int z = 0; z < destination.depth && z < source.depth; z++) |
| 1352 | { |
| 1353 | unsigned char *sourceRow = sourceSlice; |
| 1354 | unsigned char *destinationRow = destinationSlice; |
| 1355 | |
| 1356 | for(int y = 0; y < destination.height && y < source.height; y++) |
| 1357 | { |
| 1358 | unsigned char *sourceElement = sourceRow; |
| 1359 | unsigned char *destinationElement = destinationRow; |
| 1360 | |
| 1361 | for(int x = 0; x < destination.width && x < source.width; x++) |
| 1362 | { |
| 1363 | unsigned int xrgb = *(unsigned short*)sourceElement; |
| 1364 | |
| 1365 | unsigned int r = ((xrgb & 0x0F00) * 0x00001100) & 0x00FF0000; |
| 1366 | unsigned int g = ((xrgb & 0x00F0) * 0x00000110) & 0x0000FF00; |
| 1367 | unsigned int b = (xrgb & 0x000F) * 0x00000011; |
| 1368 | |
| 1369 | *(unsigned int*)destinationElement = 0xFF000000 | r | g | b; |
| 1370 | |
| 1371 | sourceElement += source.bytes; |
| 1372 | destinationElement += destination.bytes; |
| 1373 | } |
| 1374 | |
| 1375 | sourceRow += source.pitchB; |
| 1376 | destinationRow += destination.pitchB; |
| 1377 | } |
| 1378 | |
| 1379 | sourceSlice += source.sliceB; |
| 1380 | destinationSlice += destination.sliceB; |
| 1381 | } |
| 1382 | } |
| 1383 | |
| 1384 | void Surface::decodeA4R4G4B4(Buffer &destination, const Buffer &source) |
| 1385 | { |
| 1386 | unsigned char *sourceSlice = (unsigned char*)source.buffer; |
| 1387 | unsigned char *destinationSlice = (unsigned char*)destination.buffer; |
| 1388 | |
| 1389 | for(int z = 0; z < destination.depth && z < source.depth; z++) |
| 1390 | { |
| 1391 | unsigned char *sourceRow = sourceSlice; |
| 1392 | unsigned char *destinationRow = destinationSlice; |
| 1393 | |
| 1394 | for(int y = 0; y < destination.height && y < source.height; y++) |
| 1395 | { |
| 1396 | unsigned char *sourceElement = sourceRow; |
| 1397 | unsigned char *destinationElement = destinationRow; |
| 1398 | |
| 1399 | for(int x = 0; x < destination.width && x < source.width; x++) |
| 1400 | { |
| 1401 | unsigned int argb = *(unsigned short*)sourceElement; |
| 1402 | |
| 1403 | unsigned int a = ((argb & 0xF000) * 0x00011000) & 0xFF000000; |
| 1404 | unsigned int r = ((argb & 0x0F00) * 0x00001100) & 0x00FF0000; |
| 1405 | unsigned int g = ((argb & 0x00F0) * 0x00000110) & 0x0000FF00; |
| 1406 | unsigned int b = (argb & 0x000F) * 0x00000011; |
| 1407 | |
| 1408 | *(unsigned int*)destinationElement = a | r | g | b; |
| 1409 | |
| 1410 | sourceElement += source.bytes; |
| 1411 | destinationElement += destination.bytes; |
| 1412 | } |
| 1413 | |
| 1414 | sourceRow += source.pitchB; |
| 1415 | destinationRow += destination.pitchB; |
| 1416 | } |
| 1417 | |
| 1418 | sourceSlice += source.sliceB; |
| 1419 | destinationSlice += destination.sliceB; |
| 1420 | } |
| 1421 | } |
| 1422 | |
| 1423 | void Surface::decodeP8(Buffer &destination, const Buffer &source) |
| 1424 | { |
| 1425 | unsigned char *sourceSlice = (unsigned char*)source.buffer; |
| 1426 | unsigned char *destinationSlice = (unsigned char*)destination.buffer; |
| 1427 | |
| 1428 | for(int z = 0; z < destination.depth && z < source.depth; z++) |
| 1429 | { |
| 1430 | unsigned char *sourceRow = sourceSlice; |
| 1431 | unsigned char *destinationRow = destinationSlice; |
| 1432 | |
| 1433 | for(int y = 0; y < destination.height && y < source.height; y++) |
| 1434 | { |
| 1435 | unsigned char *sourceElement = sourceRow; |
| 1436 | unsigned char *destinationElement = destinationRow; |
| 1437 | |
| 1438 | for(int x = 0; x < destination.width && x < source.width; x++) |
| 1439 | { |
| 1440 | unsigned int abgr = palette[*(unsigned char*)sourceElement]; |
| 1441 | |
| 1442 | unsigned int r = (abgr & 0x000000FF) << 16; |
| 1443 | unsigned int g = (abgr & 0x0000FF00) << 0; |
| 1444 | unsigned int b = (abgr & 0x00FF0000) >> 16; |
| 1445 | unsigned int a = (abgr & 0xFF000000) >> 0; |
| 1446 | |
| 1447 | *(unsigned int*)destinationElement = a | r | g | b; |
| 1448 | |
| 1449 | sourceElement += source.bytes; |
| 1450 | destinationElement += destination.bytes; |
| 1451 | } |
| 1452 | |
| 1453 | sourceRow += source.pitchB; |
| 1454 | destinationRow += destination.pitchB; |
| 1455 | } |
| 1456 | |
| 1457 | sourceSlice += source.sliceB; |
| 1458 | destinationSlice += destination.sliceB; |
| 1459 | } |
| 1460 | } |
| 1461 | |
| 1462 | #if S3TC_SUPPORT |
| 1463 | void Surface::decodeDXT1(Buffer &internal, const Buffer &external) |
| 1464 | { |
| 1465 | unsigned int *destSlice = (unsigned int*)internal.buffer; |
| 1466 | DXT1 *source = (DXT1*)external.buffer; |
| 1467 | |
| 1468 | for(int z = 0; z < external.depth; z++) |
| 1469 | { |
| 1470 | unsigned int *dest = destSlice; |
| 1471 | |
| 1472 | for(int y = 0; y < external.height; y += 4) |
| 1473 | { |
| 1474 | for(int x = 0; x < external.width; x += 4) |
| 1475 | { |
| 1476 | Color<byte> c[4]; |
| 1477 | |
| 1478 | c[0] = source->c0; |
| 1479 | c[1] = source->c1; |
| 1480 | |
| 1481 | if(source->c0 > source->c1) // No transparency |
| 1482 | { |
| 1483 | // c2 = 2 / 3 * c0 + 1 / 3 * c1 |
| 1484 | c[2].r = (byte)((2 * (word)c[0].r + (word)c[1].r + 1) / 3); |
| 1485 | c[2].g = (byte)((2 * (word)c[0].g + (word)c[1].g + 1) / 3); |
| 1486 | c[2].b = (byte)((2 * (word)c[0].b + (word)c[1].b + 1) / 3); |
| 1487 | c[2].a = 0xFF; |
| 1488 | |
| 1489 | // c3 = 1 / 3 * c0 + 2 / 3 * c1 |
| 1490 | c[3].r = (byte)(((word)c[0].r + 2 * (word)c[1].r + 1) / 3); |
| 1491 | c[3].g = (byte)(((word)c[0].g + 2 * (word)c[1].g + 1) / 3); |
| 1492 | c[3].b = (byte)(((word)c[0].b + 2 * (word)c[1].b + 1) / 3); |
| 1493 | c[3].a = 0xFF; |
| 1494 | } |
| 1495 | else // c3 transparent |
| 1496 | { |
| 1497 | // c2 = 1 / 2 * c0 + 1 / 2 * c1 |
| 1498 | c[2].r = (byte)(((word)c[0].r + (word)c[1].r) / 2); |
| 1499 | c[2].g = (byte)(((word)c[0].g + (word)c[1].g) / 2); |
| 1500 | c[2].b = (byte)(((word)c[0].b + (word)c[1].b) / 2); |
| 1501 | c[2].a = 0xFF; |
| 1502 | |
| 1503 | c[3].r = 0; |
| 1504 | c[3].g = 0; |
| 1505 | c[3].b = 0; |
| 1506 | c[3].a = 0; |
| 1507 | } |
| 1508 | |
| 1509 | for(int j = 0; j < 4 && (y + j) < internal.height; j++) |
| 1510 | { |
| 1511 | for(int i = 0; i < 4 && (x + i) < internal.width; i++) |
| 1512 | { |
| 1513 | dest[(x + i) + (y + j) * internal.width] = c[(unsigned int)(source->lut >> 2 * (i + j * 4)) % 4]; |
| 1514 | } |
| 1515 | } |
| 1516 | |
| 1517 | source++; |
| 1518 | } |
| 1519 | } |
| 1520 | |
| 1521 | (byte*&)destSlice += internal.sliceB; |
| 1522 | } |
| 1523 | } |
| 1524 | |
| 1525 | void Surface::decodeDXT3(Buffer &internal, const Buffer &external) |
| 1526 | { |
| 1527 | unsigned int *destSlice = (unsigned int*)internal.buffer; |
| 1528 | DXT3 *source = (DXT3*)external.buffer; |
| 1529 | |
| 1530 | for(int z = 0; z < external.depth; z++) |
| 1531 | { |
| 1532 | unsigned int *dest = destSlice; |
| 1533 | |
| 1534 | for(int y = 0; y < external.height; y += 4) |
| 1535 | { |
| 1536 | for(int x = 0; x < external.width; x += 4) |
| 1537 | { |
| 1538 | Color<byte> c[4]; |
| 1539 | |
| 1540 | c[0] = source->c0; |
| 1541 | c[1] = source->c1; |
| 1542 | |
| 1543 | // c2 = 2 / 3 * c0 + 1 / 3 * c1 |
| 1544 | c[2].r = (byte)((2 * (word)c[0].r + (word)c[1].r + 1) / 3); |
| 1545 | c[2].g = (byte)((2 * (word)c[0].g + (word)c[1].g + 1) / 3); |
| 1546 | c[2].b = (byte)((2 * (word)c[0].b + (word)c[1].b + 1) / 3); |
| 1547 | |
| 1548 | // c3 = 1 / 3 * c0 + 2 / 3 * c1 |
| 1549 | c[3].r = (byte)(((word)c[0].r + 2 * (word)c[1].r + 1) / 3); |
| 1550 | c[3].g = (byte)(((word)c[0].g + 2 * (word)c[1].g + 1) / 3); |
| 1551 | c[3].b = (byte)(((word)c[0].b + 2 * (word)c[1].b + 1) / 3); |
| 1552 | |
| 1553 | for(int j = 0; j < 4 && (y + j) < internal.height; j++) |
| 1554 | { |
| 1555 | for(int i = 0; i < 4 && (x + i) < internal.width; i++) |
| 1556 | { |
| 1557 | unsigned int a = (unsigned int)(source->a >> 4 * (i + j * 4)) & 0x0F; |
| 1558 | unsigned int color = (c[(unsigned int)(source->lut >> 2 * (i + j * 4)) % 4] & 0x00FFFFFF) | ((a << 28) + (a << 24)); |
| 1559 | |
| 1560 | dest[(x + i) + (y + j) * internal.width] = color; |
| 1561 | } |
| 1562 | } |
| 1563 | |
| 1564 | source++; |
| 1565 | } |
| 1566 | } |
| 1567 | |
| 1568 | (byte*&)destSlice += internal.sliceB; |
| 1569 | } |
| 1570 | } |
| 1571 | |
| 1572 | void Surface::decodeDXT5(Buffer &internal, const Buffer &external) |
| 1573 | { |
| 1574 | unsigned int *destSlice = (unsigned int*)internal.buffer; |
| 1575 | DXT5 *source = (DXT5*)external.buffer; |
| 1576 | |
| 1577 | for(int z = 0; z < external.depth; z++) |
| 1578 | { |
| 1579 | unsigned int *dest = destSlice; |
| 1580 | |
| 1581 | for(int y = 0; y < external.height; y += 4) |
| 1582 | { |
| 1583 | for(int x = 0; x < external.width; x += 4) |
| 1584 | { |
| 1585 | Color<byte> c[4]; |
| 1586 | |
| 1587 | c[0] = source->c0; |
| 1588 | c[1] = source->c1; |
| 1589 | |
| 1590 | // c2 = 2 / 3 * c0 + 1 / 3 * c1 |
| 1591 | c[2].r = (byte)((2 * (word)c[0].r + (word)c[1].r + 1) / 3); |
| 1592 | c[2].g = (byte)((2 * (word)c[0].g + (word)c[1].g + 1) / 3); |
| 1593 | c[2].b = (byte)((2 * (word)c[0].b + (word)c[1].b + 1) / 3); |
| 1594 | |
| 1595 | // c3 = 1 / 3 * c0 + 2 / 3 * c1 |
| 1596 | c[3].r = (byte)(((word)c[0].r + 2 * (word)c[1].r + 1) / 3); |
| 1597 | c[3].g = (byte)(((word)c[0].g + 2 * (word)c[1].g + 1) / 3); |
| 1598 | c[3].b = (byte)(((word)c[0].b + 2 * (word)c[1].b + 1) / 3); |
| 1599 | |
| 1600 | byte a[8]; |
| 1601 | |
| 1602 | a[0] = source->a0; |
| 1603 | a[1] = source->a1; |
| 1604 | |
| 1605 | if(a[0] > a[1]) |
| 1606 | { |
| 1607 | a[2] = (byte)((6 * (word)a[0] + 1 * (word)a[1] + 3) / 7); |
| 1608 | a[3] = (byte)((5 * (word)a[0] + 2 * (word)a[1] + 3) / 7); |
| 1609 | a[4] = (byte)((4 * (word)a[0] + 3 * (word)a[1] + 3) / 7); |
| 1610 | a[5] = (byte)((3 * (word)a[0] + 4 * (word)a[1] + 3) / 7); |
| 1611 | a[6] = (byte)((2 * (word)a[0] + 5 * (word)a[1] + 3) / 7); |
| 1612 | a[7] = (byte)((1 * (word)a[0] + 6 * (word)a[1] + 3) / 7); |
| 1613 | } |
| 1614 | else |
| 1615 | { |
| 1616 | a[2] = (byte)((4 * (word)a[0] + 1 * (word)a[1] + 2) / 5); |
| 1617 | a[3] = (byte)((3 * (word)a[0] + 2 * (word)a[1] + 2) / 5); |
| 1618 | a[4] = (byte)((2 * (word)a[0] + 3 * (word)a[1] + 2) / 5); |
| 1619 | a[5] = (byte)((1 * (word)a[0] + 4 * (word)a[1] + 2) / 5); |
| 1620 | a[6] = 0; |
| 1621 | a[7] = 0xFF; |
| 1622 | } |
| 1623 | |
| 1624 | for(int j = 0; j < 4 && (y + j) < internal.height; j++) |
| 1625 | { |
| 1626 | for(int i = 0; i < 4 && (x + i) < internal.width; i++) |
| 1627 | { |
| 1628 | unsigned int alpha = (unsigned int)a[(unsigned int)(source->alut >> (16 + 3 * (i + j * 4))) % 8] << 24; |
| 1629 | unsigned int color = (c[(source->clut >> 2 * (i + j * 4)) % 4] & 0x00FFFFFF) | alpha; |
| 1630 | |
| 1631 | dest[(x + i) + (y + j) * internal.width] = color; |
| 1632 | } |
| 1633 | } |
| 1634 | |
| 1635 | source++; |
| 1636 | } |
| 1637 | } |
| 1638 | |
| 1639 | (byte*&)destSlice += internal.sliceB; |
| 1640 | } |
| 1641 | } |
| 1642 | |
| 1643 | void Surface::decodeATI1(Buffer &internal, const Buffer &external) |
| 1644 | { |
| 1645 | byte *destSlice = (byte*)internal.buffer; |
| 1646 | ATI1 *source = (ATI1*)external.buffer; |
| 1647 | |
| 1648 | for(int z = 0; z < external.depth; z++) |
| 1649 | { |
| 1650 | byte *dest = destSlice; |
| 1651 | |
| 1652 | for(int y = 0; y < external.height; y += 4) |
| 1653 | { |
| 1654 | for(int x = 0; x < external.width; x += 4) |
| 1655 | { |
| 1656 | byte r[8]; |
| 1657 | |
| 1658 | r[0] = source->r0; |
| 1659 | r[1] = source->r1; |
| 1660 | |
| 1661 | if(r[0] > r[1]) |
| 1662 | { |
| 1663 | r[2] = (byte)((6 * (word)r[0] + 1 * (word)r[1] + 3) / 7); |
| 1664 | r[3] = (byte)((5 * (word)r[0] + 2 * (word)r[1] + 3) / 7); |
| 1665 | r[4] = (byte)((4 * (word)r[0] + 3 * (word)r[1] + 3) / 7); |
| 1666 | r[5] = (byte)((3 * (word)r[0] + 4 * (word)r[1] + 3) / 7); |
| 1667 | r[6] = (byte)((2 * (word)r[0] + 5 * (word)r[1] + 3) / 7); |
| 1668 | r[7] = (byte)((1 * (word)r[0] + 6 * (word)r[1] + 3) / 7); |
| 1669 | } |
| 1670 | else |
| 1671 | { |
| 1672 | r[2] = (byte)((4 * (word)r[0] + 1 * (word)r[1] + 2) / 5); |
| 1673 | r[3] = (byte)((3 * (word)r[0] + 2 * (word)r[1] + 2) / 5); |
| 1674 | r[4] = (byte)((2 * (word)r[0] + 3 * (word)r[1] + 2) / 5); |
| 1675 | r[5] = (byte)((1 * (word)r[0] + 4 * (word)r[1] + 2) / 5); |
| 1676 | r[6] = 0; |
| 1677 | r[7] = 0xFF; |
| 1678 | } |
| 1679 | |
| 1680 | for(int j = 0; j < 4 && (y + j) < internal.height; j++) |
| 1681 | { |
| 1682 | for(int i = 0; i < 4 && (x + i) < internal.width; i++) |
| 1683 | { |
| 1684 | dest[(x + i) + (y + j) * internal.width] = r[(unsigned int)(source->rlut >> (16 + 3 * (i + j * 4))) % 8]; |
| 1685 | } |
| 1686 | } |
| 1687 | |
| 1688 | source++; |
| 1689 | } |
| 1690 | } |
| 1691 | |
| 1692 | destSlice += internal.sliceB; |
| 1693 | } |
| 1694 | } |
| 1695 | |
| 1696 | void Surface::decodeATI2(Buffer &internal, const Buffer &external) |
| 1697 | { |
| 1698 | word *destSlice = (word*)internal.buffer; |
| 1699 | ATI2 *source = (ATI2*)external.buffer; |
| 1700 | |
| 1701 | for(int z = 0; z < external.depth; z++) |
| 1702 | { |
| 1703 | word *dest = destSlice; |
| 1704 | |
| 1705 | for(int y = 0; y < external.height; y += 4) |
| 1706 | { |
| 1707 | for(int x = 0; x < external.width; x += 4) |
| 1708 | { |
| 1709 | byte X[8]; |
| 1710 | |
| 1711 | X[0] = source->x0; |
| 1712 | X[1] = source->x1; |
| 1713 | |
| 1714 | if(X[0] > X[1]) |
| 1715 | { |
| 1716 | X[2] = (byte)((6 * (word)X[0] + 1 * (word)X[1] + 3) / 7); |
| 1717 | X[3] = (byte)((5 * (word)X[0] + 2 * (word)X[1] + 3) / 7); |
| 1718 | X[4] = (byte)((4 * (word)X[0] + 3 * (word)X[1] + 3) / 7); |
| 1719 | X[5] = (byte)((3 * (word)X[0] + 4 * (word)X[1] + 3) / 7); |
| 1720 | X[6] = (byte)((2 * (word)X[0] + 5 * (word)X[1] + 3) / 7); |
| 1721 | X[7] = (byte)((1 * (word)X[0] + 6 * (word)X[1] + 3) / 7); |
| 1722 | } |
| 1723 | else |
| 1724 | { |
| 1725 | X[2] = (byte)((4 * (word)X[0] + 1 * (word)X[1] + 2) / 5); |
| 1726 | X[3] = (byte)((3 * (word)X[0] + 2 * (word)X[1] + 2) / 5); |
| 1727 | X[4] = (byte)((2 * (word)X[0] + 3 * (word)X[1] + 2) / 5); |
| 1728 | X[5] = (byte)((1 * (word)X[0] + 4 * (word)X[1] + 2) / 5); |
| 1729 | X[6] = 0; |
| 1730 | X[7] = 0xFF; |
| 1731 | } |
| 1732 | |
| 1733 | byte Y[8]; |
| 1734 | |
| 1735 | Y[0] = source->y0; |
| 1736 | Y[1] = source->y1; |
| 1737 | |
| 1738 | if(Y[0] > Y[1]) |
| 1739 | { |
| 1740 | Y[2] = (byte)((6 * (word)Y[0] + 1 * (word)Y[1] + 3) / 7); |
| 1741 | Y[3] = (byte)((5 * (word)Y[0] + 2 * (word)Y[1] + 3) / 7); |
| 1742 | Y[4] = (byte)((4 * (word)Y[0] + 3 * (word)Y[1] + 3) / 7); |
| 1743 | Y[5] = (byte)((3 * (word)Y[0] + 4 * (word)Y[1] + 3) / 7); |
| 1744 | Y[6] = (byte)((2 * (word)Y[0] + 5 * (word)Y[1] + 3) / 7); |
| 1745 | Y[7] = (byte)((1 * (word)Y[0] + 6 * (word)Y[1] + 3) / 7); |
| 1746 | } |
| 1747 | else |
| 1748 | { |
| 1749 | Y[2] = (byte)((4 * (word)Y[0] + 1 * (word)Y[1] + 2) / 5); |
| 1750 | Y[3] = (byte)((3 * (word)Y[0] + 2 * (word)Y[1] + 2) / 5); |
| 1751 | Y[4] = (byte)((2 * (word)Y[0] + 3 * (word)Y[1] + 2) / 5); |
| 1752 | Y[5] = (byte)((1 * (word)Y[0] + 4 * (word)Y[1] + 2) / 5); |
| 1753 | Y[6] = 0; |
| 1754 | Y[7] = 0xFF; |
| 1755 | } |
| 1756 | |
| 1757 | for(int j = 0; j < 4 && (y + j) < internal.height; j++) |
| 1758 | { |
| 1759 | for(int i = 0; i < 4 && (x + i) < internal.width; i++) |
| 1760 | { |
| 1761 | word r = X[(unsigned int)(source->xlut >> (16 + 3 * (i + j * 4))) % 8]; |
| 1762 | word g = Y[(unsigned int)(source->ylut >> (16 + 3 * (i + j * 4))) % 8]; |
| 1763 | |
| 1764 | dest[(x + i) + (y + j) * internal.width] = (g << 8) + r; |
| 1765 | } |
| 1766 | } |
| 1767 | |
| 1768 | source++; |
| 1769 | } |
| 1770 | } |
| 1771 | |
| 1772 | (byte*&)destSlice += internal.sliceB; |
| 1773 | } |
| 1774 | } |
| 1775 | #endif |
| 1776 | |
| 1777 | unsigned int Surface::size(int width, int height, int depth, Format format) |
| 1778 | { |
| 1779 | // Dimensions rounded up to multiples of 4, used for DXTC formats |
| 1780 | int width4 = (width + 3) & ~3; |
| 1781 | int height4 = (height + 3) & ~3; |
| 1782 | |
| 1783 | switch(format) |
| 1784 | { |
| 1785 | #if S3TC_SUPPORT |
| 1786 | case FORMAT_DXT1: |
John Bauman | 66b8ab2 | 2014-05-06 15:57:45 -0400 | [diff] [blame^] | 1787 | #endif |
John Bauman | 8940182 | 2014-05-06 15:04:28 -0400 | [diff] [blame] | 1788 | case FORMAT_ATI1: |
| 1789 | return width4 * height4 * depth / 2; |
John Bauman | 66b8ab2 | 2014-05-06 15:57:45 -0400 | [diff] [blame^] | 1790 | #if S3TC_SUPPORT |
John Bauman | 8940182 | 2014-05-06 15:04:28 -0400 | [diff] [blame] | 1791 | case FORMAT_DXT3: |
| 1792 | case FORMAT_DXT5: |
John Bauman | 66b8ab2 | 2014-05-06 15:57:45 -0400 | [diff] [blame^] | 1793 | #endif |
John Bauman | 8940182 | 2014-05-06 15:04:28 -0400 | [diff] [blame] | 1794 | case FORMAT_ATI2: |
| 1795 | return width4 * height4 * depth; |
John Bauman | 8940182 | 2014-05-06 15:04:28 -0400 | [diff] [blame] | 1796 | default: |
| 1797 | return bytes(format) * width * height * depth; |
| 1798 | } |
| 1799 | |
| 1800 | return 0; |
| 1801 | } |
| 1802 | |
| 1803 | bool Surface::isStencil(Format format) |
| 1804 | { |
| 1805 | switch(format) |
| 1806 | { |
| 1807 | case FORMAT_D32: |
| 1808 | case FORMAT_D16: |
| 1809 | case FORMAT_D24X8: |
| 1810 | case FORMAT_D32F: |
| 1811 | case FORMAT_D32F_COMPLEMENTARY: |
| 1812 | case FORMAT_D32F_LOCKABLE: |
| 1813 | return false; |
| 1814 | case FORMAT_D24S8: |
| 1815 | case FORMAT_D24FS8: |
| 1816 | case FORMAT_S8: |
John Bauman | 66b8ab2 | 2014-05-06 15:57:45 -0400 | [diff] [blame^] | 1817 | case FORMAT_DF24S8: |
| 1818 | case FORMAT_DF16S8: |
| 1819 | case FORMAT_D32FS8_TEXTURE: |
| 1820 | case FORMAT_D32FS8_SHADOW: |
John Bauman | 8940182 | 2014-05-06 15:04:28 -0400 | [diff] [blame] | 1821 | case FORMAT_INTZ: |
| 1822 | return true; |
| 1823 | default: |
| 1824 | return false; |
| 1825 | } |
| 1826 | } |
| 1827 | |
| 1828 | bool Surface::isDepth(Format format) |
| 1829 | { |
| 1830 | switch(format) |
| 1831 | { |
| 1832 | case FORMAT_D32: |
| 1833 | case FORMAT_D16: |
| 1834 | case FORMAT_D24X8: |
| 1835 | case FORMAT_D24S8: |
| 1836 | case FORMAT_D24FS8: |
| 1837 | case FORMAT_D32F: |
| 1838 | case FORMAT_D32F_COMPLEMENTARY: |
| 1839 | case FORMAT_D32F_LOCKABLE: |
John Bauman | 66b8ab2 | 2014-05-06 15:57:45 -0400 | [diff] [blame^] | 1840 | case FORMAT_DF24S8: |
| 1841 | case FORMAT_DF16S8: |
| 1842 | case FORMAT_D32FS8_TEXTURE: |
| 1843 | case FORMAT_D32FS8_SHADOW: |
John Bauman | 8940182 | 2014-05-06 15:04:28 -0400 | [diff] [blame] | 1844 | case FORMAT_INTZ: |
| 1845 | return true; |
| 1846 | case FORMAT_S8: |
| 1847 | return false; |
| 1848 | default: |
| 1849 | return false; |
| 1850 | } |
| 1851 | } |
| 1852 | |
| 1853 | bool Surface::isPalette(Format format) |
| 1854 | { |
| 1855 | switch(format) |
| 1856 | { |
| 1857 | case FORMAT_P8: |
| 1858 | case FORMAT_A8P8: |
| 1859 | return true; |
| 1860 | default: |
| 1861 | return false; |
| 1862 | } |
| 1863 | } |
| 1864 | |
| 1865 | bool Surface::isFloatFormat(Format format) |
| 1866 | { |
| 1867 | switch(format) |
| 1868 | { |
| 1869 | case FORMAT_X8R8G8B8: |
| 1870 | case FORMAT_A8R8G8B8: |
| 1871 | case FORMAT_G8R8: |
| 1872 | case FORMAT_G16R16: |
| 1873 | case FORMAT_A16B16G16R16: |
| 1874 | case FORMAT_V8U8: |
| 1875 | case FORMAT_Q8W8V8U8: |
| 1876 | case FORMAT_X8L8V8U8: |
| 1877 | case FORMAT_V16U16: |
| 1878 | case FORMAT_A16W16V16U16: |
| 1879 | case FORMAT_Q16W16V16U16: |
| 1880 | case FORMAT_A8: |
| 1881 | case FORMAT_R8: |
| 1882 | case FORMAT_L8: |
| 1883 | case FORMAT_L16: |
| 1884 | case FORMAT_A8L8: |
| 1885 | return false; |
| 1886 | case FORMAT_R32F: |
| 1887 | case FORMAT_G32R32F: |
| 1888 | case FORMAT_A32B32G32R32F: |
| 1889 | case FORMAT_D32F: |
| 1890 | case FORMAT_D32F_COMPLEMENTARY: |
| 1891 | case FORMAT_D32F_LOCKABLE: |
John Bauman | 66b8ab2 | 2014-05-06 15:57:45 -0400 | [diff] [blame^] | 1892 | case FORMAT_D32FS8_TEXTURE: |
| 1893 | case FORMAT_D32FS8_SHADOW: |
John Bauman | 8940182 | 2014-05-06 15:04:28 -0400 | [diff] [blame] | 1894 | return true; |
| 1895 | default: |
| 1896 | ASSERT(false); |
| 1897 | } |
| 1898 | |
| 1899 | return false; |
| 1900 | } |
| 1901 | |
| 1902 | bool Surface::isUnsignedComponent(Format format, int component) |
| 1903 | { |
| 1904 | switch(format) |
| 1905 | { |
| 1906 | case FORMAT_NULL: |
| 1907 | case FORMAT_X8R8G8B8: |
| 1908 | case FORMAT_A8R8G8B8: |
| 1909 | case FORMAT_G8R8: |
| 1910 | case FORMAT_G16R16: |
| 1911 | case FORMAT_A16B16G16R16: |
| 1912 | case FORMAT_D32F: |
| 1913 | case FORMAT_D32F_COMPLEMENTARY: |
| 1914 | case FORMAT_D32F_LOCKABLE: |
John Bauman | 66b8ab2 | 2014-05-06 15:57:45 -0400 | [diff] [blame^] | 1915 | case FORMAT_D32FS8_TEXTURE: |
| 1916 | case FORMAT_D32FS8_SHADOW: |
John Bauman | 8940182 | 2014-05-06 15:04:28 -0400 | [diff] [blame] | 1917 | case FORMAT_A8: |
| 1918 | case FORMAT_R8: |
| 1919 | case FORMAT_L8: |
| 1920 | case FORMAT_L16: |
| 1921 | case FORMAT_A8L8: |
| 1922 | return true; |
| 1923 | case FORMAT_V8U8: |
| 1924 | case FORMAT_X8L8V8U8: |
| 1925 | case FORMAT_V16U16: |
| 1926 | if(component < 2) |
| 1927 | { |
| 1928 | return false; |
| 1929 | } |
| 1930 | else |
| 1931 | { |
| 1932 | return true; |
| 1933 | } |
| 1934 | case FORMAT_A16W16V16U16: |
| 1935 | if(component < 3) |
| 1936 | { |
| 1937 | return false; |
| 1938 | } |
| 1939 | else |
| 1940 | { |
| 1941 | return true; |
| 1942 | } |
| 1943 | case FORMAT_Q8W8V8U8: |
| 1944 | case FORMAT_Q16W16V16U16: |
| 1945 | return false; |
| 1946 | case FORMAT_R32F: |
| 1947 | if(component < 1) |
| 1948 | { |
| 1949 | return false; |
| 1950 | } |
| 1951 | else |
| 1952 | { |
| 1953 | return true; |
| 1954 | } |
| 1955 | case FORMAT_G32R32F: |
| 1956 | if(component < 2) |
| 1957 | { |
| 1958 | return false; |
| 1959 | } |
| 1960 | else |
| 1961 | { |
| 1962 | return true; |
| 1963 | } |
| 1964 | case FORMAT_A32B32G32R32F: |
| 1965 | return false; |
| 1966 | default: |
| 1967 | ASSERT(false); |
| 1968 | } |
| 1969 | |
| 1970 | return false; |
| 1971 | } |
| 1972 | |
| 1973 | bool Surface::isSRGBreadable(Format format) |
| 1974 | { |
| 1975 | // Keep in sync with Capabilities::isSRGBreadable |
| 1976 | switch(format) |
| 1977 | { |
| 1978 | case FORMAT_L8: |
| 1979 | case FORMAT_A8L8: |
| 1980 | case FORMAT_R8G8B8: |
| 1981 | case FORMAT_A8R8G8B8: |
| 1982 | case FORMAT_X8R8G8B8: |
| 1983 | case FORMAT_A8B8G8R8: |
| 1984 | case FORMAT_X8B8G8R8: |
| 1985 | case FORMAT_R5G6B5: |
| 1986 | case FORMAT_X1R5G5B5: |
| 1987 | case FORMAT_A1R5G5B5: |
| 1988 | case FORMAT_A4R4G4B4: |
| 1989 | #if S3TC_SUPPORT |
| 1990 | case FORMAT_DXT1: |
| 1991 | case FORMAT_DXT3: |
| 1992 | case FORMAT_DXT5: |
John Bauman | 66b8ab2 | 2014-05-06 15:57:45 -0400 | [diff] [blame^] | 1993 | #endif |
John Bauman | 8940182 | 2014-05-06 15:04:28 -0400 | [diff] [blame] | 1994 | case FORMAT_ATI1: |
| 1995 | case FORMAT_ATI2: |
John Bauman | 8940182 | 2014-05-06 15:04:28 -0400 | [diff] [blame] | 1996 | return true; |
| 1997 | default: |
| 1998 | return false; |
| 1999 | } |
| 2000 | |
| 2001 | return false; |
| 2002 | } |
| 2003 | |
| 2004 | bool Surface::isSRGBwritable(Format format) |
| 2005 | { |
| 2006 | // Keep in sync with Capabilities::isSRGBwritable |
| 2007 | switch(format) |
| 2008 | { |
| 2009 | case FORMAT_NULL: |
| 2010 | case FORMAT_A8R8G8B8: |
| 2011 | case FORMAT_X8R8G8B8: |
| 2012 | case FORMAT_A8B8G8R8: |
| 2013 | case FORMAT_X8B8G8R8: |
| 2014 | case FORMAT_R5G6B5: |
| 2015 | return true; |
| 2016 | default: |
| 2017 | return false; |
| 2018 | } |
| 2019 | } |
| 2020 | |
| 2021 | bool Surface::isCompressed(Format format) |
| 2022 | { |
| 2023 | switch(format) |
| 2024 | { |
| 2025 | #if S3TC_SUPPORT |
| 2026 | case FORMAT_DXT1: |
| 2027 | case FORMAT_DXT3: |
| 2028 | case FORMAT_DXT5: |
John Bauman | 66b8ab2 | 2014-05-06 15:57:45 -0400 | [diff] [blame^] | 2029 | #endif |
John Bauman | 8940182 | 2014-05-06 15:04:28 -0400 | [diff] [blame] | 2030 | case FORMAT_ATI1: |
| 2031 | case FORMAT_ATI2: |
| 2032 | return true; |
John Bauman | 8940182 | 2014-05-06 15:04:28 -0400 | [diff] [blame] | 2033 | default: |
| 2034 | return false; |
| 2035 | } |
| 2036 | } |
| 2037 | |
| 2038 | int Surface::componentCount(Format format) |
| 2039 | { |
| 2040 | switch(format) |
| 2041 | { |
| 2042 | case FORMAT_X8R8G8B8: return 3; |
| 2043 | case FORMAT_A8R8G8B8: return 4; |
| 2044 | case FORMAT_G8R8: return 2; |
| 2045 | case FORMAT_G16R16: return 2; |
| 2046 | case FORMAT_A16B16G16R16: return 4; |
| 2047 | case FORMAT_V8U8: return 2; |
| 2048 | case FORMAT_Q8W8V8U8: return 4; |
| 2049 | case FORMAT_X8L8V8U8: return 3; |
| 2050 | case FORMAT_V16U16: return 2; |
| 2051 | case FORMAT_A16W16V16U16: return 4; |
| 2052 | case FORMAT_Q16W16V16U16: return 4; |
| 2053 | case FORMAT_R32F: return 1; |
| 2054 | case FORMAT_G32R32F: return 2; |
| 2055 | case FORMAT_A32B32G32R32F: return 4; |
| 2056 | case FORMAT_D32F_LOCKABLE: return 1; |
John Bauman | 66b8ab2 | 2014-05-06 15:57:45 -0400 | [diff] [blame^] | 2057 | case FORMAT_D32FS8_TEXTURE: return 1; |
| 2058 | case FORMAT_D32FS8_SHADOW: return 1; |
John Bauman | 8940182 | 2014-05-06 15:04:28 -0400 | [diff] [blame] | 2059 | case FORMAT_A8: return 1; |
| 2060 | case FORMAT_R8: return 1; |
| 2061 | case FORMAT_L8: return 1; |
| 2062 | case FORMAT_L16: return 1; |
| 2063 | case FORMAT_A8L8: return 2; |
| 2064 | default: |
| 2065 | ASSERT(false); |
| 2066 | } |
| 2067 | |
| 2068 | return 1; |
| 2069 | } |
| 2070 | |
| 2071 | void *Surface::allocateBuffer(int width, int height, int depth, Format format) |
| 2072 | { |
| 2073 | int width4 = (width + 3) & ~3; |
| 2074 | int height4 = (height + 3) & ~3; |
| 2075 | |
John Bauman | 19bac1e | 2014-05-06 15:23:49 -0400 | [diff] [blame] | 2076 | return allocate(size(width4, height4, depth, format)); |
John Bauman | 8940182 | 2014-05-06 15:04:28 -0400 | [diff] [blame] | 2077 | } |
| 2078 | |
| 2079 | void Surface::memfill(void *buffer, int pattern, int bytes) |
| 2080 | { |
| 2081 | while((size_t)buffer & 0x1 && bytes >= 1) |
| 2082 | { |
| 2083 | *(char*)buffer = (char)pattern; |
| 2084 | (char*&)buffer += 1; |
| 2085 | bytes -= 1; |
| 2086 | } |
| 2087 | |
| 2088 | while((size_t)buffer & 0x3 && bytes >= 2) |
| 2089 | { |
| 2090 | *(short*)buffer = (short)pattern; |
| 2091 | (short*&)buffer += 1; |
| 2092 | bytes -= 2; |
| 2093 | } |
| 2094 | |
| 2095 | if(CPUID::supportsSSE()) |
| 2096 | { |
| 2097 | while((size_t)buffer & 0xF && bytes >= 4) |
| 2098 | { |
| 2099 | *(int*)buffer = pattern; |
| 2100 | (int*&)buffer += 1; |
| 2101 | bytes -= 4; |
| 2102 | } |
| 2103 | |
| 2104 | __m128 quad = _mm_set_ps1((float&)pattern); |
| 2105 | |
| 2106 | float *pointer = (float*)buffer; |
| 2107 | int qxwords = bytes / 64; |
| 2108 | bytes -= qxwords * 64; |
| 2109 | |
| 2110 | while(qxwords--) |
| 2111 | { |
| 2112 | _mm_stream_ps(pointer + 0, quad); |
| 2113 | _mm_stream_ps(pointer + 4, quad); |
| 2114 | _mm_stream_ps(pointer + 8, quad); |
| 2115 | _mm_stream_ps(pointer + 12, quad); |
| 2116 | |
| 2117 | pointer += 16; |
| 2118 | } |
| 2119 | |
| 2120 | buffer = pointer; |
| 2121 | } |
| 2122 | |
| 2123 | while(bytes >= 4) |
| 2124 | { |
| 2125 | *(int*)buffer = (int)pattern; |
| 2126 | (int*&)buffer += 1; |
| 2127 | bytes -= 4; |
| 2128 | } |
| 2129 | |
| 2130 | while(bytes >= 2) |
| 2131 | { |
| 2132 | *(short*)buffer = (short)pattern; |
| 2133 | (short*&)buffer += 1; |
| 2134 | bytes -= 2; |
| 2135 | } |
| 2136 | |
| 2137 | while(bytes >= 1) |
| 2138 | { |
| 2139 | *(char*)buffer = (char)pattern; |
| 2140 | (char*&)buffer += 1; |
| 2141 | bytes -= 1; |
| 2142 | } |
| 2143 | } |
| 2144 | |
| 2145 | void Surface::clearColorBuffer(unsigned int color, unsigned int rgbaMask, int x0, int y0, int width, int height) |
| 2146 | { |
| 2147 | // FIXME: Also clear buffers in other formats? |
| 2148 | |
| 2149 | // Not overlapping |
| 2150 | if(x0 > internal.width) return; |
| 2151 | if(y0 > internal.height) return; |
| 2152 | if(x0 + width < 0) return; |
| 2153 | if(y0 + height < 0) return; |
| 2154 | |
| 2155 | // Clip against dimensions |
| 2156 | if(x0 < 0) {width += x0; x0 = 0;} |
| 2157 | if(x0 + width > internal.width) width = internal.width - x0; |
| 2158 | if(y0 < 0) {height += y0; y0 = 0;} |
| 2159 | if(y0 + height > internal.height) height = internal.height - y0; |
| 2160 | |
| 2161 | const bool entire = x0 == 0 && y0 == 0 && width == internal.width && height == internal.height; |
| 2162 | const Lock lock = entire ? LOCK_DISCARD : LOCK_WRITEONLY; |
| 2163 | |
| 2164 | int width2 = (internal.width + 1) & ~1; |
| 2165 | |
| 2166 | int x1 = x0 + width; |
| 2167 | int y1 = y0 + height; |
| 2168 | |
| 2169 | int bytes = 4 * (x1 - x0); |
| 2170 | |
| 2171 | // if(lockable || !quadLayoutEnabled) |
| 2172 | { |
| 2173 | unsigned char *buffer = (unsigned char*)lockInternal(x0, y0, 0, lock, PUBLIC); |
| 2174 | |
| 2175 | unsigned char r8 = (color & 0x00FF0000) >> 16; |
| 2176 | unsigned char g8 = (color & 0x0000FF00) >> 8; |
| 2177 | unsigned char b8 = (color & 0x000000FF) >> 0; |
| 2178 | unsigned char a8 = (color & 0xFF000000) >> 24; |
| 2179 | |
| 2180 | unsigned short r16 = (r8 << 8) + r8; |
| 2181 | unsigned short g16 = (g8 << 8) + g8; |
| 2182 | unsigned short b16 = (b8 << 8) + b8; |
| 2183 | unsigned short a16 = (a8 << 8) + a8; |
| 2184 | |
| 2185 | float r32f = r8 / 255.0f; |
| 2186 | float g32f = g8 / 255.0f; |
| 2187 | float b32f = b8 / 255.0f; |
| 2188 | float a32f = a8 / 255.0f; |
| 2189 | |
| 2190 | unsigned char g8r8[4] = {r8, g8, r8, g8}; |
| 2191 | unsigned short g16r16[2] = {r16, g16}; |
| 2192 | |
| 2193 | for(int z = 0; z < internal.depth; z++) |
| 2194 | { |
| 2195 | unsigned char *target = buffer; |
| 2196 | |
| 2197 | for(int y = y0; y < y1; y++) |
| 2198 | { |
| 2199 | switch(internal.format) |
| 2200 | { |
| 2201 | case FORMAT_NULL: |
| 2202 | break; |
| 2203 | case FORMAT_X8R8G8B8: |
| 2204 | case FORMAT_A8R8G8B8: |
| 2205 | // case FORMAT_X8G8R8B8Q: // FIXME |
| 2206 | // case FORMAT_A8G8R8B8Q: // FIXME |
John Bauman | 19bac1e | 2014-05-06 15:23:49 -0400 | [diff] [blame] | 2207 | if(rgbaMask == 0xF || (internal.format == FORMAT_X8R8G8B8 && rgbaMask == 0x7)) |
John Bauman | 8940182 | 2014-05-06 15:04:28 -0400 | [diff] [blame] | 2208 | { |
| 2209 | memfill(target, color, 4 * (x1 - x0)); |
| 2210 | } |
| 2211 | else |
| 2212 | { |
| 2213 | unsigned int bgraMask = (rgbaMask & 0x1 ? 0x00FF0000 : 0) | (rgbaMask & 0x2 ? 0x0000FF00 : 0) | (rgbaMask & 0x4 ? 0x000000FF : 0) | (rgbaMask & 0x8 ? 0xFF000000 : 0); |
| 2214 | unsigned int invMask = ~bgraMask; |
| 2215 | unsigned int maskedColor = color & bgraMask; |
| 2216 | unsigned int *target32 = (unsigned int*)target; |
| 2217 | |
| 2218 | for(int x = 0; x < width; x++) |
| 2219 | { |
| 2220 | target32[x] = maskedColor | (target32[x] & invMask); |
| 2221 | } |
| 2222 | } |
| 2223 | break; |
| 2224 | case FORMAT_G8R8: |
| 2225 | if((rgbaMask & 0x3) == 0x3) |
| 2226 | { |
| 2227 | memfill(target, (int&)g8r8, 2 * (x1 - x0)); |
| 2228 | } |
| 2229 | else |
| 2230 | { |
| 2231 | unsigned short rgMask = (rgbaMask & 0x1 ? 0x000000FF : 0) | (rgbaMask & 0x2 ? 0x0000FF00 : 0); |
| 2232 | unsigned short invMask = ~rgMask; |
| 2233 | unsigned short maskedColor = (unsigned short&)g8r8 & rgMask; |
| 2234 | unsigned short *target16 = (unsigned short*)target; |
| 2235 | |
| 2236 | for(int x = 0; x < width; x++) |
| 2237 | { |
| 2238 | target16[x] = maskedColor | (target16[x] & invMask); |
| 2239 | } |
| 2240 | } |
| 2241 | break; |
| 2242 | case FORMAT_G16R16: |
| 2243 | if((rgbaMask & 0x3) == 0x3) |
| 2244 | { |
| 2245 | memfill(target, (int&)g16r16, 4 * (x1 - x0)); |
| 2246 | } |
| 2247 | else |
| 2248 | { |
| 2249 | unsigned int rgMask = (rgbaMask & 0x1 ? 0x0000FFFF : 0) | (rgbaMask & 0x2 ? 0xFFFF0000 : 0); |
| 2250 | unsigned int invMask = ~rgMask; |
| 2251 | unsigned int maskedColor = (unsigned int&)g16r16 & rgMask; |
| 2252 | unsigned int *target32 = (unsigned int*)target; |
| 2253 | |
| 2254 | for(int x = 0; x < width; x++) |
| 2255 | { |
| 2256 | target32[x] = maskedColor | (target32[x] & invMask); |
| 2257 | } |
| 2258 | } |
| 2259 | break; |
| 2260 | case FORMAT_A16B16G16R16: |
| 2261 | if(rgbaMask == 0xF) |
| 2262 | { |
| 2263 | for(int x = 0; x < width; x++) |
| 2264 | { |
| 2265 | ((unsigned short*)target)[4 * x + 0] = r16; |
| 2266 | ((unsigned short*)target)[4 * x + 1] = g16; |
| 2267 | ((unsigned short*)target)[4 * x + 2] = b16; |
| 2268 | ((unsigned short*)target)[4 * x + 3] = a16; |
| 2269 | } |
| 2270 | } |
| 2271 | else |
| 2272 | { |
| 2273 | if(rgbaMask & 0x1) for(int x = 0; x < width; x++) ((unsigned short*)target)[4 * x + 0] = r16; |
| 2274 | if(rgbaMask & 0x2) for(int x = 0; x < width; x++) ((unsigned short*)target)[4 * x + 1] = g16; |
| 2275 | if(rgbaMask & 0x4) for(int x = 0; x < width; x++) ((unsigned short*)target)[4 * x + 2] = b16; |
| 2276 | if(rgbaMask & 0x8) for(int x = 0; x < width; x++) ((unsigned short*)target)[4 * x + 3] = a16; |
| 2277 | } |
| 2278 | break; |
| 2279 | case FORMAT_R32F: |
| 2280 | if(rgbaMask & 0x1) |
| 2281 | { |
| 2282 | for(int x = 0; x < width; x++) |
| 2283 | { |
| 2284 | ((float*)target)[x] = r32f; |
| 2285 | } |
| 2286 | } |
| 2287 | break; |
| 2288 | case FORMAT_G32R32F: |
| 2289 | if((rgbaMask & 0x3) == 0x3) |
| 2290 | { |
| 2291 | for(int x = 0; x < width; x++) |
| 2292 | { |
| 2293 | ((float*)target)[2 * x + 0] = r32f; |
| 2294 | ((float*)target)[2 * x + 1] = g32f; |
| 2295 | } |
| 2296 | } |
| 2297 | else |
| 2298 | { |
| 2299 | if(rgbaMask & 0x1) for(int x = 0; x < width; x++) ((float*)target)[2 * x + 0] = r32f; |
| 2300 | if(rgbaMask & 0x2) for(int x = 0; x < width; x++) ((float*)target)[2 * x + 1] = g32f; |
| 2301 | } |
| 2302 | break; |
| 2303 | case FORMAT_A32B32G32R32F: |
| 2304 | if(rgbaMask == 0xF) |
| 2305 | { |
| 2306 | for(int x = 0; x < width; x++) |
| 2307 | { |
| 2308 | ((float*)target)[4 * x + 0] = r32f; |
| 2309 | ((float*)target)[4 * x + 1] = g32f; |
| 2310 | ((float*)target)[4 * x + 2] = b32f; |
| 2311 | ((float*)target)[4 * x + 3] = a32f; |
| 2312 | } |
| 2313 | } |
| 2314 | else |
| 2315 | { |
| 2316 | if(rgbaMask & 0x1) for(int x = 0; x < width; x++) ((float*)target)[4 * x + 0] = r32f; |
| 2317 | if(rgbaMask & 0x2) for(int x = 0; x < width; x++) ((float*)target)[4 * x + 1] = g32f; |
| 2318 | if(rgbaMask & 0x4) for(int x = 0; x < width; x++) ((float*)target)[4 * x + 2] = b32f; |
| 2319 | if(rgbaMask & 0x8) for(int x = 0; x < width; x++) ((float*)target)[4 * x + 3] = a32f; |
| 2320 | } |
| 2321 | break; |
| 2322 | default: |
| 2323 | ASSERT(false); |
| 2324 | } |
| 2325 | |
| 2326 | target += internal.pitchB; |
| 2327 | } |
| 2328 | |
| 2329 | buffer += internal.sliceB; |
| 2330 | } |
| 2331 | |
| 2332 | unlockInternal(); |
| 2333 | } |
| 2334 | /* else |
| 2335 | { |
| 2336 | // unsigned char *target = (unsigned char*&)buffer; |
| 2337 | // |
| 2338 | // for(int y = y0; y < y1; y++) |
| 2339 | // { |
| 2340 | // for(int x = x0; x < x1; x++) |
| 2341 | // { |
| 2342 | // target[width2 * 4 * (y & ~1) + 2 * (y & 1) + 8 * (x & ~1) + (x & 1) + 0] = (color & 0x000000FF) >> 0; |
| 2343 | // target[width2 * 4 * (y & ~1) + 2 * (y & 1) + 8 * (x & ~1) + (x & 1) + 4] = (color & 0x00FF0000) >> 16; |
| 2344 | // target[width2 * 4 * (y & ~1) + 2 * (y & 1) + 8 * (x & ~1) + (x & 1) + 8] = (color & 0x0000FF00) >> 8; |
| 2345 | // target[width2 * 4 * (y & ~1) + 2 * (y & 1) + 8 * (x & ~1) + (x & 1) + 12] = (color & 0xFF000000) >> 24; |
| 2346 | // } |
| 2347 | // } |
| 2348 | |
| 2349 | unsigned char colorQ[16]; |
| 2350 | |
| 2351 | colorQ[0] = (color & 0x000000FF) >> 0; |
| 2352 | colorQ[1] = (color & 0x000000FF) >> 0; |
| 2353 | colorQ[2] = (color & 0x000000FF) >> 0; |
| 2354 | colorQ[3] = (color & 0x000000FF) >> 0; |
| 2355 | colorQ[4] = (color & 0x00FF0000) >> 16; |
| 2356 | colorQ[5] = (color & 0x00FF0000) >> 16; |
| 2357 | colorQ[6] = (color & 0x00FF0000) >> 16; |
| 2358 | colorQ[7] = (color & 0x00FF0000) >> 16; |
| 2359 | colorQ[8] = (color & 0x0000FF00) >> 8; |
| 2360 | colorQ[9] = (color & 0x0000FF00) >> 8; |
| 2361 | colorQ[10] = (color & 0x0000FF00) >> 8; |
| 2362 | colorQ[11] = (color & 0x0000FF00) >> 8; |
| 2363 | colorQ[12] = (color & 0xFF000000) >> 24; |
| 2364 | colorQ[13] = (color & 0xFF000000) >> 24; |
| 2365 | colorQ[14] = (color & 0xFF000000) >> 24; |
| 2366 | colorQ[15] = (color & 0xFF000000) >> 24; |
| 2367 | |
| 2368 | for(int y = y0; y < y1; y++) |
| 2369 | { |
| 2370 | unsigned char *target = (unsigned char*)lockInternal(0, 0, 0, lock) + width2 * 4 * (y & ~1) + 2 * (y & 1); // FIXME: Unlock |
| 2371 | |
| 2372 | if((y & 1) == 0 && y + 1 < y1) // Fill quad line at once |
| 2373 | { |
| 2374 | if((x0 & 1) != 0) |
| 2375 | { |
| 2376 | target[8 * (x0 & ~1) + 1 + 0] = (color & 0x000000FF) >> 0; |
| 2377 | target[8 * (x0 & ~1) + 1 + 4] = (color & 0x00FF0000) >> 16; |
| 2378 | target[8 * (x0 & ~1) + 1 + 8] = (color & 0x0000FF00) >> 8; |
| 2379 | target[8 * (x0 & ~1) + 1 + 12] = (color & 0xFF000000) >> 24; |
| 2380 | |
| 2381 | target[8 * (x0 & ~1) + 3 + 0] = (color & 0x000000FF) >> 0; |
| 2382 | target[8 * (x0 & ~1) + 3 + 4] = (color & 0x00FF0000) >> 16; |
| 2383 | target[8 * (x0 & ~1) + 3 + 8] = (color & 0x0000FF00) >> 8; |
| 2384 | target[8 * (x0 & ~1) + 3 + 12] = (color & 0xFF000000) >> 24; |
| 2385 | } |
| 2386 | |
| 2387 | __asm |
| 2388 | { |
| 2389 | movq mm0, colorQ+0 |
| 2390 | movq mm1, colorQ+8 |
| 2391 | |
| 2392 | mov eax, x0 |
| 2393 | add eax, 1 |
| 2394 | and eax, 0xFFFFFFFE |
| 2395 | cmp eax, x1 |
| 2396 | jge qEnd |
| 2397 | |
| 2398 | mov edi, target |
| 2399 | |
| 2400 | qLoop: |
| 2401 | movntq [edi+8*eax+0], mm0 |
| 2402 | movntq [edi+8*eax+8], mm1 |
| 2403 | |
| 2404 | add eax, 2 |
| 2405 | cmp eax, x1 |
| 2406 | jl qLoop |
| 2407 | qEnd: |
| 2408 | emms |
| 2409 | } |
| 2410 | |
| 2411 | if((x1 & 1) != 0) |
| 2412 | { |
| 2413 | target[8 * (x1 & ~1) + 0 + 0] = (color & 0x000000FF) >> 0; |
| 2414 | target[8 * (x1 & ~1) + 0 + 4] = (color & 0x00FF0000) >> 16; |
| 2415 | target[8 * (x1 & ~1) + 0 + 8] = (color & 0x0000FF00) >> 8; |
| 2416 | target[8 * (x1 & ~1) + 0 + 12] = (color & 0xFF000000) >> 24; |
| 2417 | |
| 2418 | target[8 * (x1 & ~1) + 2 + 0] = (color & 0x000000FF) >> 0; |
| 2419 | target[8 * (x1 & ~1) + 2 + 4] = (color & 0x00FF0000) >> 16; |
| 2420 | target[8 * (x1 & ~1) + 2 + 8] = (color & 0x0000FF00) >> 8; |
| 2421 | target[8 * (x1 & ~1) + 2 + 12] = (color & 0xFF000000) >> 24; |
| 2422 | } |
| 2423 | |
| 2424 | y++; |
| 2425 | } |
| 2426 | else |
| 2427 | { |
| 2428 | for(int x = x0; x < x1; x++) |
| 2429 | { |
| 2430 | target[8 * (x & ~1) + (x & 1) + 0] = (color & 0x000000FF) >> 0; |
| 2431 | target[8 * (x & ~1) + (x & 1) + 4] = (color & 0x00FF0000) >> 16; |
| 2432 | target[8 * (x & ~1) + (x & 1) + 8] = (color & 0x0000FF00) >> 8; |
| 2433 | target[8 * (x & ~1) + (x & 1) + 12] = (color & 0xFF000000) >> 24; |
| 2434 | } |
| 2435 | } |
| 2436 | } |
| 2437 | }*/ |
| 2438 | } |
| 2439 | |
| 2440 | void Surface::clearDepthBuffer(float depth, int x0, int y0, int width, int height) |
| 2441 | { |
| 2442 | // Not overlapping |
| 2443 | if(x0 > internal.width) return; |
| 2444 | if(y0 > internal.height) return; |
| 2445 | if(x0 + width < 0) return; |
| 2446 | if(y0 + height < 0) return; |
| 2447 | |
| 2448 | // Clip against dimensions |
| 2449 | if(x0 < 0) {width += x0; x0 = 0;} |
| 2450 | if(x0 + width > internal.width) width = internal.width - x0; |
| 2451 | if(y0 < 0) {height += y0; y0 = 0;} |
| 2452 | if(y0 + height > internal.height) height = internal.height - y0; |
| 2453 | |
| 2454 | const bool entire = x0 == 0 && y0 == 0 && width == internal.width && height == internal.height; |
| 2455 | const Lock lock = entire ? LOCK_DISCARD : LOCK_WRITEONLY; |
| 2456 | |
| 2457 | int width2 = (internal.width + 1) & ~1; |
| 2458 | |
| 2459 | int x1 = x0 + width; |
| 2460 | int y1 = y0 + height; |
| 2461 | |
| 2462 | if(internal.format == FORMAT_D32F_LOCKABLE || |
John Bauman | 66b8ab2 | 2014-05-06 15:57:45 -0400 | [diff] [blame^] | 2463 | internal.format == FORMAT_D32FS8_TEXTURE || |
| 2464 | internal.format == FORMAT_D32FS8_SHADOW) |
John Bauman | 8940182 | 2014-05-06 15:04:28 -0400 | [diff] [blame] | 2465 | { |
| 2466 | float *target = (float*)lockInternal(0, 0, 0, lock, PUBLIC) + x0 + width2 * y0; |
| 2467 | |
| 2468 | for(int z = 0; z < internal.depth; z++) |
| 2469 | { |
| 2470 | for(int y = y0; y < y1; y++) |
| 2471 | { |
| 2472 | memfill(target, (int&)depth, 4 * width); |
| 2473 | target += width2; |
| 2474 | } |
| 2475 | } |
| 2476 | |
| 2477 | unlockInternal(); |
| 2478 | } |
| 2479 | else // Quad layout |
| 2480 | { |
| 2481 | if(complementaryDepthBuffer) |
| 2482 | { |
| 2483 | depth = 1 - depth; |
| 2484 | } |
| 2485 | |
| 2486 | float *buffer = (float*)lockInternal(0, 0, 0, lock, PUBLIC); |
| 2487 | |
| 2488 | for(int z = 0; z < internal.depth; z++) |
| 2489 | { |
| 2490 | for(int y = y0; y < y1; y++) |
| 2491 | { |
| 2492 | float *target = buffer + (y & ~1) * width2 + (y & 1) * 2; |
| 2493 | |
| 2494 | if((y & 1) == 0 && y + 1 < y1) // Fill quad line at once |
| 2495 | { |
| 2496 | if((x0 & 1) != 0) |
| 2497 | { |
| 2498 | target[(x0 & ~1) * 2 + 1] = depth; |
| 2499 | target[(x0 & ~1) * 2 + 3] = depth; |
| 2500 | } |
| 2501 | |
| 2502 | // for(int x2 = ((x0 + 1) & ~1) * 2; x2 < x1 * 2; x2 += 4) |
| 2503 | // { |
| 2504 | // target[x2 + 0] = depth; |
| 2505 | // target[x2 + 1] = depth; |
| 2506 | // target[x2 + 2] = depth; |
| 2507 | // target[x2 + 3] = depth; |
| 2508 | // } |
| 2509 | |
| 2510 | // __asm |
| 2511 | // { |
| 2512 | // movss xmm0, depth |
| 2513 | // shufps xmm0, xmm0, 0x00 |
| 2514 | // |
| 2515 | // mov eax, x0 |
| 2516 | // add eax, 1 |
| 2517 | // and eax, 0xFFFFFFFE |
| 2518 | // cmp eax, x1 |
| 2519 | // jge qEnd |
| 2520 | // |
| 2521 | // mov edi, target |
| 2522 | // |
| 2523 | // qLoop: |
| 2524 | // movntps [edi+8*eax], xmm0 |
| 2525 | // |
| 2526 | // add eax, 2 |
| 2527 | // cmp eax, x1 |
| 2528 | // jl qLoop |
| 2529 | // qEnd: |
| 2530 | // } |
| 2531 | |
| 2532 | memfill(&target[((x0 + 1) & ~1) * 2], (int&)depth, 8 * ((x1 & ~1) - ((x0 + 1) & ~1))); |
| 2533 | |
| 2534 | if((x1 & 1) != 0) |
| 2535 | { |
| 2536 | target[(x1 & ~1) * 2 + 0] = depth; |
| 2537 | target[(x1 & ~1) * 2 + 2] = depth; |
| 2538 | } |
| 2539 | |
| 2540 | y++; |
| 2541 | } |
| 2542 | else |
| 2543 | { |
| 2544 | for(int x = x0; x < x1; x++) |
| 2545 | { |
| 2546 | target[(x & ~1) * 2 + (x & 1)] = depth; |
| 2547 | } |
| 2548 | } |
| 2549 | } |
| 2550 | |
| 2551 | buffer += internal.sliceP; |
| 2552 | } |
| 2553 | |
| 2554 | unlockInternal(); |
| 2555 | } |
| 2556 | } |
| 2557 | |
| 2558 | void Surface::clearStencilBuffer(unsigned char s, unsigned char mask, int x0, int y0, int width, int height) |
| 2559 | { |
| 2560 | // Not overlapping |
| 2561 | if(x0 > internal.width) return; |
| 2562 | if(y0 > internal.height) return; |
| 2563 | if(x0 + width < 0) return; |
| 2564 | if(y0 + height < 0) return; |
| 2565 | |
| 2566 | // Clip against dimensions |
| 2567 | if(x0 < 0) {width += x0; x0 = 0;} |
| 2568 | if(x0 + width > internal.width) width = internal.width - x0; |
| 2569 | if(y0 < 0) {height += y0; y0 = 0;} |
| 2570 | if(y0 + height > internal.height) height = internal.height - y0; |
| 2571 | |
| 2572 | int width2 = (internal.width + 1) & ~1; |
| 2573 | |
| 2574 | int x1 = x0 + width; |
| 2575 | int y1 = y0 + height; |
| 2576 | |
| 2577 | unsigned char maskedS = s & mask; |
| 2578 | unsigned char invMask = ~mask; |
| 2579 | unsigned int fill = maskedS; |
| 2580 | fill = fill | (fill << 8) | (fill << 16) + (fill << 24); |
| 2581 | |
| 2582 | if(false) |
| 2583 | { |
| 2584 | char *target = (char*)lockStencil(0, PUBLIC) + x0 + width2 * y0; |
| 2585 | |
| 2586 | for(int z = 0; z < stencil.depth; z++) |
| 2587 | { |
| 2588 | for(int y = y0; y < y0 + height; y++) |
| 2589 | { |
| 2590 | if(mask == 0xFF) |
| 2591 | { |
| 2592 | memfill(target, fill, width); |
| 2593 | } |
| 2594 | else |
| 2595 | { |
| 2596 | for(int x = 0; x < width; x++) |
| 2597 | { |
| 2598 | target[x] = maskedS | (target[x] & invMask); |
| 2599 | } |
| 2600 | } |
| 2601 | |
| 2602 | target += width2; |
| 2603 | } |
| 2604 | } |
| 2605 | |
| 2606 | unlockStencil(); |
| 2607 | } |
| 2608 | else // Quad layout |
| 2609 | { |
| 2610 | char *buffer = (char*)lockStencil(0, PUBLIC); |
| 2611 | |
| 2612 | if(mask == 0xFF) |
| 2613 | { |
| 2614 | for(int z = 0; z < stencil.depth; z++) |
| 2615 | { |
| 2616 | for(int y = y0; y < y1; y++) |
| 2617 | { |
| 2618 | char *target = buffer + (y & ~1) * width2 + (y & 1) * 2; |
| 2619 | |
| 2620 | if((y & 1) == 0 && y + 1 < y1 && mask == 0xFF) // Fill quad line at once |
| 2621 | { |
| 2622 | if((x0 & 1) != 0) |
| 2623 | { |
| 2624 | target[(x0 & ~1) * 2 + 1] = fill; |
| 2625 | target[(x0 & ~1) * 2 + 3] = fill; |
| 2626 | } |
| 2627 | |
| 2628 | memfill(&target[((x0 + 1) & ~1) * 2], fill, ((x1 + 1) & ~1) * 2 - ((x0 + 1) & ~1) * 2); |
| 2629 | |
| 2630 | if((x1 & 1) != 0) |
| 2631 | { |
| 2632 | target[(x1 & ~1) * 2 + 0] = fill; |
| 2633 | target[(x1 & ~1) * 2 + 2] = fill; |
| 2634 | } |
| 2635 | |
| 2636 | y++; |
| 2637 | } |
| 2638 | else |
| 2639 | { |
| 2640 | for(int x = x0; x < x1; x++) |
| 2641 | { |
| 2642 | target[(x & ~1) * 2 + (x & 1)] = maskedS | (target[x] & invMask); |
| 2643 | } |
| 2644 | } |
| 2645 | } |
| 2646 | |
| 2647 | buffer += stencil.sliceP; |
| 2648 | } |
| 2649 | } |
| 2650 | |
| 2651 | unlockStencil(); |
| 2652 | } |
| 2653 | } |
| 2654 | |
| 2655 | void Surface::fill(const Color<float> &color, int x0, int y0, int width, int height) |
| 2656 | { |
| 2657 | unsigned char *row; |
| 2658 | Buffer *buffer; |
| 2659 | |
| 2660 | if(internal.dirty) |
| 2661 | { |
| 2662 | row = (unsigned char*)lockInternal(x0, y0, 0, LOCK_WRITEONLY, PUBLIC); |
| 2663 | buffer = &internal; |
| 2664 | } |
| 2665 | else |
| 2666 | { |
| 2667 | row = (unsigned char*)lockExternal(x0, y0, 0, LOCK_WRITEONLY, PUBLIC); |
| 2668 | buffer = &external; |
| 2669 | } |
| 2670 | |
| 2671 | if(buffer->bytes <= 4) |
| 2672 | { |
| 2673 | int c; |
| 2674 | buffer->write(&c, color); |
| 2675 | |
| 2676 | if(buffer->bytes <= 1) c = (c << 8) | c; |
| 2677 | if(buffer->bytes <= 2) c = (c << 16) | c; |
| 2678 | |
| 2679 | for(int y = 0; y < height; y++) |
| 2680 | { |
| 2681 | memfill(row, c, width * buffer->bytes); |
| 2682 | |
| 2683 | row += buffer->pitchB; |
| 2684 | } |
| 2685 | } |
| 2686 | else // Generic |
| 2687 | { |
| 2688 | for(int y = 0; y < height; y++) |
| 2689 | { |
| 2690 | unsigned char *element = row; |
| 2691 | |
| 2692 | for(int x = 0; x < width; x++) |
| 2693 | { |
| 2694 | buffer->write(element, color); |
| 2695 | |
| 2696 | element += buffer->bytes; |
| 2697 | } |
| 2698 | |
| 2699 | row += buffer->pitchB; |
| 2700 | } |
| 2701 | } |
| 2702 | |
| 2703 | if(buffer == &internal) |
| 2704 | { |
| 2705 | unlockInternal(); |
| 2706 | } |
| 2707 | else |
| 2708 | { |
| 2709 | unlockExternal(); |
| 2710 | } |
| 2711 | } |
| 2712 | |
| 2713 | Color<float> Surface::readExternal(int x, int y, int z) const |
| 2714 | { |
| 2715 | ASSERT(external.lock != LOCK_UNLOCKED); |
| 2716 | |
| 2717 | return external.read(x, y, z); |
| 2718 | } |
| 2719 | |
| 2720 | Color<float> Surface::readExternal(int x, int y) const |
| 2721 | { |
| 2722 | ASSERT(external.lock != LOCK_UNLOCKED); |
| 2723 | |
| 2724 | return external.read(x, y); |
| 2725 | } |
| 2726 | |
| 2727 | Color<float> Surface::sampleExternal(float x, float y, float z) const |
| 2728 | { |
| 2729 | ASSERT(external.lock != LOCK_UNLOCKED); |
| 2730 | |
| 2731 | return external.sample(x, y, z); |
| 2732 | } |
| 2733 | |
| 2734 | Color<float> Surface::sampleExternal(float x, float y) const |
| 2735 | { |
| 2736 | ASSERT(external.lock != LOCK_UNLOCKED); |
| 2737 | |
| 2738 | return external.sample(x, y); |
| 2739 | } |
| 2740 | |
| 2741 | void Surface::writeExternal(int x, int y, int z, const Color<float> &color) |
| 2742 | { |
| 2743 | ASSERT(external.lock != LOCK_UNLOCKED); |
| 2744 | |
| 2745 | external.write(x, y, z, color); |
| 2746 | } |
| 2747 | |
| 2748 | void Surface::writeExternal(int x, int y, const Color<float> &color) |
| 2749 | { |
| 2750 | ASSERT(external.lock != LOCK_UNLOCKED); |
| 2751 | |
| 2752 | external.write(x, y, color); |
| 2753 | } |
| 2754 | |
| 2755 | Color<float> Surface::readInternal(int x, int y, int z) const |
| 2756 | { |
| 2757 | ASSERT(internal.lock != LOCK_UNLOCKED); |
| 2758 | |
| 2759 | return internal.read(x, y, z); |
| 2760 | } |
| 2761 | |
| 2762 | Color<float> Surface::readInternal(int x, int y) const |
| 2763 | { |
| 2764 | ASSERT(internal.lock != LOCK_UNLOCKED); |
| 2765 | |
| 2766 | return internal.read(x, y); |
| 2767 | } |
| 2768 | |
| 2769 | Color<float> Surface::sampleInternal(float x, float y, float z) const |
| 2770 | { |
| 2771 | ASSERT(internal.lock != LOCK_UNLOCKED); |
| 2772 | |
| 2773 | return internal.sample(x, y, z); |
| 2774 | } |
| 2775 | |
| 2776 | Color<float> Surface::sampleInternal(float x, float y) const |
| 2777 | { |
| 2778 | ASSERT(internal.lock != LOCK_UNLOCKED); |
| 2779 | |
| 2780 | return internal.sample(x, y); |
| 2781 | } |
| 2782 | |
| 2783 | void Surface::writeInternal(int x, int y, int z, const Color<float> &color) |
| 2784 | { |
| 2785 | ASSERT(internal.lock != LOCK_UNLOCKED); |
| 2786 | |
| 2787 | internal.write(x, y, z, color); |
| 2788 | } |
| 2789 | |
| 2790 | void Surface::writeInternal(int x, int y, const Color<float> &color) |
| 2791 | { |
| 2792 | ASSERT(internal.lock != LOCK_UNLOCKED); |
| 2793 | |
| 2794 | internal.write(x, y, color); |
| 2795 | } |
| 2796 | |
| 2797 | bool Surface::hasStencil() const |
| 2798 | { |
| 2799 | return isStencil(external.format); |
| 2800 | } |
| 2801 | |
| 2802 | bool Surface::hasDepth() const |
| 2803 | { |
| 2804 | return isDepth(external.format); |
| 2805 | } |
| 2806 | |
| 2807 | bool Surface::hasPalette() const |
| 2808 | { |
| 2809 | return isPalette(external.format); |
| 2810 | } |
| 2811 | |
| 2812 | bool Surface::isRenderTarget() const |
| 2813 | { |
| 2814 | return renderTarget; |
| 2815 | } |
| 2816 | |
| 2817 | bool Surface::hasDirtyMipmaps() const |
| 2818 | { |
| 2819 | return dirtyMipmaps; |
| 2820 | } |
| 2821 | |
| 2822 | void Surface::cleanMipmaps() |
| 2823 | { |
| 2824 | dirtyMipmaps = false; |
| 2825 | } |
| 2826 | |
| 2827 | Resource *Surface::getResource() |
| 2828 | { |
| 2829 | return resource; |
| 2830 | } |
| 2831 | |
| 2832 | bool Surface::identicalFormats() const |
| 2833 | { |
John Bauman | 66b8ab2 | 2014-05-06 15:57:45 -0400 | [diff] [blame^] | 2834 | return external.format == internal.format && |
| 2835 | external.width == internal.width && |
| 2836 | external.height == internal.height && |
| 2837 | external.depth == internal.depth && |
| 2838 | external.pitchB == internal.pitchB && |
| 2839 | external.sliceB == internal.sliceB; |
John Bauman | 8940182 | 2014-05-06 15:04:28 -0400 | [diff] [blame] | 2840 | } |
| 2841 | |
| 2842 | Format Surface::selectInternalFormat(Format format) const |
| 2843 | { |
| 2844 | switch(format) |
| 2845 | { |
| 2846 | case FORMAT_NULL: |
| 2847 | return FORMAT_NULL; |
| 2848 | case FORMAT_P8: |
| 2849 | case FORMAT_A8P8: |
| 2850 | case FORMAT_A4R4G4B4: |
| 2851 | case FORMAT_A1R5G5B5: |
| 2852 | case FORMAT_A8R3G3B2: |
| 2853 | return FORMAT_A8R8G8B8; |
| 2854 | case FORMAT_A8: |
| 2855 | return FORMAT_A8; |
| 2856 | case FORMAT_R8: |
| 2857 | return FORMAT_R8; |
| 2858 | case FORMAT_A2R10G10B10: |
| 2859 | case FORMAT_A2B10G10R10: |
| 2860 | case FORMAT_A16B16G16R16: |
| 2861 | return FORMAT_A16B16G16R16; |
| 2862 | case FORMAT_G8R8: |
| 2863 | return FORMAT_G8R8; |
| 2864 | case FORMAT_G16R16: |
| 2865 | return FORMAT_G16R16; |
| 2866 | case FORMAT_A8R8G8B8: |
| 2867 | case FORMAT_A8B8G8R8: |
| 2868 | if(lockable || !quadLayoutEnabled) |
| 2869 | { |
| 2870 | return FORMAT_A8R8G8B8; |
| 2871 | } |
| 2872 | else |
| 2873 | { |
| 2874 | return FORMAT_A8G8R8B8Q; |
| 2875 | } |
| 2876 | case FORMAT_R3G3B2: |
| 2877 | case FORMAT_R5G6B5: |
| 2878 | case FORMAT_R8G8B8: |
| 2879 | case FORMAT_X4R4G4B4: |
| 2880 | case FORMAT_X1R5G5B5: |
| 2881 | case FORMAT_X8R8G8B8: |
| 2882 | case FORMAT_X8B8G8R8: |
| 2883 | if(lockable || !quadLayoutEnabled) |
| 2884 | { |
| 2885 | return FORMAT_X8R8G8B8; |
| 2886 | } |
| 2887 | else |
| 2888 | { |
| 2889 | return FORMAT_X8G8R8B8Q; |
| 2890 | } |
| 2891 | // Compressed formats |
| 2892 | #if S3TC_SUPPORT |
| 2893 | case FORMAT_DXT1: |
| 2894 | case FORMAT_DXT3: |
| 2895 | case FORMAT_DXT5: |
| 2896 | return FORMAT_A8R8G8B8; |
John Bauman | 66b8ab2 | 2014-05-06 15:57:45 -0400 | [diff] [blame^] | 2897 | #endif |
John Bauman | 8940182 | 2014-05-06 15:04:28 -0400 | [diff] [blame] | 2898 | case FORMAT_ATI1: |
| 2899 | return FORMAT_R8; |
| 2900 | case FORMAT_ATI2: |
| 2901 | return FORMAT_G8R8; |
John Bauman | 8940182 | 2014-05-06 15:04:28 -0400 | [diff] [blame] | 2902 | // Bumpmap formats |
| 2903 | case FORMAT_V8U8: return FORMAT_V8U8; |
| 2904 | case FORMAT_L6V5U5: return FORMAT_X8L8V8U8; |
| 2905 | case FORMAT_Q8W8V8U8: return FORMAT_Q8W8V8U8; |
| 2906 | case FORMAT_X8L8V8U8: return FORMAT_X8L8V8U8; |
| 2907 | case FORMAT_V16U16: return FORMAT_V16U16; |
| 2908 | case FORMAT_A2W10V10U10: return FORMAT_A16W16V16U16; |
| 2909 | case FORMAT_Q16W16V16U16: return FORMAT_Q16W16V16U16; |
| 2910 | // Floating-point formats |
| 2911 | case FORMAT_R16F: return FORMAT_R32F; |
| 2912 | case FORMAT_G16R16F: return FORMAT_G32R32F; |
| 2913 | case FORMAT_A16B16G16R16F: return FORMAT_A32B32G32R32F; |
| 2914 | case FORMAT_R32F: return FORMAT_R32F; |
| 2915 | case FORMAT_G32R32F: return FORMAT_G32R32F; |
| 2916 | case FORMAT_A32B32G32R32F: return FORMAT_A32B32G32R32F; |
| 2917 | // Luminance formats |
| 2918 | case FORMAT_L8: return FORMAT_L8; |
| 2919 | case FORMAT_A4L4: return FORMAT_A8L8; |
| 2920 | case FORMAT_L16: return FORMAT_L16; |
| 2921 | case FORMAT_A8L8: return FORMAT_A8L8; |
| 2922 | // Depth/stencil formats |
| 2923 | case FORMAT_D16: |
| 2924 | case FORMAT_D32: |
| 2925 | case FORMAT_D24X8: |
| 2926 | case FORMAT_D24S8: |
| 2927 | case FORMAT_D24FS8: |
| 2928 | if(hasParent) // Texture |
| 2929 | { |
John Bauman | 66b8ab2 | 2014-05-06 15:57:45 -0400 | [diff] [blame^] | 2930 | return FORMAT_D32FS8_SHADOW; |
John Bauman | 8940182 | 2014-05-06 15:04:28 -0400 | [diff] [blame] | 2931 | } |
| 2932 | else if(complementaryDepthBuffer) |
| 2933 | { |
| 2934 | return FORMAT_D32F_COMPLEMENTARY; |
| 2935 | } |
| 2936 | else |
| 2937 | { |
| 2938 | return FORMAT_D32F; |
| 2939 | } |
John Bauman | 66b8ab2 | 2014-05-06 15:57:45 -0400 | [diff] [blame^] | 2940 | case FORMAT_D32F_LOCKABLE: return FORMAT_D32F_LOCKABLE; |
| 2941 | case FORMAT_D32FS8_TEXTURE: return FORMAT_D32FS8_TEXTURE; |
| 2942 | case FORMAT_INTZ: return FORMAT_D32FS8_TEXTURE; |
| 2943 | case FORMAT_DF24S8: return FORMAT_D32FS8_SHADOW; |
| 2944 | case FORMAT_DF16S8: return FORMAT_D32FS8_SHADOW; |
John Bauman | 8940182 | 2014-05-06 15:04:28 -0400 | [diff] [blame] | 2945 | default: |
| 2946 | ASSERT(false); |
| 2947 | } |
| 2948 | |
| 2949 | return FORMAT_NULL; |
| 2950 | } |
| 2951 | |
| 2952 | void Surface::setTexturePalette(unsigned int *palette) |
| 2953 | { |
| 2954 | Surface::palette = palette; |
| 2955 | Surface::paletteID++; |
| 2956 | } |
| 2957 | |
| 2958 | void Surface::resolve() |
| 2959 | { |
| 2960 | if(internal.depth <= 1 || !internal.dirty || !renderTarget || internal.format == FORMAT_NULL) |
| 2961 | { |
| 2962 | return; |
| 2963 | } |
| 2964 | |
| 2965 | void *source = internal.lockRect(0, 0, 0, LOCK_READWRITE); |
| 2966 | |
| 2967 | int quality = internal.depth; |
| 2968 | int width = internal.width; |
| 2969 | int height = internal.height; |
| 2970 | int pitch = internal.pitchB; |
| 2971 | int slice = internal.sliceB; |
| 2972 | |
| 2973 | unsigned char *source0 = (unsigned char*)source; |
| 2974 | unsigned char *source1 = source0 + slice; |
| 2975 | unsigned char *source2 = source1 + slice; |
| 2976 | unsigned char *source3 = source2 + slice; |
| 2977 | unsigned char *source4 = source3 + slice; |
| 2978 | unsigned char *source5 = source4 + slice; |
| 2979 | unsigned char *source6 = source5 + slice; |
| 2980 | unsigned char *source7 = source6 + slice; |
| 2981 | unsigned char *source8 = source7 + slice; |
| 2982 | unsigned char *source9 = source8 + slice; |
| 2983 | unsigned char *sourceA = source9 + slice; |
| 2984 | unsigned char *sourceB = sourceA + slice; |
| 2985 | unsigned char *sourceC = sourceB + slice; |
| 2986 | unsigned char *sourceD = sourceC + slice; |
| 2987 | unsigned char *sourceE = sourceD + slice; |
| 2988 | unsigned char *sourceF = sourceE + slice; |
| 2989 | |
| 2990 | if(internal.format == FORMAT_X8R8G8B8 || internal.format == FORMAT_A8R8G8B8) |
| 2991 | { |
| 2992 | if(CPUID::supportsSSE2() && (width % 4) == 0) |
| 2993 | { |
| 2994 | if(internal.depth == 2) |
| 2995 | { |
| 2996 | for(int y = 0; y < height; y++) |
| 2997 | { |
| 2998 | for(int x = 0; x < width; x += 4) |
| 2999 | { |
| 3000 | __m128i c0 = _mm_load_si128((__m128i*)(source0 + 4 * x)); |
| 3001 | __m128i c1 = _mm_load_si128((__m128i*)(source1 + 4 * x)); |
| 3002 | |
| 3003 | c0 = _mm_avg_epu8(c0, c1); |
| 3004 | |
| 3005 | _mm_store_si128((__m128i*)(source0 + 4 * x), c0); |
| 3006 | } |
| 3007 | |
| 3008 | source0 += pitch; |
| 3009 | source1 += pitch; |
| 3010 | } |
| 3011 | } |
| 3012 | else if(internal.depth == 4) |
| 3013 | { |
| 3014 | for(int y = 0; y < height; y++) |
| 3015 | { |
| 3016 | for(int x = 0; x < width; x += 4) |
| 3017 | { |
| 3018 | __m128i c0 = _mm_load_si128((__m128i*)(source0 + 4 * x)); |
| 3019 | __m128i c1 = _mm_load_si128((__m128i*)(source1 + 4 * x)); |
| 3020 | __m128i c2 = _mm_load_si128((__m128i*)(source2 + 4 * x)); |
| 3021 | __m128i c3 = _mm_load_si128((__m128i*)(source3 + 4 * x)); |
| 3022 | |
| 3023 | c0 = _mm_avg_epu8(c0, c1); |
| 3024 | c2 = _mm_avg_epu8(c2, c3); |
| 3025 | c0 = _mm_avg_epu8(c0, c2); |
| 3026 | |
| 3027 | _mm_store_si128((__m128i*)(source0 + 4 * x), c0); |
| 3028 | } |
| 3029 | |
| 3030 | source0 += pitch; |
| 3031 | source1 += pitch; |
| 3032 | source2 += pitch; |
| 3033 | source3 += pitch; |
| 3034 | } |
| 3035 | } |
| 3036 | else if(internal.depth == 8) |
| 3037 | { |
| 3038 | for(int y = 0; y < height; y++) |
| 3039 | { |
| 3040 | for(int x = 0; x < width; x += 4) |
| 3041 | { |
| 3042 | __m128i c0 = _mm_load_si128((__m128i*)(source0 + 4 * x)); |
| 3043 | __m128i c1 = _mm_load_si128((__m128i*)(source1 + 4 * x)); |
| 3044 | __m128i c2 = _mm_load_si128((__m128i*)(source2 + 4 * x)); |
| 3045 | __m128i c3 = _mm_load_si128((__m128i*)(source3 + 4 * x)); |
| 3046 | __m128i c4 = _mm_load_si128((__m128i*)(source4 + 4 * x)); |
| 3047 | __m128i c5 = _mm_load_si128((__m128i*)(source5 + 4 * x)); |
| 3048 | __m128i c6 = _mm_load_si128((__m128i*)(source6 + 4 * x)); |
| 3049 | __m128i c7 = _mm_load_si128((__m128i*)(source7 + 4 * x)); |
| 3050 | |
| 3051 | c0 = _mm_avg_epu8(c0, c1); |
| 3052 | c2 = _mm_avg_epu8(c2, c3); |
| 3053 | c4 = _mm_avg_epu8(c4, c5); |
| 3054 | c6 = _mm_avg_epu8(c6, c7); |
| 3055 | c0 = _mm_avg_epu8(c0, c2); |
| 3056 | c4 = _mm_avg_epu8(c4, c6); |
| 3057 | c0 = _mm_avg_epu8(c0, c4); |
| 3058 | |
| 3059 | _mm_store_si128((__m128i*)(source0 + 4 * x), c0); |
| 3060 | } |
| 3061 | |
| 3062 | source0 += pitch; |
| 3063 | source1 += pitch; |
| 3064 | source2 += pitch; |
| 3065 | source3 += pitch; |
| 3066 | source4 += pitch; |
| 3067 | source5 += pitch; |
| 3068 | source6 += pitch; |
| 3069 | source7 += pitch; |
| 3070 | } |
| 3071 | } |
| 3072 | else if(internal.depth == 16) |
| 3073 | { |
| 3074 | for(int y = 0; y < height; y++) |
| 3075 | { |
| 3076 | for(int x = 0; x < width; x += 4) |
| 3077 | { |
| 3078 | __m128i c0 = _mm_load_si128((__m128i*)(source0 + 4 * x)); |
| 3079 | __m128i c1 = _mm_load_si128((__m128i*)(source1 + 4 * x)); |
| 3080 | __m128i c2 = _mm_load_si128((__m128i*)(source2 + 4 * x)); |
| 3081 | __m128i c3 = _mm_load_si128((__m128i*)(source3 + 4 * x)); |
| 3082 | __m128i c4 = _mm_load_si128((__m128i*)(source4 + 4 * x)); |
| 3083 | __m128i c5 = _mm_load_si128((__m128i*)(source5 + 4 * x)); |
| 3084 | __m128i c6 = _mm_load_si128((__m128i*)(source6 + 4 * x)); |
| 3085 | __m128i c7 = _mm_load_si128((__m128i*)(source7 + 4 * x)); |
| 3086 | __m128i c8 = _mm_load_si128((__m128i*)(source8 + 4 * x)); |
| 3087 | __m128i c9 = _mm_load_si128((__m128i*)(source9 + 4 * x)); |
| 3088 | __m128i cA = _mm_load_si128((__m128i*)(sourceA + 4 * x)); |
| 3089 | __m128i cB = _mm_load_si128((__m128i*)(sourceB + 4 * x)); |
| 3090 | __m128i cC = _mm_load_si128((__m128i*)(sourceC + 4 * x)); |
| 3091 | __m128i cD = _mm_load_si128((__m128i*)(sourceD + 4 * x)); |
| 3092 | __m128i cE = _mm_load_si128((__m128i*)(sourceE + 4 * x)); |
| 3093 | __m128i cF = _mm_load_si128((__m128i*)(sourceF + 4 * x)); |
| 3094 | |
| 3095 | c0 = _mm_avg_epu8(c0, c1); |
| 3096 | c2 = _mm_avg_epu8(c2, c3); |
| 3097 | c4 = _mm_avg_epu8(c4, c5); |
| 3098 | c6 = _mm_avg_epu8(c6, c7); |
| 3099 | c8 = _mm_avg_epu8(c8, c9); |
| 3100 | cA = _mm_avg_epu8(cA, cB); |
| 3101 | cC = _mm_avg_epu8(cC, cD); |
| 3102 | cE = _mm_avg_epu8(cE, cF); |
| 3103 | c0 = _mm_avg_epu8(c0, c2); |
| 3104 | c4 = _mm_avg_epu8(c4, c6); |
| 3105 | c8 = _mm_avg_epu8(c8, cA); |
| 3106 | cC = _mm_avg_epu8(cC, cE); |
| 3107 | c0 = _mm_avg_epu8(c0, c4); |
| 3108 | c8 = _mm_avg_epu8(c8, cC); |
| 3109 | c0 = _mm_avg_epu8(c0, c8); |
| 3110 | |
| 3111 | _mm_store_si128((__m128i*)(source0 + 4 * x), c0); |
| 3112 | } |
| 3113 | |
| 3114 | source0 += pitch; |
| 3115 | source1 += pitch; |
| 3116 | source2 += pitch; |
| 3117 | source3 += pitch; |
| 3118 | source4 += pitch; |
| 3119 | source5 += pitch; |
| 3120 | source6 += pitch; |
| 3121 | source7 += pitch; |
| 3122 | source8 += pitch; |
| 3123 | source9 += pitch; |
| 3124 | sourceA += pitch; |
| 3125 | sourceB += pitch; |
| 3126 | sourceC += pitch; |
| 3127 | sourceD += pitch; |
| 3128 | sourceE += pitch; |
| 3129 | sourceF += pitch; |
| 3130 | } |
| 3131 | } |
| 3132 | else ASSERT(false); |
| 3133 | } |
| 3134 | else |
| 3135 | { |
| 3136 | #define AVERAGE(x, y) (((x) & (y)) + ((((x) ^ (y)) >> 1) & 0x7F7F7F7F) + (((x) ^ (y)) & 0x01010101)) |
| 3137 | |
| 3138 | if(internal.depth == 2) |
| 3139 | { |
| 3140 | for(int y = 0; y < height; y++) |
| 3141 | { |
| 3142 | for(int x = 0; x < width; x++) |
| 3143 | { |
| 3144 | unsigned int c0 = *(unsigned int*)(source0 + 4 * x); |
| 3145 | unsigned int c1 = *(unsigned int*)(source1 + 4 * x); |
| 3146 | |
| 3147 | c0 = AVERAGE(c0, c1); |
| 3148 | |
| 3149 | *(unsigned int*)(source0 + 4 * x) = c0; |
| 3150 | } |
| 3151 | |
| 3152 | source0 += pitch; |
| 3153 | source1 += pitch; |
| 3154 | } |
| 3155 | } |
| 3156 | else if(internal.depth == 4) |
| 3157 | { |
| 3158 | for(int y = 0; y < height; y++) |
| 3159 | { |
| 3160 | for(int x = 0; x < width; x++) |
| 3161 | { |
| 3162 | unsigned int c0 = *(unsigned int*)(source0 + 4 * x); |
| 3163 | unsigned int c1 = *(unsigned int*)(source1 + 4 * x); |
| 3164 | unsigned int c2 = *(unsigned int*)(source2 + 4 * x); |
| 3165 | unsigned int c3 = *(unsigned int*)(source3 + 4 * x); |
| 3166 | |
| 3167 | c0 = AVERAGE(c0, c1); |
| 3168 | c2 = AVERAGE(c2, c3); |
| 3169 | c0 = AVERAGE(c0, c2); |
| 3170 | |
| 3171 | *(unsigned int*)(source0 + 4 * x) = c0; |
| 3172 | } |
| 3173 | |
| 3174 | source0 += pitch; |
| 3175 | source1 += pitch; |
| 3176 | source2 += pitch; |
| 3177 | source3 += pitch; |
| 3178 | } |
| 3179 | } |
| 3180 | else if(internal.depth == 8) |
| 3181 | { |
| 3182 | for(int y = 0; y < height; y++) |
| 3183 | { |
| 3184 | for(int x = 0; x < width; x++) |
| 3185 | { |
| 3186 | unsigned int c0 = *(unsigned int*)(source0 + 4 * x); |
| 3187 | unsigned int c1 = *(unsigned int*)(source1 + 4 * x); |
| 3188 | unsigned int c2 = *(unsigned int*)(source2 + 4 * x); |
| 3189 | unsigned int c3 = *(unsigned int*)(source3 + 4 * x); |
| 3190 | unsigned int c4 = *(unsigned int*)(source4 + 4 * x); |
| 3191 | unsigned int c5 = *(unsigned int*)(source5 + 4 * x); |
| 3192 | unsigned int c6 = *(unsigned int*)(source6 + 4 * x); |
| 3193 | unsigned int c7 = *(unsigned int*)(source7 + 4 * x); |
| 3194 | |
| 3195 | c0 = AVERAGE(c0, c1); |
| 3196 | c2 = AVERAGE(c2, c3); |
| 3197 | c4 = AVERAGE(c4, c5); |
| 3198 | c6 = AVERAGE(c6, c7); |
| 3199 | c0 = AVERAGE(c0, c2); |
| 3200 | c4 = AVERAGE(c4, c6); |
| 3201 | c0 = AVERAGE(c0, c4); |
| 3202 | |
| 3203 | *(unsigned int*)(source0 + 4 * x) = c0; |
| 3204 | } |
| 3205 | |
| 3206 | source0 += pitch; |
| 3207 | source1 += pitch; |
| 3208 | source2 += pitch; |
| 3209 | source3 += pitch; |
| 3210 | source4 += pitch; |
| 3211 | source5 += pitch; |
| 3212 | source6 += pitch; |
| 3213 | source7 += pitch; |
| 3214 | } |
| 3215 | } |
| 3216 | else if(internal.depth == 16) |
| 3217 | { |
| 3218 | for(int y = 0; y < height; y++) |
| 3219 | { |
| 3220 | for(int x = 0; x < width; x++) |
| 3221 | { |
| 3222 | unsigned int c0 = *(unsigned int*)(source0 + 4 * x); |
| 3223 | unsigned int c1 = *(unsigned int*)(source1 + 4 * x); |
| 3224 | unsigned int c2 = *(unsigned int*)(source2 + 4 * x); |
| 3225 | unsigned int c3 = *(unsigned int*)(source3 + 4 * x); |
| 3226 | unsigned int c4 = *(unsigned int*)(source4 + 4 * x); |
| 3227 | unsigned int c5 = *(unsigned int*)(source5 + 4 * x); |
| 3228 | unsigned int c6 = *(unsigned int*)(source6 + 4 * x); |
| 3229 | unsigned int c7 = *(unsigned int*)(source7 + 4 * x); |
| 3230 | unsigned int c8 = *(unsigned int*)(source8 + 4 * x); |
| 3231 | unsigned int c9 = *(unsigned int*)(source9 + 4 * x); |
| 3232 | unsigned int cA = *(unsigned int*)(sourceA + 4 * x); |
| 3233 | unsigned int cB = *(unsigned int*)(sourceB + 4 * x); |
| 3234 | unsigned int cC = *(unsigned int*)(sourceC + 4 * x); |
| 3235 | unsigned int cD = *(unsigned int*)(sourceD + 4 * x); |
| 3236 | unsigned int cE = *(unsigned int*)(sourceE + 4 * x); |
| 3237 | unsigned int cF = *(unsigned int*)(sourceF + 4 * x); |
| 3238 | |
| 3239 | c0 = AVERAGE(c0, c1); |
| 3240 | c2 = AVERAGE(c2, c3); |
| 3241 | c4 = AVERAGE(c4, c5); |
| 3242 | c6 = AVERAGE(c6, c7); |
| 3243 | c8 = AVERAGE(c8, c9); |
| 3244 | cA = AVERAGE(cA, cB); |
| 3245 | cC = AVERAGE(cC, cD); |
| 3246 | cE = AVERAGE(cE, cF); |
| 3247 | c0 = AVERAGE(c0, c2); |
| 3248 | c4 = AVERAGE(c4, c6); |
| 3249 | c8 = AVERAGE(c8, cA); |
| 3250 | cC = AVERAGE(cC, cE); |
| 3251 | c0 = AVERAGE(c0, c4); |
| 3252 | c8 = AVERAGE(c8, cC); |
| 3253 | c0 = AVERAGE(c0, c8); |
| 3254 | |
| 3255 | *(unsigned int*)(source0 + 4 * x) = c0; |
| 3256 | } |
| 3257 | |
| 3258 | source0 += pitch; |
| 3259 | source1 += pitch; |
| 3260 | source2 += pitch; |
| 3261 | source3 += pitch; |
| 3262 | source4 += pitch; |
| 3263 | source5 += pitch; |
| 3264 | source6 += pitch; |
| 3265 | source7 += pitch; |
| 3266 | source8 += pitch; |
| 3267 | source9 += pitch; |
| 3268 | sourceA += pitch; |
| 3269 | sourceB += pitch; |
| 3270 | sourceC += pitch; |
| 3271 | sourceD += pitch; |
| 3272 | sourceE += pitch; |
| 3273 | sourceF += pitch; |
| 3274 | } |
| 3275 | } |
| 3276 | else ASSERT(false); |
| 3277 | |
| 3278 | #undef AVERAGE |
| 3279 | } |
| 3280 | } |
| 3281 | else if(internal.format == FORMAT_G16R16) |
| 3282 | { |
| 3283 | if(CPUID::supportsSSE2() && (width % 4) == 0) |
| 3284 | { |
| 3285 | if(internal.depth == 2) |
| 3286 | { |
| 3287 | for(int y = 0; y < height; y++) |
| 3288 | { |
| 3289 | for(int x = 0; x < width; x += 4) |
| 3290 | { |
| 3291 | __m128i c0 = _mm_load_si128((__m128i*)(source0 + 4 * x)); |
| 3292 | __m128i c1 = _mm_load_si128((__m128i*)(source1 + 4 * x)); |
| 3293 | |
| 3294 | c0 = _mm_avg_epu16(c0, c1); |
| 3295 | |
| 3296 | _mm_store_si128((__m128i*)(source0 + 4 * x), c0); |
| 3297 | } |
| 3298 | |
| 3299 | source0 += pitch; |
| 3300 | source1 += pitch; |
| 3301 | } |
| 3302 | } |
| 3303 | else if(internal.depth == 4) |
| 3304 | { |
| 3305 | for(int y = 0; y < height; y++) |
| 3306 | { |
| 3307 | for(int x = 0; x < width; x += 4) |
| 3308 | { |
| 3309 | __m128i c0 = _mm_load_si128((__m128i*)(source0 + 4 * x)); |
| 3310 | __m128i c1 = _mm_load_si128((__m128i*)(source1 + 4 * x)); |
| 3311 | __m128i c2 = _mm_load_si128((__m128i*)(source2 + 4 * x)); |
| 3312 | __m128i c3 = _mm_load_si128((__m128i*)(source3 + 4 * x)); |
| 3313 | |
| 3314 | c0 = _mm_avg_epu16(c0, c1); |
| 3315 | c2 = _mm_avg_epu16(c2, c3); |
| 3316 | c0 = _mm_avg_epu16(c0, c2); |
| 3317 | |
| 3318 | _mm_store_si128((__m128i*)(source0 + 4 * x), c0); |
| 3319 | } |
| 3320 | |
| 3321 | source0 += pitch; |
| 3322 | source1 += pitch; |
| 3323 | source2 += pitch; |
| 3324 | source3 += pitch; |
| 3325 | } |
| 3326 | } |
| 3327 | else if(internal.depth == 8) |
| 3328 | { |
| 3329 | for(int y = 0; y < height; y++) |
| 3330 | { |
| 3331 | for(int x = 0; x < width; x += 4) |
| 3332 | { |
| 3333 | __m128i c0 = _mm_load_si128((__m128i*)(source0 + 4 * x)); |
| 3334 | __m128i c1 = _mm_load_si128((__m128i*)(source1 + 4 * x)); |
| 3335 | __m128i c2 = _mm_load_si128((__m128i*)(source2 + 4 * x)); |
| 3336 | __m128i c3 = _mm_load_si128((__m128i*)(source3 + 4 * x)); |
| 3337 | __m128i c4 = _mm_load_si128((__m128i*)(source4 + 4 * x)); |
| 3338 | __m128i c5 = _mm_load_si128((__m128i*)(source5 + 4 * x)); |
| 3339 | __m128i c6 = _mm_load_si128((__m128i*)(source6 + 4 * x)); |
| 3340 | __m128i c7 = _mm_load_si128((__m128i*)(source7 + 4 * x)); |
| 3341 | |
| 3342 | c0 = _mm_avg_epu16(c0, c1); |
| 3343 | c2 = _mm_avg_epu16(c2, c3); |
| 3344 | c4 = _mm_avg_epu16(c4, c5); |
| 3345 | c6 = _mm_avg_epu16(c6, c7); |
| 3346 | c0 = _mm_avg_epu16(c0, c2); |
| 3347 | c4 = _mm_avg_epu16(c4, c6); |
| 3348 | c0 = _mm_avg_epu16(c0, c4); |
| 3349 | |
| 3350 | _mm_store_si128((__m128i*)(source0 + 4 * x), c0); |
| 3351 | } |
| 3352 | |
| 3353 | source0 += pitch; |
| 3354 | source1 += pitch; |
| 3355 | source2 += pitch; |
| 3356 | source3 += pitch; |
| 3357 | source4 += pitch; |
| 3358 | source5 += pitch; |
| 3359 | source6 += pitch; |
| 3360 | source7 += pitch; |
| 3361 | } |
| 3362 | } |
| 3363 | else if(internal.depth == 16) |
| 3364 | { |
| 3365 | for(int y = 0; y < height; y++) |
| 3366 | { |
| 3367 | for(int x = 0; x < width; x += 4) |
| 3368 | { |
| 3369 | __m128i c0 = _mm_load_si128((__m128i*)(source0 + 4 * x)); |
| 3370 | __m128i c1 = _mm_load_si128((__m128i*)(source1 + 4 * x)); |
| 3371 | __m128i c2 = _mm_load_si128((__m128i*)(source2 + 4 * x)); |
| 3372 | __m128i c3 = _mm_load_si128((__m128i*)(source3 + 4 * x)); |
| 3373 | __m128i c4 = _mm_load_si128((__m128i*)(source4 + 4 * x)); |
| 3374 | __m128i c5 = _mm_load_si128((__m128i*)(source5 + 4 * x)); |
| 3375 | __m128i c6 = _mm_load_si128((__m128i*)(source6 + 4 * x)); |
| 3376 | __m128i c7 = _mm_load_si128((__m128i*)(source7 + 4 * x)); |
| 3377 | __m128i c8 = _mm_load_si128((__m128i*)(source8 + 4 * x)); |
| 3378 | __m128i c9 = _mm_load_si128((__m128i*)(source9 + 4 * x)); |
| 3379 | __m128i cA = _mm_load_si128((__m128i*)(sourceA + 4 * x)); |
| 3380 | __m128i cB = _mm_load_si128((__m128i*)(sourceB + 4 * x)); |
| 3381 | __m128i cC = _mm_load_si128((__m128i*)(sourceC + 4 * x)); |
| 3382 | __m128i cD = _mm_load_si128((__m128i*)(sourceD + 4 * x)); |
| 3383 | __m128i cE = _mm_load_si128((__m128i*)(sourceE + 4 * x)); |
| 3384 | __m128i cF = _mm_load_si128((__m128i*)(sourceF + 4 * x)); |
| 3385 | |
| 3386 | c0 = _mm_avg_epu16(c0, c1); |
| 3387 | c2 = _mm_avg_epu16(c2, c3); |
| 3388 | c4 = _mm_avg_epu16(c4, c5); |
| 3389 | c6 = _mm_avg_epu16(c6, c7); |
| 3390 | c8 = _mm_avg_epu16(c8, c9); |
| 3391 | cA = _mm_avg_epu16(cA, cB); |
| 3392 | cC = _mm_avg_epu16(cC, cD); |
| 3393 | cE = _mm_avg_epu16(cE, cF); |
| 3394 | c0 = _mm_avg_epu16(c0, c2); |
| 3395 | c4 = _mm_avg_epu16(c4, c6); |
| 3396 | c8 = _mm_avg_epu16(c8, cA); |
| 3397 | cC = _mm_avg_epu16(cC, cE); |
| 3398 | c0 = _mm_avg_epu16(c0, c4); |
| 3399 | c8 = _mm_avg_epu16(c8, cC); |
| 3400 | c0 = _mm_avg_epu16(c0, c8); |
| 3401 | |
| 3402 | _mm_store_si128((__m128i*)(source0 + 4 * x), c0); |
| 3403 | } |
| 3404 | |
| 3405 | source0 += pitch; |
| 3406 | source1 += pitch; |
| 3407 | source2 += pitch; |
| 3408 | source3 += pitch; |
| 3409 | source4 += pitch; |
| 3410 | source5 += pitch; |
| 3411 | source6 += pitch; |
| 3412 | source7 += pitch; |
| 3413 | source8 += pitch; |
| 3414 | source9 += pitch; |
| 3415 | sourceA += pitch; |
| 3416 | sourceB += pitch; |
| 3417 | sourceC += pitch; |
| 3418 | sourceD += pitch; |
| 3419 | sourceE += pitch; |
| 3420 | sourceF += pitch; |
| 3421 | } |
| 3422 | } |
| 3423 | else ASSERT(false); |
| 3424 | } |
| 3425 | else |
| 3426 | { |
| 3427 | #define AVERAGE(x, y) (((x) & (y)) + ((((x) ^ (y)) >> 1) & 0x7FFF7FFF) + (((x) ^ (y)) & 0x00010001)) |
| 3428 | |
| 3429 | if(internal.depth == 2) |
| 3430 | { |
| 3431 | for(int y = 0; y < height; y++) |
| 3432 | { |
| 3433 | for(int x = 0; x < width; x++) |
| 3434 | { |
| 3435 | unsigned int c0 = *(unsigned int*)(source0 + 4 * x); |
| 3436 | unsigned int c1 = *(unsigned int*)(source1 + 4 * x); |
| 3437 | |
| 3438 | c0 = AVERAGE(c0, c1); |
| 3439 | |
| 3440 | *(unsigned int*)(source0 + 4 * x) = c0; |
| 3441 | } |
| 3442 | |
| 3443 | source0 += pitch; |
| 3444 | source1 += pitch; |
| 3445 | } |
| 3446 | } |
| 3447 | else if(internal.depth == 4) |
| 3448 | { |
| 3449 | for(int y = 0; y < height; y++) |
| 3450 | { |
| 3451 | for(int x = 0; x < width; x++) |
| 3452 | { |
| 3453 | unsigned int c0 = *(unsigned int*)(source0 + 4 * x); |
| 3454 | unsigned int c1 = *(unsigned int*)(source1 + 4 * x); |
| 3455 | unsigned int c2 = *(unsigned int*)(source2 + 4 * x); |
| 3456 | unsigned int c3 = *(unsigned int*)(source3 + 4 * x); |
| 3457 | |
| 3458 | c0 = AVERAGE(c0, c1); |
| 3459 | c2 = AVERAGE(c2, c3); |
| 3460 | c0 = AVERAGE(c0, c2); |
| 3461 | |
| 3462 | *(unsigned int*)(source0 + 4 * x) = c0; |
| 3463 | } |
| 3464 | |
| 3465 | source0 += pitch; |
| 3466 | source1 += pitch; |
| 3467 | source2 += pitch; |
| 3468 | source3 += pitch; |
| 3469 | } |
| 3470 | } |
| 3471 | else if(internal.depth == 8) |
| 3472 | { |
| 3473 | for(int y = 0; y < height; y++) |
| 3474 | { |
| 3475 | for(int x = 0; x < width; x++) |
| 3476 | { |
| 3477 | unsigned int c0 = *(unsigned int*)(source0 + 4 * x); |
| 3478 | unsigned int c1 = *(unsigned int*)(source1 + 4 * x); |
| 3479 | unsigned int c2 = *(unsigned int*)(source2 + 4 * x); |
| 3480 | unsigned int c3 = *(unsigned int*)(source3 + 4 * x); |
| 3481 | unsigned int c4 = *(unsigned int*)(source4 + 4 * x); |
| 3482 | unsigned int c5 = *(unsigned int*)(source5 + 4 * x); |
| 3483 | unsigned int c6 = *(unsigned int*)(source6 + 4 * x); |
| 3484 | unsigned int c7 = *(unsigned int*)(source7 + 4 * x); |
| 3485 | |
| 3486 | c0 = AVERAGE(c0, c1); |
| 3487 | c2 = AVERAGE(c2, c3); |
| 3488 | c4 = AVERAGE(c4, c5); |
| 3489 | c6 = AVERAGE(c6, c7); |
| 3490 | c0 = AVERAGE(c0, c2); |
| 3491 | c4 = AVERAGE(c4, c6); |
| 3492 | c0 = AVERAGE(c0, c4); |
| 3493 | |
| 3494 | *(unsigned int*)(source0 + 4 * x) = c0; |
| 3495 | } |
| 3496 | |
| 3497 | source0 += pitch; |
| 3498 | source1 += pitch; |
| 3499 | source2 += pitch; |
| 3500 | source3 += pitch; |
| 3501 | source4 += pitch; |
| 3502 | source5 += pitch; |
| 3503 | source6 += pitch; |
| 3504 | source7 += pitch; |
| 3505 | } |
| 3506 | } |
| 3507 | else if(internal.depth == 16) |
| 3508 | { |
| 3509 | for(int y = 0; y < height; y++) |
| 3510 | { |
| 3511 | for(int x = 0; x < width; x++) |
| 3512 | { |
| 3513 | unsigned int c0 = *(unsigned int*)(source0 + 4 * x); |
| 3514 | unsigned int c1 = *(unsigned int*)(source1 + 4 * x); |
| 3515 | unsigned int c2 = *(unsigned int*)(source2 + 4 * x); |
| 3516 | unsigned int c3 = *(unsigned int*)(source3 + 4 * x); |
| 3517 | unsigned int c4 = *(unsigned int*)(source4 + 4 * x); |
| 3518 | unsigned int c5 = *(unsigned int*)(source5 + 4 * x); |
| 3519 | unsigned int c6 = *(unsigned int*)(source6 + 4 * x); |
| 3520 | unsigned int c7 = *(unsigned int*)(source7 + 4 * x); |
| 3521 | unsigned int c8 = *(unsigned int*)(source8 + 4 * x); |
| 3522 | unsigned int c9 = *(unsigned int*)(source9 + 4 * x); |
| 3523 | unsigned int cA = *(unsigned int*)(sourceA + 4 * x); |
| 3524 | unsigned int cB = *(unsigned int*)(sourceB + 4 * x); |
| 3525 | unsigned int cC = *(unsigned int*)(sourceC + 4 * x); |
| 3526 | unsigned int cD = *(unsigned int*)(sourceD + 4 * x); |
| 3527 | unsigned int cE = *(unsigned int*)(sourceE + 4 * x); |
| 3528 | unsigned int cF = *(unsigned int*)(sourceF + 4 * x); |
| 3529 | |
| 3530 | c0 = AVERAGE(c0, c1); |
| 3531 | c2 = AVERAGE(c2, c3); |
| 3532 | c4 = AVERAGE(c4, c5); |
| 3533 | c6 = AVERAGE(c6, c7); |
| 3534 | c8 = AVERAGE(c8, c9); |
| 3535 | cA = AVERAGE(cA, cB); |
| 3536 | cC = AVERAGE(cC, cD); |
| 3537 | cE = AVERAGE(cE, cF); |
| 3538 | c0 = AVERAGE(c0, c2); |
| 3539 | c4 = AVERAGE(c4, c6); |
| 3540 | c8 = AVERAGE(c8, cA); |
| 3541 | cC = AVERAGE(cC, cE); |
| 3542 | c0 = AVERAGE(c0, c4); |
| 3543 | c8 = AVERAGE(c8, cC); |
| 3544 | c0 = AVERAGE(c0, c8); |
| 3545 | |
| 3546 | *(unsigned int*)(source0 + 4 * x) = c0; |
| 3547 | } |
| 3548 | |
| 3549 | source0 += pitch; |
| 3550 | source1 += pitch; |
| 3551 | source2 += pitch; |
| 3552 | source3 += pitch; |
| 3553 | source4 += pitch; |
| 3554 | source5 += pitch; |
| 3555 | source6 += pitch; |
| 3556 | source7 += pitch; |
| 3557 | source8 += pitch; |
| 3558 | source9 += pitch; |
| 3559 | sourceA += pitch; |
| 3560 | sourceB += pitch; |
| 3561 | sourceC += pitch; |
| 3562 | sourceD += pitch; |
| 3563 | sourceE += pitch; |
| 3564 | sourceF += pitch; |
| 3565 | } |
| 3566 | } |
| 3567 | else ASSERT(false); |
| 3568 | |
| 3569 | #undef AVERAGE |
| 3570 | } |
| 3571 | } |
| 3572 | else if(internal.format == FORMAT_A16B16G16R16) |
| 3573 | { |
| 3574 | if(CPUID::supportsSSE2() && (width % 2) == 0) |
| 3575 | { |
| 3576 | if(internal.depth == 2) |
| 3577 | { |
| 3578 | for(int y = 0; y < height; y++) |
| 3579 | { |
| 3580 | for(int x = 0; x < width; x += 2) |
| 3581 | { |
| 3582 | __m128i c0 = _mm_load_si128((__m128i*)(source0 + 8 * x)); |
| 3583 | __m128i c1 = _mm_load_si128((__m128i*)(source1 + 8 * x)); |
| 3584 | |
| 3585 | c0 = _mm_avg_epu16(c0, c1); |
| 3586 | |
| 3587 | _mm_store_si128((__m128i*)(source0 + 8 * x), c0); |
| 3588 | } |
| 3589 | |
| 3590 | source0 += pitch; |
| 3591 | source1 += pitch; |
| 3592 | } |
| 3593 | } |
| 3594 | else if(internal.depth == 4) |
| 3595 | { |
| 3596 | for(int y = 0; y < height; y++) |
| 3597 | { |
| 3598 | for(int x = 0; x < width; x += 2) |
| 3599 | { |
| 3600 | __m128i c0 = _mm_load_si128((__m128i*)(source0 + 8 * x)); |
| 3601 | __m128i c1 = _mm_load_si128((__m128i*)(source1 + 8 * x)); |
| 3602 | __m128i c2 = _mm_load_si128((__m128i*)(source2 + 8 * x)); |
| 3603 | __m128i c3 = _mm_load_si128((__m128i*)(source3 + 8 * x)); |
| 3604 | |
| 3605 | c0 = _mm_avg_epu16(c0, c1); |
| 3606 | c2 = _mm_avg_epu16(c2, c3); |
| 3607 | c0 = _mm_avg_epu16(c0, c2); |
| 3608 | |
| 3609 | _mm_store_si128((__m128i*)(source0 + 8 * x), c0); |
| 3610 | } |
| 3611 | |
| 3612 | source0 += pitch; |
| 3613 | source1 += pitch; |
| 3614 | source2 += pitch; |
| 3615 | source3 += pitch; |
| 3616 | } |
| 3617 | } |
| 3618 | else if(internal.depth == 8) |
| 3619 | { |
| 3620 | for(int y = 0; y < height; y++) |
| 3621 | { |
| 3622 | for(int x = 0; x < width; x += 2) |
| 3623 | { |
| 3624 | __m128i c0 = _mm_load_si128((__m128i*)(source0 + 8 * x)); |
| 3625 | __m128i c1 = _mm_load_si128((__m128i*)(source1 + 8 * x)); |
| 3626 | __m128i c2 = _mm_load_si128((__m128i*)(source2 + 8 * x)); |
| 3627 | __m128i c3 = _mm_load_si128((__m128i*)(source3 + 8 * x)); |
| 3628 | __m128i c4 = _mm_load_si128((__m128i*)(source4 + 8 * x)); |
| 3629 | __m128i c5 = _mm_load_si128((__m128i*)(source5 + 8 * x)); |
| 3630 | __m128i c6 = _mm_load_si128((__m128i*)(source6 + 8 * x)); |
| 3631 | __m128i c7 = _mm_load_si128((__m128i*)(source7 + 8 * x)); |
| 3632 | |
| 3633 | c0 = _mm_avg_epu16(c0, c1); |
| 3634 | c2 = _mm_avg_epu16(c2, c3); |
| 3635 | c4 = _mm_avg_epu16(c4, c5); |
| 3636 | c6 = _mm_avg_epu16(c6, c7); |
| 3637 | c0 = _mm_avg_epu16(c0, c2); |
| 3638 | c4 = _mm_avg_epu16(c4, c6); |
| 3639 | c0 = _mm_avg_epu16(c0, c4); |
| 3640 | |
| 3641 | _mm_store_si128((__m128i*)(source0 + 8 * x), c0); |
| 3642 | } |
| 3643 | |
| 3644 | source0 += pitch; |
| 3645 | source1 += pitch; |
| 3646 | source2 += pitch; |
| 3647 | source3 += pitch; |
| 3648 | source4 += pitch; |
| 3649 | source5 += pitch; |
| 3650 | source6 += pitch; |
| 3651 | source7 += pitch; |
| 3652 | } |
| 3653 | } |
| 3654 | else if(internal.depth == 16) |
| 3655 | { |
| 3656 | for(int y = 0; y < height; y++) |
| 3657 | { |
| 3658 | for(int x = 0; x < width; x += 2) |
| 3659 | { |
| 3660 | __m128i c0 = _mm_load_si128((__m128i*)(source0 + 8 * x)); |
| 3661 | __m128i c1 = _mm_load_si128((__m128i*)(source1 + 8 * x)); |
| 3662 | __m128i c2 = _mm_load_si128((__m128i*)(source2 + 8 * x)); |
| 3663 | __m128i c3 = _mm_load_si128((__m128i*)(source3 + 8 * x)); |
| 3664 | __m128i c4 = _mm_load_si128((__m128i*)(source4 + 8 * x)); |
| 3665 | __m128i c5 = _mm_load_si128((__m128i*)(source5 + 8 * x)); |
| 3666 | __m128i c6 = _mm_load_si128((__m128i*)(source6 + 8 * x)); |
| 3667 | __m128i c7 = _mm_load_si128((__m128i*)(source7 + 8 * x)); |
| 3668 | __m128i c8 = _mm_load_si128((__m128i*)(source8 + 8 * x)); |
| 3669 | __m128i c9 = _mm_load_si128((__m128i*)(source9 + 8 * x)); |
| 3670 | __m128i cA = _mm_load_si128((__m128i*)(sourceA + 8 * x)); |
| 3671 | __m128i cB = _mm_load_si128((__m128i*)(sourceB + 8 * x)); |
| 3672 | __m128i cC = _mm_load_si128((__m128i*)(sourceC + 8 * x)); |
| 3673 | __m128i cD = _mm_load_si128((__m128i*)(sourceD + 8 * x)); |
| 3674 | __m128i cE = _mm_load_si128((__m128i*)(sourceE + 8 * x)); |
| 3675 | __m128i cF = _mm_load_si128((__m128i*)(sourceF + 8 * x)); |
| 3676 | |
| 3677 | c0 = _mm_avg_epu16(c0, c1); |
| 3678 | c2 = _mm_avg_epu16(c2, c3); |
| 3679 | c4 = _mm_avg_epu16(c4, c5); |
| 3680 | c6 = _mm_avg_epu16(c6, c7); |
| 3681 | c8 = _mm_avg_epu16(c8, c9); |
| 3682 | cA = _mm_avg_epu16(cA, cB); |
| 3683 | cC = _mm_avg_epu16(cC, cD); |
| 3684 | cE = _mm_avg_epu16(cE, cF); |
| 3685 | c0 = _mm_avg_epu16(c0, c2); |
| 3686 | c4 = _mm_avg_epu16(c4, c6); |
| 3687 | c8 = _mm_avg_epu16(c8, cA); |
| 3688 | cC = _mm_avg_epu16(cC, cE); |
| 3689 | c0 = _mm_avg_epu16(c0, c4); |
| 3690 | c8 = _mm_avg_epu16(c8, cC); |
| 3691 | c0 = _mm_avg_epu16(c0, c8); |
| 3692 | |
| 3693 | _mm_store_si128((__m128i*)(source0 + 8 * x), c0); |
| 3694 | } |
| 3695 | |
| 3696 | source0 += pitch; |
| 3697 | source1 += pitch; |
| 3698 | source2 += pitch; |
| 3699 | source3 += pitch; |
| 3700 | source4 += pitch; |
| 3701 | source5 += pitch; |
| 3702 | source6 += pitch; |
| 3703 | source7 += pitch; |
| 3704 | source8 += pitch; |
| 3705 | source9 += pitch; |
| 3706 | sourceA += pitch; |
| 3707 | sourceB += pitch; |
| 3708 | sourceC += pitch; |
| 3709 | sourceD += pitch; |
| 3710 | sourceE += pitch; |
| 3711 | sourceF += pitch; |
| 3712 | } |
| 3713 | } |
| 3714 | else ASSERT(false); |
| 3715 | } |
| 3716 | else |
| 3717 | { |
| 3718 | #define AVERAGE(x, y) (((x) & (y)) + ((((x) ^ (y)) >> 1) & 0x7FFF7FFF) + (((x) ^ (y)) & 0x00010001)) |
| 3719 | |
| 3720 | if(internal.depth == 2) |
| 3721 | { |
| 3722 | for(int y = 0; y < height; y++) |
| 3723 | { |
| 3724 | for(int x = 0; x < 2 * width; x++) |
| 3725 | { |
| 3726 | unsigned int c0 = *(unsigned int*)(source0 + 4 * x); |
| 3727 | unsigned int c1 = *(unsigned int*)(source1 + 4 * x); |
| 3728 | |
| 3729 | c0 = AVERAGE(c0, c1); |
| 3730 | |
| 3731 | *(unsigned int*)(source0 + 4 * x) = c0; |
| 3732 | } |
| 3733 | |
| 3734 | source0 += pitch; |
| 3735 | source1 += pitch; |
| 3736 | } |
| 3737 | } |
| 3738 | else if(internal.depth == 4) |
| 3739 | { |
| 3740 | for(int y = 0; y < height; y++) |
| 3741 | { |
| 3742 | for(int x = 0; x < 2 * width; x++) |
| 3743 | { |
| 3744 | unsigned int c0 = *(unsigned int*)(source0 + 4 * x); |
| 3745 | unsigned int c1 = *(unsigned int*)(source1 + 4 * x); |
| 3746 | unsigned int c2 = *(unsigned int*)(source2 + 4 * x); |
| 3747 | unsigned int c3 = *(unsigned int*)(source3 + 4 * x); |
| 3748 | |
| 3749 | c0 = AVERAGE(c0, c1); |
| 3750 | c2 = AVERAGE(c2, c3); |
| 3751 | c0 = AVERAGE(c0, c2); |
| 3752 | |
| 3753 | *(unsigned int*)(source0 + 4 * x) = c0; |
| 3754 | } |
| 3755 | |
| 3756 | source0 += pitch; |
| 3757 | source1 += pitch; |
| 3758 | source2 += pitch; |
| 3759 | source3 += pitch; |
| 3760 | } |
| 3761 | } |
| 3762 | else if(internal.depth == 8) |
| 3763 | { |
| 3764 | for(int y = 0; y < height; y++) |
| 3765 | { |
| 3766 | for(int x = 0; x < 2 * width; x++) |
| 3767 | { |
| 3768 | unsigned int c0 = *(unsigned int*)(source0 + 4 * x); |
| 3769 | unsigned int c1 = *(unsigned int*)(source1 + 4 * x); |
| 3770 | unsigned int c2 = *(unsigned int*)(source2 + 4 * x); |
| 3771 | unsigned int c3 = *(unsigned int*)(source3 + 4 * x); |
| 3772 | unsigned int c4 = *(unsigned int*)(source4 + 4 * x); |
| 3773 | unsigned int c5 = *(unsigned int*)(source5 + 4 * x); |
| 3774 | unsigned int c6 = *(unsigned int*)(source6 + 4 * x); |
| 3775 | unsigned int c7 = *(unsigned int*)(source7 + 4 * x); |
| 3776 | |
| 3777 | c0 = AVERAGE(c0, c1); |
| 3778 | c2 = AVERAGE(c2, c3); |
| 3779 | c4 = AVERAGE(c4, c5); |
| 3780 | c6 = AVERAGE(c6, c7); |
| 3781 | c0 = AVERAGE(c0, c2); |
| 3782 | c4 = AVERAGE(c4, c6); |
| 3783 | c0 = AVERAGE(c0, c4); |
| 3784 | |
| 3785 | *(unsigned int*)(source0 + 4 * x) = c0; |
| 3786 | } |
| 3787 | |
| 3788 | source0 += pitch; |
| 3789 | source1 += pitch; |
| 3790 | source2 += pitch; |
| 3791 | source3 += pitch; |
| 3792 | source4 += pitch; |
| 3793 | source5 += pitch; |
| 3794 | source6 += pitch; |
| 3795 | source7 += pitch; |
| 3796 | } |
| 3797 | } |
| 3798 | else if(internal.depth == 16) |
| 3799 | { |
| 3800 | for(int y = 0; y < height; y++) |
| 3801 | { |
| 3802 | for(int x = 0; x < 2 * width; x++) |
| 3803 | { |
| 3804 | unsigned int c0 = *(unsigned int*)(source0 + 4 * x); |
| 3805 | unsigned int c1 = *(unsigned int*)(source1 + 4 * x); |
| 3806 | unsigned int c2 = *(unsigned int*)(source2 + 4 * x); |
| 3807 | unsigned int c3 = *(unsigned int*)(source3 + 4 * x); |
| 3808 | unsigned int c4 = *(unsigned int*)(source4 + 4 * x); |
| 3809 | unsigned int c5 = *(unsigned int*)(source5 + 4 * x); |
| 3810 | unsigned int c6 = *(unsigned int*)(source6 + 4 * x); |
| 3811 | unsigned int c7 = *(unsigned int*)(source7 + 4 * x); |
| 3812 | unsigned int c8 = *(unsigned int*)(source8 + 4 * x); |
| 3813 | unsigned int c9 = *(unsigned int*)(source9 + 4 * x); |
| 3814 | unsigned int cA = *(unsigned int*)(sourceA + 4 * x); |
| 3815 | unsigned int cB = *(unsigned int*)(sourceB + 4 * x); |
| 3816 | unsigned int cC = *(unsigned int*)(sourceC + 4 * x); |
| 3817 | unsigned int cD = *(unsigned int*)(sourceD + 4 * x); |
| 3818 | unsigned int cE = *(unsigned int*)(sourceE + 4 * x); |
| 3819 | unsigned int cF = *(unsigned int*)(sourceF + 4 * x); |
| 3820 | |
| 3821 | c0 = AVERAGE(c0, c1); |
| 3822 | c2 = AVERAGE(c2, c3); |
| 3823 | c4 = AVERAGE(c4, c5); |
| 3824 | c6 = AVERAGE(c6, c7); |
| 3825 | c8 = AVERAGE(c8, c9); |
| 3826 | cA = AVERAGE(cA, cB); |
| 3827 | cC = AVERAGE(cC, cD); |
| 3828 | cE = AVERAGE(cE, cF); |
| 3829 | c0 = AVERAGE(c0, c2); |
| 3830 | c4 = AVERAGE(c4, c6); |
| 3831 | c8 = AVERAGE(c8, cA); |
| 3832 | cC = AVERAGE(cC, cE); |
| 3833 | c0 = AVERAGE(c0, c4); |
| 3834 | c8 = AVERAGE(c8, cC); |
| 3835 | c0 = AVERAGE(c0, c8); |
| 3836 | |
| 3837 | *(unsigned int*)(source0 + 4 * x) = c0; |
| 3838 | } |
| 3839 | |
| 3840 | source0 += pitch; |
| 3841 | source1 += pitch; |
| 3842 | source2 += pitch; |
| 3843 | source3 += pitch; |
| 3844 | source4 += pitch; |
| 3845 | source5 += pitch; |
| 3846 | source6 += pitch; |
| 3847 | source7 += pitch; |
| 3848 | source8 += pitch; |
| 3849 | source9 += pitch; |
| 3850 | sourceA += pitch; |
| 3851 | sourceB += pitch; |
| 3852 | sourceC += pitch; |
| 3853 | sourceD += pitch; |
| 3854 | sourceE += pitch; |
| 3855 | sourceF += pitch; |
| 3856 | } |
| 3857 | } |
| 3858 | else ASSERT(false); |
| 3859 | |
| 3860 | #undef AVERAGE |
| 3861 | } |
| 3862 | } |
| 3863 | else if(internal.format == FORMAT_R32F) |
| 3864 | { |
| 3865 | if(CPUID::supportsSSE() && (width % 4) == 0) |
| 3866 | { |
| 3867 | if(internal.depth == 2) |
| 3868 | { |
| 3869 | for(int y = 0; y < height; y++) |
| 3870 | { |
| 3871 | for(int x = 0; x < width; x += 4) |
| 3872 | { |
| 3873 | __m128 c0 = _mm_load_ps((float*)(source0 + 4 * x)); |
| 3874 | __m128 c1 = _mm_load_ps((float*)(source1 + 4 * x)); |
| 3875 | |
| 3876 | c0 = _mm_add_ps(c0, c1); |
| 3877 | c0 = _mm_mul_ps(c0, _mm_set1_ps(1.0f / 2.0f)); |
| 3878 | |
| 3879 | _mm_store_ps((float*)(source0 + 4 * x), c0); |
| 3880 | } |
| 3881 | |
| 3882 | source0 += pitch; |
| 3883 | source1 += pitch; |
| 3884 | } |
| 3885 | } |
| 3886 | else if(internal.depth == 4) |
| 3887 | { |
| 3888 | for(int y = 0; y < height; y++) |
| 3889 | { |
| 3890 | for(int x = 0; x < width; x += 4) |
| 3891 | { |
| 3892 | __m128 c0 = _mm_load_ps((float*)(source0 + 4 * x)); |
| 3893 | __m128 c1 = _mm_load_ps((float*)(source1 + 4 * x)); |
| 3894 | __m128 c2 = _mm_load_ps((float*)(source2 + 4 * x)); |
| 3895 | __m128 c3 = _mm_load_ps((float*)(source3 + 4 * x)); |
| 3896 | |
| 3897 | c0 = _mm_add_ps(c0, c1); |
| 3898 | c2 = _mm_add_ps(c2, c3); |
| 3899 | c0 = _mm_add_ps(c0, c2); |
| 3900 | c0 = _mm_mul_ps(c0, _mm_set1_ps(1.0f / 4.0f)); |
| 3901 | |
| 3902 | _mm_store_ps((float*)(source0 + 4 * x), c0); |
| 3903 | } |
| 3904 | |
| 3905 | source0 += pitch; |
| 3906 | source1 += pitch; |
| 3907 | source2 += pitch; |
| 3908 | source3 += pitch; |
| 3909 | } |
| 3910 | } |
| 3911 | else if(internal.depth == 8) |
| 3912 | { |
| 3913 | for(int y = 0; y < height; y++) |
| 3914 | { |
| 3915 | for(int x = 0; x < width; x += 4) |
| 3916 | { |
| 3917 | __m128 c0 = _mm_load_ps((float*)(source0 + 4 * x)); |
| 3918 | __m128 c1 = _mm_load_ps((float*)(source1 + 4 * x)); |
| 3919 | __m128 c2 = _mm_load_ps((float*)(source2 + 4 * x)); |
| 3920 | __m128 c3 = _mm_load_ps((float*)(source3 + 4 * x)); |
| 3921 | __m128 c4 = _mm_load_ps((float*)(source4 + 4 * x)); |
| 3922 | __m128 c5 = _mm_load_ps((float*)(source5 + 4 * x)); |
| 3923 | __m128 c6 = _mm_load_ps((float*)(source6 + 4 * x)); |
| 3924 | __m128 c7 = _mm_load_ps((float*)(source7 + 4 * x)); |
| 3925 | |
| 3926 | c0 = _mm_add_ps(c0, c1); |
| 3927 | c2 = _mm_add_ps(c2, c3); |
| 3928 | c4 = _mm_add_ps(c4, c5); |
| 3929 | c6 = _mm_add_ps(c6, c7); |
| 3930 | c0 = _mm_add_ps(c0, c2); |
| 3931 | c4 = _mm_add_ps(c4, c6); |
| 3932 | c0 = _mm_add_ps(c0, c4); |
| 3933 | c0 = _mm_mul_ps(c0, _mm_set1_ps(1.0f / 8.0f)); |
| 3934 | |
| 3935 | _mm_store_ps((float*)(source0 + 4 * x), c0); |
| 3936 | } |
| 3937 | |
| 3938 | source0 += pitch; |
| 3939 | source1 += pitch; |
| 3940 | source2 += pitch; |
| 3941 | source3 += pitch; |
| 3942 | source4 += pitch; |
| 3943 | source5 += pitch; |
| 3944 | source6 += pitch; |
| 3945 | source7 += pitch; |
| 3946 | } |
| 3947 | } |
| 3948 | else if(internal.depth == 16) |
| 3949 | { |
| 3950 | for(int y = 0; y < height; y++) |
| 3951 | { |
| 3952 | for(int x = 0; x < width; x += 4) |
| 3953 | { |
| 3954 | __m128 c0 = _mm_load_ps((float*)(source0 + 4 * x)); |
| 3955 | __m128 c1 = _mm_load_ps((float*)(source1 + 4 * x)); |
| 3956 | __m128 c2 = _mm_load_ps((float*)(source2 + 4 * x)); |
| 3957 | __m128 c3 = _mm_load_ps((float*)(source3 + 4 * x)); |
| 3958 | __m128 c4 = _mm_load_ps((float*)(source4 + 4 * x)); |
| 3959 | __m128 c5 = _mm_load_ps((float*)(source5 + 4 * x)); |
| 3960 | __m128 c6 = _mm_load_ps((float*)(source6 + 4 * x)); |
| 3961 | __m128 c7 = _mm_load_ps((float*)(source7 + 4 * x)); |
| 3962 | __m128 c8 = _mm_load_ps((float*)(source8 + 4 * x)); |
| 3963 | __m128 c9 = _mm_load_ps((float*)(source9 + 4 * x)); |
| 3964 | __m128 cA = _mm_load_ps((float*)(sourceA + 4 * x)); |
| 3965 | __m128 cB = _mm_load_ps((float*)(sourceB + 4 * x)); |
| 3966 | __m128 cC = _mm_load_ps((float*)(sourceC + 4 * x)); |
| 3967 | __m128 cD = _mm_load_ps((float*)(sourceD + 4 * x)); |
| 3968 | __m128 cE = _mm_load_ps((float*)(sourceE + 4 * x)); |
| 3969 | __m128 cF = _mm_load_ps((float*)(sourceF + 4 * x)); |
| 3970 | |
| 3971 | c0 = _mm_add_ps(c0, c1); |
| 3972 | c2 = _mm_add_ps(c2, c3); |
| 3973 | c4 = _mm_add_ps(c4, c5); |
| 3974 | c6 = _mm_add_ps(c6, c7); |
| 3975 | c8 = _mm_add_ps(c8, c9); |
| 3976 | cA = _mm_add_ps(cA, cB); |
| 3977 | cC = _mm_add_ps(cC, cD); |
| 3978 | cE = _mm_add_ps(cE, cF); |
| 3979 | c0 = _mm_add_ps(c0, c2); |
| 3980 | c4 = _mm_add_ps(c4, c6); |
| 3981 | c8 = _mm_add_ps(c8, cA); |
| 3982 | cC = _mm_add_ps(cC, cE); |
| 3983 | c0 = _mm_add_ps(c0, c4); |
| 3984 | c8 = _mm_add_ps(c8, cC); |
| 3985 | c0 = _mm_add_ps(c0, c8); |
| 3986 | c0 = _mm_mul_ps(c0, _mm_set1_ps(1.0f / 16.0f)); |
| 3987 | |
| 3988 | _mm_store_ps((float*)(source0 + 4 * x), c0); |
| 3989 | } |
| 3990 | |
| 3991 | source0 += pitch; |
| 3992 | source1 += pitch; |
| 3993 | source2 += pitch; |
| 3994 | source3 += pitch; |
| 3995 | source4 += pitch; |
| 3996 | source5 += pitch; |
| 3997 | source6 += pitch; |
| 3998 | source7 += pitch; |
| 3999 | source8 += pitch; |
| 4000 | source9 += pitch; |
| 4001 | sourceA += pitch; |
| 4002 | sourceB += pitch; |
| 4003 | sourceC += pitch; |
| 4004 | sourceD += pitch; |
| 4005 | sourceE += pitch; |
| 4006 | sourceF += pitch; |
| 4007 | } |
| 4008 | } |
| 4009 | else ASSERT(false); |
| 4010 | } |
| 4011 | else |
| 4012 | { |
| 4013 | if(internal.depth == 2) |
| 4014 | { |
| 4015 | for(int y = 0; y < height; y++) |
| 4016 | { |
| 4017 | for(int x = 0; x < width; x++) |
| 4018 | { |
| 4019 | float c0 = *(float*)(source0 + 4 * x); |
| 4020 | float c1 = *(float*)(source1 + 4 * x); |
| 4021 | |
| 4022 | c0 = c0 + c1; |
| 4023 | c0 *= 1.0f / 2.0f; |
| 4024 | |
| 4025 | *(float*)(source0 + 4 * x) = c0; |
| 4026 | } |
| 4027 | |
| 4028 | source0 += pitch; |
| 4029 | source1 += pitch; |
| 4030 | } |
| 4031 | } |
| 4032 | else if(internal.depth == 4) |
| 4033 | { |
| 4034 | for(int y = 0; y < height; y++) |
| 4035 | { |
| 4036 | for(int x = 0; x < width; x++) |
| 4037 | { |
| 4038 | float c0 = *(float*)(source0 + 4 * x); |
| 4039 | float c1 = *(float*)(source1 + 4 * x); |
| 4040 | float c2 = *(float*)(source2 + 4 * x); |
| 4041 | float c3 = *(float*)(source3 + 4 * x); |
| 4042 | |
| 4043 | c0 = c0 + c1; |
| 4044 | c2 = c2 + c3; |
| 4045 | c0 = c0 + c2; |
| 4046 | c0 *= 1.0f / 4.0f; |
| 4047 | |
| 4048 | *(float*)(source0 + 4 * x) = c0; |
| 4049 | } |
| 4050 | |
| 4051 | source0 += pitch; |
| 4052 | source1 += pitch; |
| 4053 | source2 += pitch; |
| 4054 | source3 += pitch; |
| 4055 | } |
| 4056 | } |
| 4057 | else if(internal.depth == 8) |
| 4058 | { |
| 4059 | for(int y = 0; y < height; y++) |
| 4060 | { |
| 4061 | for(int x = 0; x < width; x++) |
| 4062 | { |
| 4063 | float c0 = *(float*)(source0 + 4 * x); |
| 4064 | float c1 = *(float*)(source1 + 4 * x); |
| 4065 | float c2 = *(float*)(source2 + 4 * x); |
| 4066 | float c3 = *(float*)(source3 + 4 * x); |
| 4067 | float c4 = *(float*)(source4 + 4 * x); |
| 4068 | float c5 = *(float*)(source5 + 4 * x); |
| 4069 | float c6 = *(float*)(source6 + 4 * x); |
| 4070 | float c7 = *(float*)(source7 + 4 * x); |
| 4071 | |
| 4072 | c0 = c0 + c1; |
| 4073 | c2 = c2 + c3; |
| 4074 | c4 = c4 + c5; |
| 4075 | c6 = c6 + c7; |
| 4076 | c0 = c0 + c2; |
| 4077 | c4 = c4 + c6; |
| 4078 | c0 = c0 + c4; |
| 4079 | c0 *= 1.0f / 8.0f; |
| 4080 | |
| 4081 | *(float*)(source0 + 4 * x) = c0; |
| 4082 | } |
| 4083 | |
| 4084 | source0 += pitch; |
| 4085 | source1 += pitch; |
| 4086 | source2 += pitch; |
| 4087 | source3 += pitch; |
| 4088 | source4 += pitch; |
| 4089 | source5 += pitch; |
| 4090 | source6 += pitch; |
| 4091 | source7 += pitch; |
| 4092 | } |
| 4093 | } |
| 4094 | else if(internal.depth == 16) |
| 4095 | { |
| 4096 | for(int y = 0; y < height; y++) |
| 4097 | { |
| 4098 | for(int x = 0; x < width; x++) |
| 4099 | { |
| 4100 | float c0 = *(float*)(source0 + 4 * x); |
| 4101 | float c1 = *(float*)(source1 + 4 * x); |
| 4102 | float c2 = *(float*)(source2 + 4 * x); |
| 4103 | float c3 = *(float*)(source3 + 4 * x); |
| 4104 | float c4 = *(float*)(source4 + 4 * x); |
| 4105 | float c5 = *(float*)(source5 + 4 * x); |
| 4106 | float c6 = *(float*)(source6 + 4 * x); |
| 4107 | float c7 = *(float*)(source7 + 4 * x); |
| 4108 | float c8 = *(float*)(source8 + 4 * x); |
| 4109 | float c9 = *(float*)(source9 + 4 * x); |
| 4110 | float cA = *(float*)(sourceA + 4 * x); |
| 4111 | float cB = *(float*)(sourceB + 4 * x); |
| 4112 | float cC = *(float*)(sourceC + 4 * x); |
| 4113 | float cD = *(float*)(sourceD + 4 * x); |
| 4114 | float cE = *(float*)(sourceE + 4 * x); |
| 4115 | float cF = *(float*)(sourceF + 4 * x); |
| 4116 | |
| 4117 | c0 = c0 + c1; |
| 4118 | c2 = c2 + c3; |
| 4119 | c4 = c4 + c5; |
| 4120 | c6 = c6 + c7; |
| 4121 | c8 = c8 + c9; |
| 4122 | cA = cA + cB; |
| 4123 | cC = cC + cD; |
| 4124 | cE = cE + cF; |
| 4125 | c0 = c0 + c2; |
| 4126 | c4 = c4 + c6; |
| 4127 | c8 = c8 + cA; |
| 4128 | cC = cC + cE; |
| 4129 | c0 = c0 + c4; |
| 4130 | c8 = c8 + cC; |
| 4131 | c0 = c0 + c8; |
| 4132 | c0 *= 1.0f / 16.0f; |
| 4133 | |
| 4134 | *(float*)(source0 + 4 * x) = c0; |
| 4135 | } |
| 4136 | |
| 4137 | source0 += pitch; |
| 4138 | source1 += pitch; |
| 4139 | source2 += pitch; |
| 4140 | source3 += pitch; |
| 4141 | source4 += pitch; |
| 4142 | source5 += pitch; |
| 4143 | source6 += pitch; |
| 4144 | source7 += pitch; |
| 4145 | source8 += pitch; |
| 4146 | source9 += pitch; |
| 4147 | sourceA += pitch; |
| 4148 | sourceB += pitch; |
| 4149 | sourceC += pitch; |
| 4150 | sourceD += pitch; |
| 4151 | sourceE += pitch; |
| 4152 | sourceF += pitch; |
| 4153 | } |
| 4154 | } |
| 4155 | else ASSERT(false); |
| 4156 | } |
| 4157 | } |
| 4158 | else if(internal.format == FORMAT_G32R32F) |
| 4159 | { |
| 4160 | if(CPUID::supportsSSE() && (width % 2) == 0) |
| 4161 | { |
| 4162 | if(internal.depth == 2) |
| 4163 | { |
| 4164 | for(int y = 0; y < height; y++) |
| 4165 | { |
| 4166 | for(int x = 0; x < width; x += 2) |
| 4167 | { |
| 4168 | __m128 c0 = _mm_load_ps((float*)(source0 + 8 * x)); |
| 4169 | __m128 c1 = _mm_load_ps((float*)(source1 + 8 * x)); |
| 4170 | |
| 4171 | c0 = _mm_add_ps(c0, c1); |
| 4172 | c0 = _mm_mul_ps(c0, _mm_set1_ps(1.0f / 2.0f)); |
| 4173 | |
| 4174 | _mm_store_ps((float*)(source0 + 8 * x), c0); |
| 4175 | } |
| 4176 | |
| 4177 | source0 += pitch; |
| 4178 | source1 += pitch; |
| 4179 | } |
| 4180 | } |
| 4181 | else if(internal.depth == 4) |
| 4182 | { |
| 4183 | for(int y = 0; y < height; y++) |
| 4184 | { |
| 4185 | for(int x = 0; x < width; x += 2) |
| 4186 | { |
| 4187 | __m128 c0 = _mm_load_ps((float*)(source0 + 8 * x)); |
| 4188 | __m128 c1 = _mm_load_ps((float*)(source1 + 8 * x)); |
| 4189 | __m128 c2 = _mm_load_ps((float*)(source2 + 8 * x)); |
| 4190 | __m128 c3 = _mm_load_ps((float*)(source3 + 8 * x)); |
| 4191 | |
| 4192 | c0 = _mm_add_ps(c0, c1); |
| 4193 | c2 = _mm_add_ps(c2, c3); |
| 4194 | c0 = _mm_add_ps(c0, c2); |
| 4195 | c0 = _mm_mul_ps(c0, _mm_set1_ps(1.0f / 4.0f)); |
| 4196 | |
| 4197 | _mm_store_ps((float*)(source0 + 8 * x), c0); |
| 4198 | } |
| 4199 | |
| 4200 | source0 += pitch; |
| 4201 | source1 += pitch; |
| 4202 | source2 += pitch; |
| 4203 | source3 += pitch; |
| 4204 | } |
| 4205 | } |
| 4206 | else if(internal.depth == 8) |
| 4207 | { |
| 4208 | for(int y = 0; y < height; y++) |
| 4209 | { |
| 4210 | for(int x = 0; x < width; x += 2) |
| 4211 | { |
| 4212 | __m128 c0 = _mm_load_ps((float*)(source0 + 8 * x)); |
| 4213 | __m128 c1 = _mm_load_ps((float*)(source1 + 8 * x)); |
| 4214 | __m128 c2 = _mm_load_ps((float*)(source2 + 8 * x)); |
| 4215 | __m128 c3 = _mm_load_ps((float*)(source3 + 8 * x)); |
| 4216 | __m128 c4 = _mm_load_ps((float*)(source4 + 8 * x)); |
| 4217 | __m128 c5 = _mm_load_ps((float*)(source5 + 8 * x)); |
| 4218 | __m128 c6 = _mm_load_ps((float*)(source6 + 8 * x)); |
| 4219 | __m128 c7 = _mm_load_ps((float*)(source7 + 8 * x)); |
| 4220 | |
| 4221 | c0 = _mm_add_ps(c0, c1); |
| 4222 | c2 = _mm_add_ps(c2, c3); |
| 4223 | c4 = _mm_add_ps(c4, c5); |
| 4224 | c6 = _mm_add_ps(c6, c7); |
| 4225 | c0 = _mm_add_ps(c0, c2); |
| 4226 | c4 = _mm_add_ps(c4, c6); |
| 4227 | c0 = _mm_add_ps(c0, c4); |
| 4228 | c0 = _mm_mul_ps(c0, _mm_set1_ps(1.0f / 8.0f)); |
| 4229 | |
| 4230 | _mm_store_ps((float*)(source0 + 8 * x), c0); |
| 4231 | } |
| 4232 | |
| 4233 | source0 += pitch; |
| 4234 | source1 += pitch; |
| 4235 | source2 += pitch; |
| 4236 | source3 += pitch; |
| 4237 | source4 += pitch; |
| 4238 | source5 += pitch; |
| 4239 | source6 += pitch; |
| 4240 | source7 += pitch; |
| 4241 | } |
| 4242 | } |
| 4243 | else if(internal.depth == 16) |
| 4244 | { |
| 4245 | for(int y = 0; y < height; y++) |
| 4246 | { |
| 4247 | for(int x = 0; x < width; x += 2) |
| 4248 | { |
| 4249 | __m128 c0 = _mm_load_ps((float*)(source0 + 8 * x)); |
| 4250 | __m128 c1 = _mm_load_ps((float*)(source1 + 8 * x)); |
| 4251 | __m128 c2 = _mm_load_ps((float*)(source2 + 8 * x)); |
| 4252 | __m128 c3 = _mm_load_ps((float*)(source3 + 8 * x)); |
| 4253 | __m128 c4 = _mm_load_ps((float*)(source4 + 8 * x)); |
| 4254 | __m128 c5 = _mm_load_ps((float*)(source5 + 8 * x)); |
| 4255 | __m128 c6 = _mm_load_ps((float*)(source6 + 8 * x)); |
| 4256 | __m128 c7 = _mm_load_ps((float*)(source7 + 8 * x)); |
| 4257 | __m128 c8 = _mm_load_ps((float*)(source8 + 8 * x)); |
| 4258 | __m128 c9 = _mm_load_ps((float*)(source9 + 8 * x)); |
| 4259 | __m128 cA = _mm_load_ps((float*)(sourceA + 8 * x)); |
| 4260 | __m128 cB = _mm_load_ps((float*)(sourceB + 8 * x)); |
| 4261 | __m128 cC = _mm_load_ps((float*)(sourceC + 8 * x)); |
| 4262 | __m128 cD = _mm_load_ps((float*)(sourceD + 8 * x)); |
| 4263 | __m128 cE = _mm_load_ps((float*)(sourceE + 8 * x)); |
| 4264 | __m128 cF = _mm_load_ps((float*)(sourceF + 8 * x)); |
| 4265 | |
| 4266 | c0 = _mm_add_ps(c0, c1); |
| 4267 | c2 = _mm_add_ps(c2, c3); |
| 4268 | c4 = _mm_add_ps(c4, c5); |
| 4269 | c6 = _mm_add_ps(c6, c7); |
| 4270 | c8 = _mm_add_ps(c8, c9); |
| 4271 | cA = _mm_add_ps(cA, cB); |
| 4272 | cC = _mm_add_ps(cC, cD); |
| 4273 | cE = _mm_add_ps(cE, cF); |
| 4274 | c0 = _mm_add_ps(c0, c2); |
| 4275 | c4 = _mm_add_ps(c4, c6); |
| 4276 | c8 = _mm_add_ps(c8, cA); |
| 4277 | cC = _mm_add_ps(cC, cE); |
| 4278 | c0 = _mm_add_ps(c0, c4); |
| 4279 | c8 = _mm_add_ps(c8, cC); |
| 4280 | c0 = _mm_add_ps(c0, c8); |
| 4281 | c0 = _mm_mul_ps(c0, _mm_set1_ps(1.0f / 16.0f)); |
| 4282 | |
| 4283 | _mm_store_ps((float*)(source0 + 8 * x), c0); |
| 4284 | } |
| 4285 | |
| 4286 | source0 += pitch; |
| 4287 | source1 += pitch; |
| 4288 | source2 += pitch; |
| 4289 | source3 += pitch; |
| 4290 | source4 += pitch; |
| 4291 | source5 += pitch; |
| 4292 | source6 += pitch; |
| 4293 | source7 += pitch; |
| 4294 | source8 += pitch; |
| 4295 | source9 += pitch; |
| 4296 | sourceA += pitch; |
| 4297 | sourceB += pitch; |
| 4298 | sourceC += pitch; |
| 4299 | sourceD += pitch; |
| 4300 | sourceE += pitch; |
| 4301 | sourceF += pitch; |
| 4302 | } |
| 4303 | } |
| 4304 | else ASSERT(false); |
| 4305 | } |
| 4306 | else |
| 4307 | { |
| 4308 | if(internal.depth == 2) |
| 4309 | { |
| 4310 | for(int y = 0; y < height; y++) |
| 4311 | { |
| 4312 | for(int x = 0; x < 2 * width; x++) |
| 4313 | { |
| 4314 | float c0 = *(float*)(source0 + 4 * x); |
| 4315 | float c1 = *(float*)(source1 + 4 * x); |
| 4316 | |
| 4317 | c0 = c0 + c1; |
| 4318 | c0 *= 1.0f / 2.0f; |
| 4319 | |
| 4320 | *(float*)(source0 + 4 * x) = c0; |
| 4321 | } |
| 4322 | |
| 4323 | source0 += pitch; |
| 4324 | source1 += pitch; |
| 4325 | } |
| 4326 | } |
| 4327 | else if(internal.depth == 4) |
| 4328 | { |
| 4329 | for(int y = 0; y < height; y++) |
| 4330 | { |
| 4331 | for(int x = 0; x < 2 * width; x++) |
| 4332 | { |
| 4333 | float c0 = *(float*)(source0 + 4 * x); |
| 4334 | float c1 = *(float*)(source1 + 4 * x); |
| 4335 | float c2 = *(float*)(source2 + 4 * x); |
| 4336 | float c3 = *(float*)(source3 + 4 * x); |
| 4337 | |
| 4338 | c0 = c0 + c1; |
| 4339 | c2 = c2 + c3; |
| 4340 | c0 = c0 + c2; |
| 4341 | c0 *= 1.0f / 4.0f; |
| 4342 | |
| 4343 | *(float*)(source0 + 4 * x) = c0; |
| 4344 | } |
| 4345 | |
| 4346 | source0 += pitch; |
| 4347 | source1 += pitch; |
| 4348 | source2 += pitch; |
| 4349 | source3 += pitch; |
| 4350 | } |
| 4351 | } |
| 4352 | else if(internal.depth == 8) |
| 4353 | { |
| 4354 | for(int y = 0; y < height; y++) |
| 4355 | { |
| 4356 | for(int x = 0; x < 2 * width; x++) |
| 4357 | { |
| 4358 | float c0 = *(float*)(source0 + 4 * x); |
| 4359 | float c1 = *(float*)(source1 + 4 * x); |
| 4360 | float c2 = *(float*)(source2 + 4 * x); |
| 4361 | float c3 = *(float*)(source3 + 4 * x); |
| 4362 | float c4 = *(float*)(source4 + 4 * x); |
| 4363 | float c5 = *(float*)(source5 + 4 * x); |
| 4364 | float c6 = *(float*)(source6 + 4 * x); |
| 4365 | float c7 = *(float*)(source7 + 4 * x); |
| 4366 | |
| 4367 | c0 = c0 + c1; |
| 4368 | c2 = c2 + c3; |
| 4369 | c4 = c4 + c5; |
| 4370 | c6 = c6 + c7; |
| 4371 | c0 = c0 + c2; |
| 4372 | c4 = c4 + c6; |
| 4373 | c0 = c0 + c4; |
| 4374 | c0 *= 1.0f / 8.0f; |
| 4375 | |
| 4376 | *(float*)(source0 + 4 * x) = c0; |
| 4377 | } |
| 4378 | |
| 4379 | source0 += pitch; |
| 4380 | source1 += pitch; |
| 4381 | source2 += pitch; |
| 4382 | source3 += pitch; |
| 4383 | source4 += pitch; |
| 4384 | source5 += pitch; |
| 4385 | source6 += pitch; |
| 4386 | source7 += pitch; |
| 4387 | } |
| 4388 | } |
| 4389 | else if(internal.depth == 16) |
| 4390 | { |
| 4391 | for(int y = 0; y < height; y++) |
| 4392 | { |
| 4393 | for(int x = 0; x < 2 * width; x++) |
| 4394 | { |
| 4395 | float c0 = *(float*)(source0 + 4 * x); |
| 4396 | float c1 = *(float*)(source1 + 4 * x); |
| 4397 | float c2 = *(float*)(source2 + 4 * x); |
| 4398 | float c3 = *(float*)(source3 + 4 * x); |
| 4399 | float c4 = *(float*)(source4 + 4 * x); |
| 4400 | float c5 = *(float*)(source5 + 4 * x); |
| 4401 | float c6 = *(float*)(source6 + 4 * x); |
| 4402 | float c7 = *(float*)(source7 + 4 * x); |
| 4403 | float c8 = *(float*)(source8 + 4 * x); |
| 4404 | float c9 = *(float*)(source9 + 4 * x); |
| 4405 | float cA = *(float*)(sourceA + 4 * x); |
| 4406 | float cB = *(float*)(sourceB + 4 * x); |
| 4407 | float cC = *(float*)(sourceC + 4 * x); |
| 4408 | float cD = *(float*)(sourceD + 4 * x); |
| 4409 | float cE = *(float*)(sourceE + 4 * x); |
| 4410 | float cF = *(float*)(sourceF + 4 * x); |
| 4411 | |
| 4412 | c0 = c0 + c1; |
| 4413 | c2 = c2 + c3; |
| 4414 | c4 = c4 + c5; |
| 4415 | c6 = c6 + c7; |
| 4416 | c8 = c8 + c9; |
| 4417 | cA = cA + cB; |
| 4418 | cC = cC + cD; |
| 4419 | cE = cE + cF; |
| 4420 | c0 = c0 + c2; |
| 4421 | c4 = c4 + c6; |
| 4422 | c8 = c8 + cA; |
| 4423 | cC = cC + cE; |
| 4424 | c0 = c0 + c4; |
| 4425 | c8 = c8 + cC; |
| 4426 | c0 = c0 + c8; |
| 4427 | c0 *= 1.0f / 16.0f; |
| 4428 | |
| 4429 | *(float*)(source0 + 4 * x) = c0; |
| 4430 | } |
| 4431 | |
| 4432 | source0 += pitch; |
| 4433 | source1 += pitch; |
| 4434 | source2 += pitch; |
| 4435 | source3 += pitch; |
| 4436 | source4 += pitch; |
| 4437 | source5 += pitch; |
| 4438 | source6 += pitch; |
| 4439 | source7 += pitch; |
| 4440 | source8 += pitch; |
| 4441 | source9 += pitch; |
| 4442 | sourceA += pitch; |
| 4443 | sourceB += pitch; |
| 4444 | sourceC += pitch; |
| 4445 | sourceD += pitch; |
| 4446 | sourceE += pitch; |
| 4447 | sourceF += pitch; |
| 4448 | } |
| 4449 | } |
| 4450 | else ASSERT(false); |
| 4451 | } |
| 4452 | } |
| 4453 | else if(internal.format == FORMAT_A32B32G32R32F) |
| 4454 | { |
| 4455 | if(CPUID::supportsSSE()) |
| 4456 | { |
| 4457 | if(internal.depth == 2) |
| 4458 | { |
| 4459 | for(int y = 0; y < height; y++) |
| 4460 | { |
| 4461 | for(int x = 0; x < width; x++) |
| 4462 | { |
| 4463 | __m128 c0 = _mm_load_ps((float*)(source0 + 16 * x)); |
| 4464 | __m128 c1 = _mm_load_ps((float*)(source1 + 16 * x)); |
| 4465 | |
| 4466 | c0 = _mm_add_ps(c0, c1); |
| 4467 | c0 = _mm_mul_ps(c0, _mm_set1_ps(1.0f / 2.0f)); |
| 4468 | |
| 4469 | _mm_store_ps((float*)(source0 + 16 * x), c0); |
| 4470 | } |
| 4471 | |
| 4472 | source0 += pitch; |
| 4473 | source1 += pitch; |
| 4474 | } |
| 4475 | } |
| 4476 | else if(internal.depth == 4) |
| 4477 | { |
| 4478 | for(int y = 0; y < height; y++) |
| 4479 | { |
| 4480 | for(int x = 0; x < width; x++) |
| 4481 | { |
| 4482 | __m128 c0 = _mm_load_ps((float*)(source0 + 16 * x)); |
| 4483 | __m128 c1 = _mm_load_ps((float*)(source1 + 16 * x)); |
| 4484 | __m128 c2 = _mm_load_ps((float*)(source2 + 16 * x)); |
| 4485 | __m128 c3 = _mm_load_ps((float*)(source3 + 16 * x)); |
| 4486 | |
| 4487 | c0 = _mm_add_ps(c0, c1); |
| 4488 | c2 = _mm_add_ps(c2, c3); |
| 4489 | c0 = _mm_add_ps(c0, c2); |
| 4490 | c0 = _mm_mul_ps(c0, _mm_set1_ps(1.0f / 4.0f)); |
| 4491 | |
| 4492 | _mm_store_ps((float*)(source0 + 16 * x), c0); |
| 4493 | } |
| 4494 | |
| 4495 | source0 += pitch; |
| 4496 | source1 += pitch; |
| 4497 | source2 += pitch; |
| 4498 | source3 += pitch; |
| 4499 | } |
| 4500 | } |
| 4501 | else if(internal.depth == 8) |
| 4502 | { |
| 4503 | for(int y = 0; y < height; y++) |
| 4504 | { |
| 4505 | for(int x = 0; x < width; x++) |
| 4506 | { |
| 4507 | __m128 c0 = _mm_load_ps((float*)(source0 + 16 * x)); |
| 4508 | __m128 c1 = _mm_load_ps((float*)(source1 + 16 * x)); |
| 4509 | __m128 c2 = _mm_load_ps((float*)(source2 + 16 * x)); |
| 4510 | __m128 c3 = _mm_load_ps((float*)(source3 + 16 * x)); |
| 4511 | __m128 c4 = _mm_load_ps((float*)(source4 + 16 * x)); |
| 4512 | __m128 c5 = _mm_load_ps((float*)(source5 + 16 * x)); |
| 4513 | __m128 c6 = _mm_load_ps((float*)(source6 + 16 * x)); |
| 4514 | __m128 c7 = _mm_load_ps((float*)(source7 + 16 * x)); |
| 4515 | |
| 4516 | c0 = _mm_add_ps(c0, c1); |
| 4517 | c2 = _mm_add_ps(c2, c3); |
| 4518 | c4 = _mm_add_ps(c4, c5); |
| 4519 | c6 = _mm_add_ps(c6, c7); |
| 4520 | c0 = _mm_add_ps(c0, c2); |
| 4521 | c4 = _mm_add_ps(c4, c6); |
| 4522 | c0 = _mm_add_ps(c0, c4); |
| 4523 | c0 = _mm_mul_ps(c0, _mm_set1_ps(1.0f / 8.0f)); |
| 4524 | |
| 4525 | _mm_store_ps((float*)(source0 + 16 * x), c0); |
| 4526 | } |
| 4527 | |
| 4528 | source0 += pitch; |
| 4529 | source1 += pitch; |
| 4530 | source2 += pitch; |
| 4531 | source3 += pitch; |
| 4532 | source4 += pitch; |
| 4533 | source5 += pitch; |
| 4534 | source6 += pitch; |
| 4535 | source7 += pitch; |
| 4536 | } |
| 4537 | } |
| 4538 | else if(internal.depth == 16) |
| 4539 | { |
| 4540 | for(int y = 0; y < height; y++) |
| 4541 | { |
| 4542 | for(int x = 0; x < width; x++) |
| 4543 | { |
| 4544 | __m128 c0 = _mm_load_ps((float*)(source0 + 16 * x)); |
| 4545 | __m128 c1 = _mm_load_ps((float*)(source1 + 16 * x)); |
| 4546 | __m128 c2 = _mm_load_ps((float*)(source2 + 16 * x)); |
| 4547 | __m128 c3 = _mm_load_ps((float*)(source3 + 16 * x)); |
| 4548 | __m128 c4 = _mm_load_ps((float*)(source4 + 16 * x)); |
| 4549 | __m128 c5 = _mm_load_ps((float*)(source5 + 16 * x)); |
| 4550 | __m128 c6 = _mm_load_ps((float*)(source6 + 16 * x)); |
| 4551 | __m128 c7 = _mm_load_ps((float*)(source7 + 16 * x)); |
| 4552 | __m128 c8 = _mm_load_ps((float*)(source8 + 16 * x)); |
| 4553 | __m128 c9 = _mm_load_ps((float*)(source9 + 16 * x)); |
| 4554 | __m128 cA = _mm_load_ps((float*)(sourceA + 16 * x)); |
| 4555 | __m128 cB = _mm_load_ps((float*)(sourceB + 16 * x)); |
| 4556 | __m128 cC = _mm_load_ps((float*)(sourceC + 16 * x)); |
| 4557 | __m128 cD = _mm_load_ps((float*)(sourceD + 16 * x)); |
| 4558 | __m128 cE = _mm_load_ps((float*)(sourceE + 16 * x)); |
| 4559 | __m128 cF = _mm_load_ps((float*)(sourceF + 16 * x)); |
| 4560 | |
| 4561 | c0 = _mm_add_ps(c0, c1); |
| 4562 | c2 = _mm_add_ps(c2, c3); |
| 4563 | c4 = _mm_add_ps(c4, c5); |
| 4564 | c6 = _mm_add_ps(c6, c7); |
| 4565 | c8 = _mm_add_ps(c8, c9); |
| 4566 | cA = _mm_add_ps(cA, cB); |
| 4567 | cC = _mm_add_ps(cC, cD); |
| 4568 | cE = _mm_add_ps(cE, cF); |
| 4569 | c0 = _mm_add_ps(c0, c2); |
| 4570 | c4 = _mm_add_ps(c4, c6); |
| 4571 | c8 = _mm_add_ps(c8, cA); |
| 4572 | cC = _mm_add_ps(cC, cE); |
| 4573 | c0 = _mm_add_ps(c0, c4); |
| 4574 | c8 = _mm_add_ps(c8, cC); |
| 4575 | c0 = _mm_add_ps(c0, c8); |
| 4576 | c0 = _mm_mul_ps(c0, _mm_set1_ps(1.0f / 16.0f)); |
| 4577 | |
| 4578 | _mm_store_ps((float*)(source0 + 16 * x), c0); |
| 4579 | } |
| 4580 | |
| 4581 | source0 += pitch; |
| 4582 | source1 += pitch; |
| 4583 | source2 += pitch; |
| 4584 | source3 += pitch; |
| 4585 | source4 += pitch; |
| 4586 | source5 += pitch; |
| 4587 | source6 += pitch; |
| 4588 | source7 += pitch; |
| 4589 | source8 += pitch; |
| 4590 | source9 += pitch; |
| 4591 | sourceA += pitch; |
| 4592 | sourceB += pitch; |
| 4593 | sourceC += pitch; |
| 4594 | sourceD += pitch; |
| 4595 | sourceE += pitch; |
| 4596 | sourceF += pitch; |
| 4597 | } |
| 4598 | } |
| 4599 | else ASSERT(false); |
| 4600 | } |
| 4601 | else |
| 4602 | { |
| 4603 | if(internal.depth == 2) |
| 4604 | { |
| 4605 | for(int y = 0; y < height; y++) |
| 4606 | { |
| 4607 | for(int x = 0; x < 4 * width; x++) |
| 4608 | { |
| 4609 | float c0 = *(float*)(source0 + 4 * x); |
| 4610 | float c1 = *(float*)(source1 + 4 * x); |
| 4611 | |
| 4612 | c0 = c0 + c1; |
| 4613 | c0 *= 1.0f / 2.0f; |
| 4614 | |
| 4615 | *(float*)(source0 + 4 * x) = c0; |
| 4616 | } |
| 4617 | |
| 4618 | source0 += pitch; |
| 4619 | source1 += pitch; |
| 4620 | } |
| 4621 | } |
| 4622 | else if(internal.depth == 4) |
| 4623 | { |
| 4624 | for(int y = 0; y < height; y++) |
| 4625 | { |
| 4626 | for(int x = 0; x < 4 * width; x++) |
| 4627 | { |
| 4628 | float c0 = *(float*)(source0 + 4 * x); |
| 4629 | float c1 = *(float*)(source1 + 4 * x); |
| 4630 | float c2 = *(float*)(source2 + 4 * x); |
| 4631 | float c3 = *(float*)(source3 + 4 * x); |
| 4632 | |
| 4633 | c0 = c0 + c1; |
| 4634 | c2 = c2 + c3; |
| 4635 | c0 = c0 + c2; |
| 4636 | c0 *= 1.0f / 4.0f; |
| 4637 | |
| 4638 | *(float*)(source0 + 4 * x) = c0; |
| 4639 | } |
| 4640 | |
| 4641 | source0 += pitch; |
| 4642 | source1 += pitch; |
| 4643 | source2 += pitch; |
| 4644 | source3 += pitch; |
| 4645 | } |
| 4646 | } |
| 4647 | else if(internal.depth == 8) |
| 4648 | { |
| 4649 | for(int y = 0; y < height; y++) |
| 4650 | { |
| 4651 | for(int x = 0; x < 4 * width; x++) |
| 4652 | { |
| 4653 | float c0 = *(float*)(source0 + 4 * x); |
| 4654 | float c1 = *(float*)(source1 + 4 * x); |
| 4655 | float c2 = *(float*)(source2 + 4 * x); |
| 4656 | float c3 = *(float*)(source3 + 4 * x); |
| 4657 | float c4 = *(float*)(source4 + 4 * x); |
| 4658 | float c5 = *(float*)(source5 + 4 * x); |
| 4659 | float c6 = *(float*)(source6 + 4 * x); |
| 4660 | float c7 = *(float*)(source7 + 4 * x); |
| 4661 | |
| 4662 | c0 = c0 + c1; |
| 4663 | c2 = c2 + c3; |
| 4664 | c4 = c4 + c5; |
| 4665 | c6 = c6 + c7; |
| 4666 | c0 = c0 + c2; |
| 4667 | c4 = c4 + c6; |
| 4668 | c0 = c0 + c4; |
| 4669 | c0 *= 1.0f / 8.0f; |
| 4670 | |
| 4671 | *(float*)(source0 + 4 * x) = c0; |
| 4672 | } |
| 4673 | |
| 4674 | source0 += pitch; |
| 4675 | source1 += pitch; |
| 4676 | source2 += pitch; |
| 4677 | source3 += pitch; |
| 4678 | source4 += pitch; |
| 4679 | source5 += pitch; |
| 4680 | source6 += pitch; |
| 4681 | source7 += pitch; |
| 4682 | } |
| 4683 | } |
| 4684 | else if(internal.depth == 16) |
| 4685 | { |
| 4686 | for(int y = 0; y < height; y++) |
| 4687 | { |
| 4688 | for(int x = 0; x < 4 * width; x++) |
| 4689 | { |
| 4690 | float c0 = *(float*)(source0 + 4 * x); |
| 4691 | float c1 = *(float*)(source1 + 4 * x); |
| 4692 | float c2 = *(float*)(source2 + 4 * x); |
| 4693 | float c3 = *(float*)(source3 + 4 * x); |
| 4694 | float c4 = *(float*)(source4 + 4 * x); |
| 4695 | float c5 = *(float*)(source5 + 4 * x); |
| 4696 | float c6 = *(float*)(source6 + 4 * x); |
| 4697 | float c7 = *(float*)(source7 + 4 * x); |
| 4698 | float c8 = *(float*)(source8 + 4 * x); |
| 4699 | float c9 = *(float*)(source9 + 4 * x); |
| 4700 | float cA = *(float*)(sourceA + 4 * x); |
| 4701 | float cB = *(float*)(sourceB + 4 * x); |
| 4702 | float cC = *(float*)(sourceC + 4 * x); |
| 4703 | float cD = *(float*)(sourceD + 4 * x); |
| 4704 | float cE = *(float*)(sourceE + 4 * x); |
| 4705 | float cF = *(float*)(sourceF + 4 * x); |
| 4706 | |
| 4707 | c0 = c0 + c1; |
| 4708 | c2 = c2 + c3; |
| 4709 | c4 = c4 + c5; |
| 4710 | c6 = c6 + c7; |
| 4711 | c8 = c8 + c9; |
| 4712 | cA = cA + cB; |
| 4713 | cC = cC + cD; |
| 4714 | cE = cE + cF; |
| 4715 | c0 = c0 + c2; |
| 4716 | c4 = c4 + c6; |
| 4717 | c8 = c8 + cA; |
| 4718 | cC = cC + cE; |
| 4719 | c0 = c0 + c4; |
| 4720 | c8 = c8 + cC; |
| 4721 | c0 = c0 + c8; |
| 4722 | c0 *= 1.0f / 16.0f; |
| 4723 | |
| 4724 | *(float*)(source0 + 4 * x) = c0; |
| 4725 | } |
| 4726 | |
| 4727 | source0 += pitch; |
| 4728 | source1 += pitch; |
| 4729 | source2 += pitch; |
| 4730 | source3 += pitch; |
| 4731 | source4 += pitch; |
| 4732 | source5 += pitch; |
| 4733 | source6 += pitch; |
| 4734 | source7 += pitch; |
| 4735 | source8 += pitch; |
| 4736 | source9 += pitch; |
| 4737 | sourceA += pitch; |
| 4738 | sourceB += pitch; |
| 4739 | sourceC += pitch; |
| 4740 | sourceD += pitch; |
| 4741 | sourceE += pitch; |
| 4742 | sourceF += pitch; |
| 4743 | } |
| 4744 | } |
| 4745 | else ASSERT(false); |
| 4746 | } |
| 4747 | } |
| 4748 | else |
| 4749 | { |
| 4750 | // UNIMPLEMENTED(); |
| 4751 | } |
| 4752 | } |
| 4753 | } |