dom/canvas/WebGLTexelConversions.h

   1 /*
   2  * Copyright (C) 2010 Apple Inc. All rights reserved.
   3  * Copyright (C) 2010 Google Inc. All rights reserved.
   4  * Copyright (C) 2010 Mozilla Corporation. All rights reserved.
   5  *
   6  * Redistribution and use in source and binary forms, with or without
   7  * modification, are permitted provided that the following conditions
   8  * are met:
   9  * 1. Redistributions of source code must retain the above copyright
  10  *    notice, this list of conditions and the following disclaimer.
  11  * 2. Redistributions in binary form must reproduce the above copyright
  12  *    notice, this list of conditions and the following disclaimer in the
  13  *    documentation and/or other materials provided with the distribution.
  14  *
  15  * THIS SOFTWARE IS PROVIDED BY APPLE COMPUTER, INC. ``AS IS'' AND ANY
  16  * EXPRESS OR IMPLIED WARRANTIES, INCLUDING, BUT NOT LIMITED TO, THE
  17  * IMPLIED WARRANTIES OF MERCHANTABILITY AND FITNESS FOR A PARTICULAR
  18  * PURPOSE ARE DISCLAIMED.  IN NO EVENT SHALL APPLE COMPUTER, INC. OR
  19  * CONTRIBUTORS BE LIABLE FOR ANY DIRECT, INDIRECT, INCIDENTAL, SPECIAL,
  20  * EXEMPLARY, OR CONSEQUENTIAL DAMAGES (INCLUDING, BUT NOT LIMITED TO,
  21  * PROCUREMENT OF SUBSTITUTE GOODS OR SERVICES; LOSS OF USE, DATA, OR
  22  * PROFITS; OR BUSINESS INTERRUPTION) HOWEVER CAUSED AND ON ANY THEORY
  23  * OF LIABILITY, WHETHER IN CONTRACT, STRICT LIABILITY, OR TORT
  24  * (INCLUDING NEGLIGENCE OR OTHERWISE) ARISING IN ANY WAY OUT OF THE USE
  25  * OF THIS SOFTWARE, EVEN IF ADVISED OF THE POSSIBILITY OF SUCH DAMAGE.
  26  */
  27
  28 #ifndef WEBGLTEXELCONVERSIONS_H_
  29 #define WEBGLTEXELCONVERSIONS_H_
  30
  31 #ifdef __SUNPRO_CC
  32 #  define __restrict
  33 #endif
  34
  35 #include "WebGLTypes.h"
  36 #include <stdint.h>
  37 #include "mozilla/Attributes.h"
  38 #include "mozilla/Casting.h"
  39
  40 namespace mozilla {
  41
  42 namespace gl {
  43 enum class OriginPos : uint8_t;
  44 }
  45
  46 bool ConvertImage(size_t width, size_t height, const void* srcBegin,
  47                   size_t srcStride, gl::OriginPos srcOrigin,
  48                   WebGLTexelFormat srcFormat, bool srcPremultiplied,
  49                   void* dstBegin, size_t dstStride, gl::OriginPos dstOrigin,
  50                   WebGLTexelFormat dstFormat, bool dstPremultiplied,
  51                   bool* out_wasTrivial);
  52
  53 //////////////////////////////////////////////////////////////////////////////////////////
  54
  55 // single precision float
  56 // seeeeeeeemmmmmmmmmmmmmmmmmmmmmmm
  57
  58 // half precision float
  59 // seeeeemmmmmmmmmm
  60
  61 // IEEE 16bits floating point:
  62 const uint16_t kFloat16Value_Zero = 0x0000;      // = 0000000000000000b
  63 const uint16_t kFloat16Value_One = 0x3C00;       // = 0011110000000000b
  64 const uint16_t kFloat16Value_Infinity = 0x7C00;  // = 0111110000000000b
  65 const uint16_t kFloat16Value_NaN = 0x7FFF;  // = 011111yyyyyyyyyyb (nonzero y)
  66
  67 MOZ_ALWAYS_INLINE uint16_t packToFloat16(float v) {
  68   union {
  69     float f32Value;
  70     uint32_t f32Bits;
  71   };
  72
  73   f32Value = v;
  74
  75   // pull the sign from v into f16bits
  76   uint16_t f16Bits = uint16_t(f32Bits >> 16) & 0x8000;
  77   const uint32_t mantissa = f32Bits & 0x7FFFFF;
  78   const uint32_t exp = (f32Bits >> 23) & 0xFF;
  79
  80   // Adapted from: OpenGL ES 2.0 Programming Guide Appx.
  81   // Converting Float to Half-Float
  82   // 143 = 255 - 127 + 15
  83   //     = sp_max - sp_bias + hp_bias
  84   if (exp >= 143) {
  85     if (mantissa && exp == 0xFF) {
  86       // Single precision was NaN
  87       return f16Bits | kFloat16Value_NaN;
  88     } else {
  89       // Outside range, store as infinity
  90       return f16Bits | kFloat16Value_Infinity;
  91     }
  92   }
  93
  94   // too small, try to make a denormalized number
  95   // 112 = 255 - 127 - (15 + 1)
  96   //     = sp_max - sp_bias - (hp_bias + 1)
  97   if (exp <= 112) {
  98     return f16Bits | uint16_t(mantissa >> (14 + 112 - exp));
  99   }
 100
 101   f16Bits |= uint16_t(exp - 112) << 10;
 102   f16Bits |= uint16_t(mantissa >> 13) & 0x03FF;
 103
 104   return f16Bits;
 105 }
 106
 107 MOZ_ALWAYS_INLINE float unpackFromFloat16(uint16_t v) {
 108   union {
 109     float f32Value;
 110     uint32_t f32Bits;
 111   };
 112
 113   // grab sign bit
 114   f32Bits = uint32_t(v & 0x8000) << 16;
 115   uint16_t exp = (v >> 10) & 0x001F;
 116   uint16_t mantissa = v & 0x03FF;
 117
 118   if (!exp) {
 119     // Handle denormalized numbers
 120     // Adapted from: OpenGL ES 2.0 Programming Guide Appx.
 121     // Converting Float to Half-Float
 122     if (mantissa) {
 123       exp = 112;  // See packToFloat16
 124       mantissa <<= 1;
 125       // For every leading zero, decrement the exponent
 126       // and shift the mantissa to the left
 127       while ((mantissa & (1 << 10)) == 0) {
 128         mantissa <<= 1;
 129         --exp;
 130       }
 131       mantissa &= 0x03FF;
 132
 133       f32Bits |= (exp << 23) | (mantissa << 13);
 134
 135       // Denormalized number
 136       return f32Value;
 137     }
 138
 139     // +/- zero
 140     return f32Value;
 141   }
 142
 143   if (exp == 0x001F) {
 144     if (v & 0x03FF) {
 145       // this is a NaN
 146       f32Bits |= 0x7FFFFFFF;
 147     } else {
 148       // this is -inf or +inf
 149       f32Bits |= 0x7F800000;
 150     }
 151     return f32Value;
 152   }
 153
 154   f32Bits |= uint32_t(exp + (-15 + 127)) << 23;
 155   f32Bits |= uint32_t(v & 0x03FF) << 13;
 156
 157   return f32Value;
 158 }
 159
 160 // These routines come from angle/common/mathutil.h
 161 // They are copied here to remove the dependency on ANGLE headers
 162 // included from mathutil.h
 163 MOZ_ALWAYS_INLINE uint16_t packToFloat11(float fp32) {
 164   const unsigned int float32MantissaMask = 0x7FFFFF;
 165   const unsigned int float32ExponentMask = 0x7F800000;
 166   const unsigned int float32SignMask = 0x80000000;
 167   const unsigned int float32ValueMask = ~float32SignMask;
 168   const unsigned int float32ExponentFirstBit = 23;
 169   const unsigned int float32ExponentBias = 127;
 170
 171   const unsigned short float11Max = 0x7BF;
 172   const unsigned short float11MantissaMask = 0x3F;
 173   const unsigned short float11ExponentMask = 0x7C0;
 174   const unsigned short float11BitMask = 0x7FF;
 175   const unsigned int float11ExponentBias = 14;
 176
 177   const unsigned int float32Maxfloat11 = 0x477E0000;
 178   const unsigned int float32Minfloat11 = 0x38800000;
 179
 180   const unsigned int float32Bits = BitwiseCast<unsigned int>(fp32);
 181   const bool float32Sign = (float32Bits & float32SignMask) == float32SignMask;
 182
 183   unsigned int float32Val = float32Bits & float32ValueMask;
 184
 185   if ((float32Val & float32ExponentMask) == float32ExponentMask) {
 186     // INF or NAN
 187     if ((float32Val & float32MantissaMask) != 0) {
 188       return float11ExponentMask | (((float32Val >> 17) | (float32Val >> 11) |
 189                                      (float32Val >> 6) | (float32Val)) &
 190                                     float11MantissaMask);
 191     } else if (float32Sign) {
 192       // -INF is clamped to 0 since float11 is positive only
 193       return 0;
 194     } else {
 195       return float11ExponentMask;
 196     }
 197   } else if (float32Sign) {
 198     // float11 is positive only, so clamp to zero
 199     return 0;
 200   } else if (float32Val > float32Maxfloat11) {
 201     // The number is too large to be represented as a float11, set to max
 202     return float11Max;
 203   } else {
 204     if (float32Val < float32Minfloat11) {
 205       // The number is too small to be represented as a normalized float11
 206       // Convert it to a denormalized value.
 207       const unsigned int shift = (float32ExponentBias - float11ExponentBias) -
 208                                  (float32Val >> float32ExponentFirstBit);
 209       float32Val = ((1 << float32ExponentFirstBit) |
 210                     (float32Val & float32MantissaMask)) >>
 211                    shift;
 212     } else {
 213       // Rebias the exponent to represent the value as a normalized float11
 214       float32Val += 0xC8000000;
 215     }
 216
 217     return ((float32Val + 0xFFFF + ((float32Val >> 17) & 1)) >> 17) &
 218            float11BitMask;
 219   }
 220 }
 221
 222 MOZ_ALWAYS_INLINE uint16_t packToFloat10(float fp32) {
 223   const unsigned int float32MantissaMask = 0x7FFFFF;
 224   const unsigned int float32ExponentMask = 0x7F800000;
 225   const unsigned int float32SignMask = 0x80000000;
 226   const unsigned int float32ValueMask = ~float32SignMask;
 227   const unsigned int float32ExponentFirstBit = 23;
 228   const unsigned int float32ExponentBias = 127;
 229
 230   const unsigned short float10Max = 0x3DF;
 231   const unsigned short float10MantissaMask = 0x1F;
 232   const unsigned short float10ExponentMask = 0x3E0;
 233   const unsigned short float10BitMask = 0x3FF;
 234   const unsigned int float10ExponentBias = 14;
 235
 236   const unsigned int float32Maxfloat10 = 0x477C0000;
 237   const unsigned int float32Minfloat10 = 0x38800000;
 238
 239   const unsigned int float32Bits = BitwiseCast<unsigned int>(fp32);
 240   const bool float32Sign = (float32Bits & float32SignMask) == float32SignMask;
 241
 242   unsigned int float32Val = float32Bits & float32ValueMask;
 243
 244   if ((float32Val & float32ExponentMask) == float32ExponentMask) {
 245     // INF or NAN
 246     if ((float32Val & float32MantissaMask) != 0) {
 247       return float10ExponentMask | (((float32Val >> 18) | (float32Val >> 13) |
 248                                      (float32Val >> 3) | (float32Val)) &
 249                                     float10MantissaMask);
 250     } else if (float32Sign) {
 251       // -INF is clamped to 0 since float11 is positive only
 252       return 0;
 253     } else {
 254       return float10ExponentMask;
 255     }
 256   } else if (float32Sign) {
 257     // float10 is positive only, so clamp to zero
 258     return 0;
 259   } else if (float32Val > float32Maxfloat10) {
 260     // The number is too large to be represented as a float11, set to max
 261     return float10Max;
 262   } else {
 263     if (float32Val < float32Minfloat10) {
 264       // The number is too small to be represented as a normalized float11
 265       // Convert it to a denormalized value.
 266       const unsigned int shift = (float32ExponentBias - float10ExponentBias) -
 267                                  (float32Val >> float32ExponentFirstBit);
 268       float32Val = ((1 << float32ExponentFirstBit) |
 269                     (float32Val & float32MantissaMask)) >>
 270                    shift;
 271     } else {
 272       // Rebias the exponent to represent the value as a normalized float11
 273       float32Val += 0xC8000000;
 274     }
 275
 276     return ((float32Val + 0x1FFFF + ((float32Val >> 18) & 1)) >> 18) &
 277            float10BitMask;
 278   }
 279 }
 280
 281 enum class WebGLTexelPremultiplicationOp : int {
 282   None,
 283   Premultiply,
 284   Unpremultiply
 285 };
 286
 287 namespace WebGLTexelConversions {
 288
 289 template <WebGLTexelFormat Format>
 290 struct IsFloatFormat {
 291   static const bool Value =
 292       Format == WebGLTexelFormat::A32F || Format == WebGLTexelFormat::R32F ||
 293       Format == WebGLTexelFormat::RA32F || Format == WebGLTexelFormat::RG32F ||
 294       Format == WebGLTexelFormat::RGB11F11F10F ||
 295       Format == WebGLTexelFormat::RGB32F || Format == WebGLTexelFormat::RGBA32F;
 296 };
 297
 298 template <WebGLTexelFormat Format>
 299 struct IsHalfFloatFormat {
 300   static const bool Value =
 301       Format == WebGLTexelFormat::A16F || Format == WebGLTexelFormat::R16F ||
 302       Format == WebGLTexelFormat::RA16F || Format == WebGLTexelFormat::RG16F ||
 303       Format == WebGLTexelFormat::RGB16F || Format == WebGLTexelFormat::RGBA16F;
 304 };
 305
 306 template <WebGLTexelFormat Format>
 307 struct Is16bppFormat {
 308   static const bool Value = Format == WebGLTexelFormat::RGB565 ||
 309                             Format == WebGLTexelFormat::RGBA4444 ||
 310                             Format == WebGLTexelFormat::RGBA5551;
 311 };
 312
 313 template <WebGLTexelFormat Format, bool IsFloat = IsFloatFormat<Format>::Value,
 314           bool Is16bpp = Is16bppFormat<Format>::Value,
 315           bool IsHalfFloat = IsHalfFloatFormat<Format>::Value>
 316 struct DataTypeForFormat {
 317   using Type = uint8_t;
 318 };
 319
 320 template <WebGLTexelFormat Format>
 321 struct DataTypeForFormat<Format, true, false, false> {
 322   using Type = float;
 323 };
 324
 325 template <WebGLTexelFormat Format>
 326 struct DataTypeForFormat<Format, false, true, false> {
 327   using Type = uint16_t;
 328 };
 329
 330 template <WebGLTexelFormat Format>
 331 struct DataTypeForFormat<Format, false, false, true> {
 332   using Type = uint16_t;
 333 };
 334
 335 template <>
 336 struct DataTypeForFormat<WebGLTexelFormat::RGB11F11F10F, true, false, false> {
 337   using Type = uint32_t;
 338 };
 339
 340 template <WebGLTexelFormat Format>
 341 struct IntermediateFormat {
 342   static const WebGLTexelFormat Value =
 343       IsFloatFormat<Format>::Value       ? WebGLTexelFormat::RGBA32F
 344       : IsHalfFloatFormat<Format>::Value ? WebGLTexelFormat::RGBA16F
 345                                          : WebGLTexelFormat::RGBA8;
 346 };
 347
 348 inline size_t TexelBytesForFormat(WebGLTexelFormat format) {
 349   switch (format) {
 350     case WebGLTexelFormat::A8:
 351     case WebGLTexelFormat::R8:
 352       return 1;
 353     case WebGLTexelFormat::A16F:
 354     case WebGLTexelFormat::R16F:
 355     case WebGLTexelFormat::RA8:
 356     case WebGLTexelFormat::RG8:
 357     case WebGLTexelFormat::RGB565:
 358     case WebGLTexelFormat::RGBA4444:
 359     case WebGLTexelFormat::RGBA5551:
 360       return 2;
 361     case WebGLTexelFormat::RGB8:
 362       return 3;
 363     case WebGLTexelFormat::A32F:
 364     case WebGLTexelFormat::R32F:
 365     case WebGLTexelFormat::RA16F:
 366     case WebGLTexelFormat::RG16F:
 367     case WebGLTexelFormat::RGB11F11F10F:
 368     case WebGLTexelFormat::RGBA8:
 369     case WebGLTexelFormat::BGRX8:
 370     case WebGLTexelFormat::BGRA8:
 371       return 4;
 372     case WebGLTexelFormat::RGB16F:
 373       return 6;
 374     case WebGLTexelFormat::RA32F:
 375     case WebGLTexelFormat::RG32F:
 376     case WebGLTexelFormat::RGBA16F:
 377       return 8;
 378     case WebGLTexelFormat::RGB32F:
 379       return 12;
 380     case WebGLTexelFormat::RGBA32F:
 381       return 16;
 382     default:
 383       MOZ_ASSERT(false, "Unknown texel format. Coding mistake?");
 384       return 0;
 385   }
 386 }
 387
 388 MOZ_ALWAYS_INLINE bool HasAlpha(WebGLTexelFormat format) {
 389   return (
 390       format == WebGLTexelFormat::A8 || format == WebGLTexelFormat::A16F ||
 391       format == WebGLTexelFormat::A32F || format == WebGLTexelFormat::RA8 ||
 392       format == WebGLTexelFormat::RA16F || format == WebGLTexelFormat::RA32F ||
 393       format == WebGLTexelFormat::RGBA4444 ||
 394       format == WebGLTexelFormat::RGBA5551 ||
 395       format == WebGLTexelFormat::RGBA8 ||
 396       format == WebGLTexelFormat::RGBA16F ||
 397       format == WebGLTexelFormat::RGBA32F || format == WebGLTexelFormat::BGRA8);
 398 }
 399
 400 MOZ_ALWAYS_INLINE bool HasColor(WebGLTexelFormat format) {
 401   return (
 402       format == WebGLTexelFormat::R8 || format == WebGLTexelFormat::R16F ||
 403       format == WebGLTexelFormat::R32F || format == WebGLTexelFormat::RA8 ||
 404       format == WebGLTexelFormat::RA16F || format == WebGLTexelFormat::RA32F ||
 405       format == WebGLTexelFormat::RG8 || format == WebGLTexelFormat::RG16F ||
 406       format == WebGLTexelFormat::RG32F || format == WebGLTexelFormat::RGB565 ||
 407       format == WebGLTexelFormat::RGB8 ||
 408       format == WebGLTexelFormat::RGB11F11F10F ||
 409       format == WebGLTexelFormat::RGB16F ||
 410       format == WebGLTexelFormat::RGB32F ||
 411       format == WebGLTexelFormat::RGBA4444 ||
 412       format == WebGLTexelFormat::RGBA5551 ||
 413       format == WebGLTexelFormat::RGBA8 ||
 414       format == WebGLTexelFormat::RGBA16F ||
 415       format == WebGLTexelFormat::RGBA32F ||
 416       format == WebGLTexelFormat::BGRX8 || format == WebGLTexelFormat::BGRA8);
 417 }
 418
 419 /****** BEGIN CODE SHARED WITH WEBKIT ******/
 420
 421 // the pack/unpack functions here are originally from this file:
 422 //   http://trac.webkit.org/browser/trunk/WebCore/platform/graphics/GraphicsContext3D.cpp
 423
 424 //----------------------------------------------------------------------
 425 // Pixel unpacking routines.
 426
 427 template <WebGLTexelFormat Format, typename SrcType, typename DstType>
 428 MOZ_ALWAYS_INLINE void unpack(const SrcType* __restrict src,
 429                               DstType* __restrict dst) {
 430   MOZ_ASSERT(false, "Unimplemented texture format conversion");
 431 }
 432
 433 ////////////////////////////////////////////////////////////////////////////////
 434 // 1-channel formats
 435 template <>
 436 MOZ_ALWAYS_INLINE void unpack<WebGLTexelFormat::A8, uint8_t, uint8_t>(
 437     const uint8_t* __restrict src, uint8_t* __restrict dst) {
 438   dst[0] = 0;
 439   dst[1] = 0;
 440   dst[2] = 0;
 441   dst[3] = src[0];
 442 }
 443
 444 template <>
 445 MOZ_ALWAYS_INLINE void unpack<WebGLTexelFormat::A16F, uint16_t, uint16_t>(
 446     const uint16_t* __restrict src, uint16_t* __restrict dst) {
 447   dst[0] = kFloat16Value_Zero;
 448   dst[1] = kFloat16Value_Zero;
 449   dst[2] = kFloat16Value_Zero;
 450   dst[3] = src[0];
 451 }
 452
 453 template <>
 454 MOZ_ALWAYS_INLINE void unpack<WebGLTexelFormat::A32F, float, float>(
 455     const float* __restrict src, float* __restrict dst) {
 456   dst[0] = 0;
 457   dst[1] = 0;
 458   dst[2] = 0;
 459   dst[3] = src[0];
 460 }
 461
 462 template <>
 463 MOZ_ALWAYS_INLINE void unpack<WebGLTexelFormat::R8, uint8_t, uint8_t>(
 464     const uint8_t* __restrict src, uint8_t* __restrict dst) {
 465   dst[0] = src[0];
 466   dst[1] = src[0];
 467   dst[2] = src[0];
 468   dst[3] = 0xFF;
 469 }
 470
 471 template <>
 472 MOZ_ALWAYS_INLINE void unpack<WebGLTexelFormat::R16F, uint16_t, uint16_t>(
 473     const uint16_t* __restrict src, uint16_t* __restrict dst) {
 474   dst[0] = src[0];
 475   dst[1] = src[0];
 476   dst[2] = src[0];
 477   dst[3] = kFloat16Value_One;
 478 }
 479
 480 template <>
 481 MOZ_ALWAYS_INLINE void unpack<WebGLTexelFormat::R32F, float, float>(
 482     const float* __restrict src, float* __restrict dst) {
 483   dst[0] = src[0];
 484   dst[1] = src[0];
 485   dst[2] = src[0];
 486   dst[3] = 1.0f;
 487 }
 488
 489 ////////////////////////////////////////////////////////////////////////////////
 490 // 2-channel formats
 491 template <>
 492 MOZ_ALWAYS_INLINE void unpack<WebGLTexelFormat::RA8, uint8_t, uint8_t>(
 493     const uint8_t* __restrict src, uint8_t* __restrict dst) {
 494   dst[0] = src[0];
 495   dst[1] = src[0];
 496   dst[2] = src[0];
 497   dst[3] = src[1];
 498 }
 499
 500 template <>
 501 MOZ_ALWAYS_INLINE void unpack<WebGLTexelFormat::RA16F, uint16_t, uint16_t>(
 502     const uint16_t* __restrict src, uint16_t* __restrict dst) {
 503   dst[0] = src[0];
 504   dst[1] = src[0];
 505   dst[2] = src[0];
 506   dst[3] = src[1];
 507 }
 508
 509 template <>
 510 MOZ_ALWAYS_INLINE void unpack<WebGLTexelFormat::RA32F, float, float>(
 511     const float* __restrict src, float* __restrict dst) {
 512   dst[0] = src[0];
 513   dst[1] = src[0];
 514   dst[2] = src[0];
 515   dst[3] = src[1];
 516 }
 517
 518 ////////////////////////////////////////////////////////////////////////////////
 519 // 3-channel formats
 520 template <>
 521 MOZ_ALWAYS_INLINE void unpack<WebGLTexelFormat::RGB565, uint16_t, uint8_t>(
 522     const uint16_t* __restrict src, uint8_t* __restrict dst) {
 523   uint16_t packedValue = src[0];
 524   uint8_t r = (packedValue >> 11) & 0x1F;
 525   uint8_t g = (packedValue >> 5) & 0x3F;
 526   uint8_t b = packedValue & 0x1F;
 527   dst[0] = uint8_t(r << 3) | (r & 0x7);
 528   dst[1] = uint8_t(g << 2) | (g & 0x3);
 529   dst[2] = uint8_t(b << 3) | (b & 0x7);
 530   dst[3] = 0xFF;
 531 }
 532
 533 template <>
 534 MOZ_ALWAYS_INLINE void unpack<WebGLTexelFormat::RGB8, uint8_t, uint8_t>(
 535     const uint8_t* __restrict src, uint8_t* __restrict dst) {
 536   dst[0] = src[0];
 537   dst[1] = src[1];
 538   dst[2] = src[2];
 539   dst[3] = 0xFF;
 540 }
 541
 542 template <>
 543 MOZ_ALWAYS_INLINE void unpack<WebGLTexelFormat::RGB16F, uint16_t, uint16_t>(
 544     const uint16_t* __restrict src, uint16_t* __restrict dst) {
 545   dst[0] = src[0];
 546   dst[1] = src[1];
 547   dst[2] = src[2];
 548   dst[3] = kFloat16Value_One;
 549 }
 550
 551 template <>
 552 MOZ_ALWAYS_INLINE void unpack<WebGLTexelFormat::RGB32F, float, float>(
 553     const float* __restrict src, float* __restrict dst) {
 554   dst[0] = src[0];
 555   dst[1] = src[1];
 556   dst[2] = src[2];
 557   dst[3] = 1.0f;
 558 }
 559
 560 ////////////////////////////////////////////////////////////////////////////////
 561 // 4-channel formats
 562 template <>
 563 MOZ_ALWAYS_INLINE void unpack<WebGLTexelFormat::RGBA4444, uint16_t, uint8_t>(
 564     const uint16_t* __restrict src, uint8_t* __restrict dst) {
 565   uint16_t packedValue = src[0];
 566   uint8_t r = (packedValue >> 12) & 0x0F;
 567   uint8_t g = (packedValue >> 8) & 0x0F;
 568   uint8_t b = (packedValue >> 4) & 0x0F;
 569   uint8_t a = packedValue & 0x0F;
 570   dst[0] = uint8_t(r << 4) | r;
 571   dst[1] = uint8_t(g << 4) | g;
 572   dst[2] = uint8_t(b << 4) | b;
 573   dst[3] = uint8_t(a << 4) | a;
 574 }
 575
 576 template <>
 577 MOZ_ALWAYS_INLINE void unpack<WebGLTexelFormat::RGBA5551, uint16_t, uint8_t>(
 578     const uint16_t* __restrict src, uint8_t* __restrict dst) {
 579   uint16_t packedValue = src[0];
 580   uint8_t r = (packedValue >> 11) & 0x1F;
 581   uint8_t g = (packedValue >> 6) & 0x1F;
 582   uint8_t b = (packedValue >> 1) & 0x1F;
 583   dst[0] = uint8_t(r << 3) | (r & 0x7);
 584   dst[1] = uint8_t(g << 3) | (g & 0x7);
 585   dst[2] = uint8_t(b << 3) | (b & 0x7);
 586   dst[3] = (packedValue & 0x1) ? 0xFF : 0;
 587 }
 588
 589 template <>
 590 MOZ_ALWAYS_INLINE void unpack<WebGLTexelFormat::RGBA8, uint8_t, uint8_t>(
 591     const uint8_t* __restrict src, uint8_t* __restrict dst) {
 592   dst[0] = src[0];
 593   dst[1] = src[1];
 594   dst[2] = src[2];
 595   dst[3] = src[3];
 596 }
 597
 598 template <>
 599 MOZ_ALWAYS_INLINE void unpack<WebGLTexelFormat::RGBA16F, uint16_t, uint16_t>(
 600     const uint16_t* __restrict src, uint16_t* __restrict dst) {
 601   dst[0] = src[0];
 602   dst[1] = src[1];
 603   dst[2] = src[2];
 604   dst[3] = src[3];
 605 }
 606
 607 template <>
 608 MOZ_ALWAYS_INLINE void unpack<WebGLTexelFormat::RGBA32F, float, float>(
 609     const float* __restrict src, float* __restrict dst) {
 610   dst[0] = src[0];
 611   dst[1] = src[1];
 612   dst[2] = src[2];
 613   dst[3] = src[3];
 614 }
 615
 616 ////////////////////////////////////////////////////////////////////////////////
 617 // DOM element formats
 618 template <>
 619 MOZ_ALWAYS_INLINE void unpack<WebGLTexelFormat::BGRX8, uint8_t, uint8_t>(
 620     const uint8_t* __restrict src, uint8_t* __restrict dst) {
 621   dst[0] = src[2];
 622   dst[1] = src[1];
 623   dst[2] = src[0];
 624   dst[3] = 0xFF;
 625 }
 626
 627 template <>
 628 MOZ_ALWAYS_INLINE void unpack<WebGLTexelFormat::BGRA8, uint8_t, uint8_t>(
 629     const uint8_t* __restrict src, uint8_t* __restrict dst) {
 630   dst[0] = src[2];
 631   dst[1] = src[1];
 632   dst[2] = src[0];
 633   dst[3] = src[3];
 634 }
 635
 636 //----------------------------------------------------------------------
 637 // Pixel packing routines.
 638 //
 639
 640 template <WebGLTexelFormat Format,
 641           WebGLTexelPremultiplicationOp PremultiplicationOp, typename SrcType,
 642           typename DstType>
 643 MOZ_ALWAYS_INLINE void pack(const SrcType* __restrict src,
 644                             DstType* __restrict dst) {
 645   MOZ_CRASH("GFX: Unimplemented texture format conversion");
 646 }
 647
 648 ////////////////////////////////////////////////////////////////////////////////
 649 // 1-channel formats
 650 template <>
 651 MOZ_ALWAYS_INLINE void
 652 pack<WebGLTexelFormat::A8, WebGLTexelPremultiplicationOp::None, uint8_t,
 653      uint8_t>(const uint8_t* __restrict src, uint8_t* __restrict dst) {
 654   dst[0] = src[3];
 655 }
 656
 657 template <>
 658 MOZ_ALWAYS_INLINE void
 659 pack<WebGLTexelFormat::A8, WebGLTexelPremultiplicationOp::Premultiply, uint8_t,
 660      uint8_t>(const uint8_t* __restrict src, uint8_t* __restrict dst) {
 661   dst[0] = src[3];
 662 }
 663
 664 template <>
 665 MOZ_ALWAYS_INLINE void
 666 pack<WebGLTexelFormat::A8, WebGLTexelPremultiplicationOp::Unpremultiply,
 667      uint8_t, uint8_t>(const uint8_t* __restrict src, uint8_t* __restrict dst) {
 668   dst[0] = src[3];
 669 }
 670
 671 template <>
 672 MOZ_ALWAYS_INLINE void
 673 pack<WebGLTexelFormat::A16F, WebGLTexelPremultiplicationOp::None, uint16_t,
 674      uint16_t>(const uint16_t* __restrict src, uint16_t* __restrict dst) {
 675   dst[0] = src[3];
 676 }
 677
 678 template <>
 679 MOZ_ALWAYS_INLINE void
 680 pack<WebGLTexelFormat::A16F, WebGLTexelPremultiplicationOp::Premultiply,
 681      uint16_t, uint16_t>(const uint16_t* __restrict src,
 682                          uint16_t* __restrict dst) {
 683   dst[0] = src[3];
 684 }
 685
 686 template <>
 687 MOZ_ALWAYS_INLINE void
 688 pack<WebGLTexelFormat::A16F, WebGLTexelPremultiplicationOp::Unpremultiply,
 689      uint16_t, uint16_t>(const uint16_t* __restrict src,
 690                          uint16_t* __restrict dst) {
 691   dst[0] = src[3];
 692 }
 693
 694 template <>
 695 MOZ_ALWAYS_INLINE void
 696 pack<WebGLTexelFormat::A32F, WebGLTexelPremultiplicationOp::None, float, float>(
 697     const float* __restrict src, float* __restrict dst) {
 698   dst[0] = src[3];
 699 }
 700
 701 template <>
 702 MOZ_ALWAYS_INLINE void
 703 pack<WebGLTexelFormat::A32F, WebGLTexelPremultiplicationOp::Premultiply, float,
 704      float>(const float* __restrict src, float* __restrict dst) {
 705   dst[0] = src[3];
 706 }
 707
 708 template <>
 709 MOZ_ALWAYS_INLINE void
 710 pack<WebGLTexelFormat::A32F, WebGLTexelPremultiplicationOp::Unpremultiply,
 711      float, float>(const float* __restrict src, float* __restrict dst) {
 712   dst[0] = src[3];
 713 }
 714
 715 template <>
 716 MOZ_ALWAYS_INLINE void
 717 pack<WebGLTexelFormat::R8, WebGLTexelPremultiplicationOp::None, uint8_t,
 718      uint8_t>(const uint8_t* __restrict src, uint8_t* __restrict dst) {
 719   dst[0] = src[0];
 720 }
 721
 722 template <>
 723 MOZ_ALWAYS_INLINE void
 724 pack<WebGLTexelFormat::R8, WebGLTexelPremultiplicationOp::Premultiply, uint8_t,
 725      uint8_t>(const uint8_t* __restrict src, uint8_t* __restrict dst) {
 726   float scaleFactor = src[3] / 255.0f;
 727   uint8_t srcR = static_cast<uint8_t>(src[0] * scaleFactor);
 728   dst[0] = srcR;
 729 }
 730
 731 template <>
 732 MOZ_ALWAYS_INLINE void
 733 pack<WebGLTexelFormat::R8, WebGLTexelPremultiplicationOp::Unpremultiply,
 734      uint8_t, uint8_t>(const uint8_t* __restrict src, uint8_t* __restrict dst) {
 735   float scaleFactor = src[3] ? 255.0f / src[3] : 1.0f;
 736   uint8_t srcR = static_cast<uint8_t>(src[0] * scaleFactor);
 737   dst[0] = srcR;
 738 }
 739
 740 template <>
 741 MOZ_ALWAYS_INLINE void
 742 pack<WebGLTexelFormat::R16F, WebGLTexelPremultiplicationOp::None, uint16_t,
 743      uint16_t>(const uint16_t* __restrict src, uint16_t* __restrict dst) {
 744   dst[0] = src[0];
 745 }
 746
 747 template <>
 748 MOZ_ALWAYS_INLINE void
 749 pack<WebGLTexelFormat::R16F, WebGLTexelPremultiplicationOp::Premultiply,
 750      uint16_t, uint16_t>(const uint16_t* __restrict src,
 751                          uint16_t* __restrict dst) {
 752   float scaleFactor = unpackFromFloat16(src[3]);
 753   dst[0] = packToFloat16(unpackFromFloat16(src[0]) * scaleFactor);
 754 }
 755
 756 template <>
 757 MOZ_ALWAYS_INLINE void
 758 pack<WebGLTexelFormat::R16F, WebGLTexelPremultiplicationOp::Unpremultiply,
 759      uint16_t, uint16_t>(const uint16_t* __restrict src,
 760                          uint16_t* __restrict dst) {
 761   float unpackedAlpha = unpackFromFloat16(src[3]);
 762   float scaleFactor = unpackedAlpha ? 1.0f / unpackedAlpha : 1.0f;
 763   dst[0] = packToFloat16(unpackFromFloat16(src[0]) * scaleFactor);
 764 }
 765
 766 template <>
 767 MOZ_ALWAYS_INLINE void
 768 pack<WebGLTexelFormat::R32F, WebGLTexelPremultiplicationOp::None, float, float>(
 769     const float* __restrict src, float* __restrict dst) {
 770   dst[0] = src[0];
 771 }
 772
 773 template <>
 774 MOZ_ALWAYS_INLINE void
 775 pack<WebGLTexelFormat::R32F, WebGLTexelPremultiplicationOp::Premultiply, float,
 776      float>(const float* __restrict src, float* __restrict dst) {
 777   float scaleFactor = src[3];
 778   dst[0] = src[0] * scaleFactor;
 779 }
 780
 781 template <>
 782 MOZ_ALWAYS_INLINE void
 783 pack<WebGLTexelFormat::R32F, WebGLTexelPremultiplicationOp::Unpremultiply,
 784      float, float>(const float* __restrict src, float* __restrict dst) {
 785   float scaleFactor = src[3] ? 1.0f / src[3] : 1.0f;
 786   dst[0] = src[0] * scaleFactor;
 787 }
 788
 789 ////////////////////////////////////////////////////////////////////////////////
 790 // 2-channel formats
 791 template <>
 792 MOZ_ALWAYS_INLINE void
 793 pack<WebGLTexelFormat::RA8, WebGLTexelPremultiplicationOp::None, uint8_t,
 794      uint8_t>(const uint8_t* __restrict src, uint8_t* __restrict dst) {
 795   dst[0] = src[0];
 796   dst[1] = src[3];
 797 }
 798
 799 template <>
 800 MOZ_ALWAYS_INLINE void
 801 pack<WebGLTexelFormat::RA8, WebGLTexelPremultiplicationOp::Premultiply, uint8_t,
 802      uint8_t>(const uint8_t* __restrict src, uint8_t* __restrict dst) {
 803   float scaleFactor = src[3] / 255.0f;
 804   uint8_t srcR = static_cast<uint8_t>(src[0] * scaleFactor);
 805   dst[0] = srcR;
 806   dst[1] = src[3];
 807 }
 808
 809 // FIXME: this routine is lossy and must be removed.
 810 template <>
 811 MOZ_ALWAYS_INLINE void
 812 pack<WebGLTexelFormat::RA8, WebGLTexelPremultiplicationOp::Unpremultiply,
 813      uint8_t, uint8_t>(const uint8_t* __restrict src, uint8_t* __restrict dst) {
 814   float scaleFactor = src[3] ? 255.0f / src[3] : 1.0f;
 815   uint8_t srcR = static_cast<uint8_t>(src[0] * scaleFactor);
 816   dst[0] = srcR;
 817   dst[1] = src[3];
 818 }
 819
 820 template <>
 821 MOZ_ALWAYS_INLINE void
 822 pack<WebGLTexelFormat::RA16F, WebGLTexelPremultiplicationOp::None, uint16_t,
 823      uint16_t>(const uint16_t* __restrict src, uint16_t* __restrict dst) {
 824   dst[0] = src[0];
 825   dst[1] = src[3];
 826 }
 827
 828 template <>
 829 MOZ_ALWAYS_INLINE void
 830 pack<WebGLTexelFormat::RA16F, WebGLTexelPremultiplicationOp::Premultiply,
 831      uint16_t, uint16_t>(const uint16_t* __restrict src,
 832                          uint16_t* __restrict dst) {
 833   float scaleFactor = unpackFromFloat16(src[3]);
 834   dst[0] = packToFloat16(unpackFromFloat16(src[0]) * scaleFactor);
 835   dst[1] = src[3];
 836 }
 837
 838 template <>
 839 MOZ_ALWAYS_INLINE void
 840 pack<WebGLTexelFormat::RA16F, WebGLTexelPremultiplicationOp::Unpremultiply,
 841      uint16_t, uint16_t>(const uint16_t* __restrict src,
 842                          uint16_t* __restrict dst) {
 843   float unpackedAlpha = unpackFromFloat16(src[3]);
 844   float scaleFactor = unpackedAlpha ? 1.0f / unpackedAlpha : 1.0f;
 845   dst[0] = packToFloat16(unpackFromFloat16(src[0]) * scaleFactor);
 846   dst[1] = src[3];
 847 }
 848
 849 template <>
 850 MOZ_ALWAYS_INLINE void pack<WebGLTexelFormat::RA32F,
 851                             WebGLTexelPremultiplicationOp::None, float, float>(
 852     const float* __restrict src, float* __restrict dst) {
 853   dst[0] = src[0];
 854   dst[1] = src[3];
 855 }
 856
 857 template <>
 858 MOZ_ALWAYS_INLINE void
 859 pack<WebGLTexelFormat::RA32F, WebGLTexelPremultiplicationOp::Premultiply, float,
 860      float>(const float* __restrict src, float* __restrict dst) {
 861   float scaleFactor = src[3];
 862   dst[0] = src[0] * scaleFactor;
 863   dst[1] = src[3];
 864 }
 865
 866 template <>
 867 MOZ_ALWAYS_INLINE void
 868 pack<WebGLTexelFormat::RA32F, WebGLTexelPremultiplicationOp::Unpremultiply,
 869      float, float>(const float* __restrict src, float* __restrict dst) {
 870   float scaleFactor = src[3] ? 1.0f / src[3] : 1.0f;
 871   dst[0] = src[0] * scaleFactor;
 872   dst[1] = src[3];
 873 }
 874
 875 template <>
 876 MOZ_ALWAYS_INLINE void
 877 pack<WebGLTexelFormat::RG8, WebGLTexelPremultiplicationOp::None, uint8_t,
 878      uint8_t>(const uint8_t* __restrict src, uint8_t* __restrict dst) {
 879   dst[0] = src[0];
 880   dst[1] = src[1];
 881 }
 882
 883 template <>
 884 MOZ_ALWAYS_INLINE void
 885 pack<WebGLTexelFormat::RG8, WebGLTexelPremultiplicationOp::Premultiply, uint8_t,
 886      uint8_t>(const uint8_t* __restrict src, uint8_t* __restrict dst) {
 887   float scaleFactor = src[3] / 255.0f;
 888   uint8_t srcR = static_cast<uint8_t>(src[0] * scaleFactor);
 889   uint8_t srcG = static_cast<uint8_t>(src[1] * scaleFactor);
 890   dst[0] = srcR;
 891   dst[1] = srcG;
 892 }
 893
 894 // FIXME: this routine is lossy and must be removed.
 895 template <>
 896 MOZ_ALWAYS_INLINE void
 897 pack<WebGLTexelFormat::RG8, WebGLTexelPremultiplicationOp::Unpremultiply,
 898      uint8_t, uint8_t>(const uint8_t* __restrict src, uint8_t* __restrict dst) {
 899   float scaleFactor = src[3] ? 255.0f / src[3] : 1.0f;
 900   uint8_t srcR = static_cast<uint8_t>(src[0] * scaleFactor);
 901   uint8_t srcG = static_cast<uint8_t>(src[1] * scaleFactor);
 902   dst[0] = srcR;
 903   dst[1] = srcG;
 904 }
 905
 906 template <>
 907 MOZ_ALWAYS_INLINE void
 908 pack<WebGLTexelFormat::RG16F, WebGLTexelPremultiplicationOp::None, uint16_t,
 909      uint16_t>(const uint16_t* __restrict src, uint16_t* __restrict dst) {
 910   dst[0] = src[0];
 911   dst[1] = src[1];
 912 }
 913
 914 template <>
 915 MOZ_ALWAYS_INLINE void
 916 pack<WebGLTexelFormat::RG16F, WebGLTexelPremultiplicationOp::Premultiply,
 917      uint16_t, uint16_t>(const uint16_t* __restrict src,
 918                          uint16_t* __restrict dst) {
 919   float scaleFactor = unpackFromFloat16(src[3]);
 920   dst[0] = packToFloat16(unpackFromFloat16(src[0]) * scaleFactor);
 921   dst[1] = packToFloat16(unpackFromFloat16(src[1]) * scaleFactor);
 922 }
 923
 924 template <>
 925 MOZ_ALWAYS_INLINE void
 926 pack<WebGLTexelFormat::RG16F, WebGLTexelPremultiplicationOp::Unpremultiply,
 927      uint16_t, uint16_t>(const uint16_t* __restrict src,
 928                          uint16_t* __restrict dst) {
 929   float unpackedAlpha = unpackFromFloat16(src[3]);
 930   float scaleFactor = unpackedAlpha ? 1.0f / unpackedAlpha : 1.0f;
 931   dst[0] = packToFloat16(unpackFromFloat16(src[0]) * scaleFactor);
 932   dst[1] = packToFloat16(unpackFromFloat16(src[1]) * scaleFactor);
 933 }
 934
 935 template <>
 936 MOZ_ALWAYS_INLINE void pack<WebGLTexelFormat::RG32F,
 937                             WebGLTexelPremultiplicationOp::None, float, float>(
 938     const float* __restrict src, float* __restrict dst) {
 939   dst[0] = src[0];
 940   dst[1] = src[1];
 941 }
 942
 943 template <>
 944 MOZ_ALWAYS_INLINE void
 945 pack<WebGLTexelFormat::RG32F, WebGLTexelPremultiplicationOp::Premultiply, float,
 946      float>(const float* __restrict src, float* __restrict dst) {
 947   float scaleFactor = src[3];
 948   dst[0] = src[0] * scaleFactor;
 949   dst[1] = src[1] * scaleFactor;
 950 }
 951
 952 template <>
 953 MOZ_ALWAYS_INLINE void
 954 pack<WebGLTexelFormat::RG32F, WebGLTexelPremultiplicationOp::Unpremultiply,
 955      float, float>(const float* __restrict src, float* __restrict dst) {
 956   float scaleFactor = src[3] ? 1.0f / src[3] : 1.0f;
 957   dst[0] = src[0] * scaleFactor;
 958   dst[1] = src[1] * scaleFactor;
 959 }
 960
 961 ////////////////////////////////////////////////////////////////////////////////
 962 // 3-channel formats
 963 template <>
 964 MOZ_ALWAYS_INLINE void
 965 pack<WebGLTexelFormat::RGB565, WebGLTexelPremultiplicationOp::None, uint8_t,
 966      uint16_t>(const uint8_t* __restrict src, uint16_t* __restrict dst) {
 967   *dst = uint16_t(((src[0] & 0xF8) << 8) | ((src[1] & 0xFC) << 3) |
 968                   ((src[2] & 0xF8) >> 3));
 969 }
 970
 971 template <>
 972 MOZ_ALWAYS_INLINE void
 973 pack<WebGLTexelFormat::RGB565, WebGLTexelPremultiplicationOp::Premultiply,
 974      uint8_t, uint16_t>(const uint8_t* __restrict src,
 975                         uint16_t* __restrict dst) {
 976   float scaleFactor = src[3] / 255.0f;
 977   uint8_t srcR = static_cast<uint8_t>(src[0] * scaleFactor);
 978   uint8_t srcG = static_cast<uint8_t>(src[1] * scaleFactor);
 979   uint8_t srcB = static_cast<uint8_t>(src[2] * scaleFactor);
 980   *dst = uint16_t(((srcR & 0xF8) << 8) | ((srcG & 0xFC) << 3) |
 981                   ((srcB & 0xF8) >> 3));
 982 }
 983
 984 // FIXME: this routine is lossy and must be removed.
 985 template <>
 986 MOZ_ALWAYS_INLINE void
 987 pack<WebGLTexelFormat::RGB565, WebGLTexelPremultiplicationOp::Unpremultiply,
 988      uint8_t, uint16_t>(const uint8_t* __restrict src,
 989                         uint16_t* __restrict dst) {
 990   float scaleFactor = src[3] ? 255.0f / src[3] : 1.0f;
 991   uint8_t srcR = static_cast<uint8_t>(src[0] * scaleFactor);
 992   uint8_t srcG = static_cast<uint8_t>(src[1] * scaleFactor);
 993   uint8_t srcB = static_cast<uint8_t>(src[2] * scaleFactor);
 994   *dst = uint16_t(((srcR & 0xF8) << 8) | ((srcG & 0xFC) << 3) |
 995                   ((srcB & 0xF8) >> 3));
 996 }
 997
 998 template <>
 999 MOZ_ALWAYS_INLINE void
1000 pack<WebGLTexelFormat::RGB8, WebGLTexelPremultiplicationOp::None, uint8_t,
1001      uint8_t>(const uint8_t* __restrict src, uint8_t* __restrict dst) {
1002   dst[0] = src[0];
1003   dst[1] = src[1];
1004   dst[2] = src[2];
1005 }
1006
1007 template <>
1008 MOZ_ALWAYS_INLINE void
1009 pack<WebGLTexelFormat::RGB8, WebGLTexelPremultiplicationOp::Premultiply,
1010      uint8_t, uint8_t>(const uint8_t* __restrict src, uint8_t* __restrict dst) {
1011   float scaleFactor = src[3] / 255.0f;
1012   uint8_t srcR = static_cast<uint8_t>(src[0] * scaleFactor);
1013   uint8_t srcG = static_cast<uint8_t>(src[1] * scaleFactor);
1014   uint8_t srcB = static_cast<uint8_t>(src[2] * scaleFactor);
1015   dst[0] = srcR;
1016   dst[1] = srcG;
1017   dst[2] = srcB;
1018 }
1019
1020 template <>
1021 MOZ_ALWAYS_INLINE void
1022 pack<WebGLTexelFormat::RGB8, WebGLTexelPremultiplicationOp::Unpremultiply,
1023      uint8_t, uint8_t>(const uint8_t* __restrict src, uint8_t* __restrict dst) {
1024   float scaleFactor = src[3] ? 255.0f / src[3] : 1.0f;
1025   uint8_t srcR = static_cast<uint8_t>(src[0] * scaleFactor);
1026   uint8_t srcG = static_cast<uint8_t>(src[1] * scaleFactor);
1027   uint8_t srcB = static_cast<uint8_t>(src[2] * scaleFactor);
1028   dst[0] = srcR;
1029   dst[1] = srcG;
1030   dst[2] = srcB;
1031 }
1032
1033 template <>
1034 MOZ_ALWAYS_INLINE void
1035 pack<WebGLTexelFormat::RGB11F11F10F, WebGLTexelPremultiplicationOp::None, float,
1036      uint32_t>(const float* __restrict src, uint32_t* __restrict dst) {
1037   dst[0] = ((packToFloat11(src[0]) << 0) | (packToFloat11(src[1]) << 11) |
1038             (packToFloat10(src[2]) << 22));
1039 }
1040
1041 template <>
1042 MOZ_ALWAYS_INLINE void
1043 pack<WebGLTexelFormat::RGB11F11F10F, WebGLTexelPremultiplicationOp::Premultiply,
1044      float, uint32_t>(const float* __restrict src, uint32_t* __restrict dst) {
1045   float scaleFactor = src[3];
1046   dst[0] = ((packToFloat11(src[0] * scaleFactor) << 0) |
1047             (packToFloat11(src[1] * scaleFactor) << 11) |
1048             (packToFloat10(src[2] * scaleFactor) << 22));
1049 }
1050
1051 template <>
1052 MOZ_ALWAYS_INLINE void pack<WebGLTexelFormat::RGB11F11F10F,
1053                             WebGLTexelPremultiplicationOp::Unpremultiply, float,
1054                             uint32_t>(const float* __restrict src,
1055                                       uint32_t* __restrict dst) {
1056   float scaleFactor = src[3] ? 1.0f / src[3] : 1.0f;
1057   dst[0] = ((packToFloat11(src[0] * scaleFactor) << 0) |
1058             (packToFloat11(src[1] * scaleFactor) << 11) |
1059             (packToFloat10(src[2] * scaleFactor) << 22));
1060 }
1061
1062 template <>
1063 MOZ_ALWAYS_INLINE void
1064 pack<WebGLTexelFormat::RGB16F, WebGLTexelPremultiplicationOp::None, uint16_t,
1065      uint16_t>(const uint16_t* __restrict src, uint16_t* __restrict dst) {
1066   dst[0] = src[0];
1067   dst[1] = src[1];
1068   dst[2] = src[2];
1069 }
1070
1071 template <>
1072 MOZ_ALWAYS_INLINE void
1073 pack<WebGLTexelFormat::RGB16F, WebGLTexelPremultiplicationOp::Premultiply,
1074      uint16_t, uint16_t>(const uint16_t* __restrict src,
1075                          uint16_t* __restrict dst) {
1076   float scaleFactor = unpackFromFloat16(src[3]);
1077   dst[0] = packToFloat16(unpackFromFloat16(src[0]) * scaleFactor);
1078   dst[1] = packToFloat16(unpackFromFloat16(src[1]) * scaleFactor);
1079   dst[2] = packToFloat16(unpackFromFloat16(src[2]) * scaleFactor);
1080 }
1081
1082 template <>
1083 MOZ_ALWAYS_INLINE void
1084 pack<WebGLTexelFormat::RGB16F, WebGLTexelPremultiplicationOp::Unpremultiply,
1085      uint16_t, uint16_t>(const uint16_t* __restrict src,
1086                          uint16_t* __restrict dst) {
1087   float unpackedAlpha = unpackFromFloat16(src[3]);
1088   float scaleFactor = unpackedAlpha ? 1.0f / unpackedAlpha : 1.0f;
1089   dst[0] = packToFloat16(unpackFromFloat16(src[0]) * scaleFactor);
1090   dst[1] = packToFloat16(unpackFromFloat16(src[1]) * scaleFactor);
1091   dst[2] = packToFloat16(unpackFromFloat16(src[2]) * scaleFactor);
1092 }
1093
1094 template <>
1095 MOZ_ALWAYS_INLINE void pack<WebGLTexelFormat::RGB32F,
1096                             WebGLTexelPremultiplicationOp::None, float, float>(
1097     const float* __restrict src, float* __restrict dst) {
1098   dst[0] = src[0];
1099   dst[1] = src[1];
1100   dst[2] = src[2];
1101 }
1102
1103 template <>
1104 MOZ_ALWAYS_INLINE void
1105 pack<WebGLTexelFormat::RGB32F, WebGLTexelPremultiplicationOp::Premultiply,
1106      float, float>(const float* __restrict src, float* __restrict dst) {
1107   float scaleFactor = src[3];
1108   dst[0] = src[0] * scaleFactor;
1109   dst[1] = src[1] * scaleFactor;
1110   dst[2] = src[2] * scaleFactor;
1111 }
1112
1113 template <>
1114 MOZ_ALWAYS_INLINE void
1115 pack<WebGLTexelFormat::RGB32F, WebGLTexelPremultiplicationOp::Unpremultiply,
1116      float, float>(const float* __restrict src, float* __restrict dst) {
1117   float scaleFactor = src[3] ? 1.0f / src[3] : 1.0f;
1118   dst[0] = src[0] * scaleFactor;
1119   dst[1] = src[1] * scaleFactor;
1120   dst[2] = src[2] * scaleFactor;
1121 }
1122
1123 ////////////////////////////////////////////////////////////////////////////////
1124 // 4-channel formats
1125 template <>
1126 MOZ_ALWAYS_INLINE void
1127 pack<WebGLTexelFormat::RGBA4444, WebGLTexelPremultiplicationOp::None, uint8_t,
1128      uint16_t>(const uint8_t* __restrict src, uint16_t* __restrict dst) {
1129   *dst = uint16_t(((src[0] & 0xF0) << 8) | ((src[1] & 0xF0) << 4) |
1130                   (src[2] & 0xF0) | (src[3] >> 4));
1131 }
1132
1133 template <>
1134 MOZ_ALWAYS_INLINE void
1135 pack<WebGLTexelFormat::RGBA4444, WebGLTexelPremultiplicationOp::Premultiply,
1136      uint8_t, uint16_t>(const uint8_t* __restrict src,
1137                         uint16_t* __restrict dst) {
1138   float scaleFactor = src[3] / 255.0f;
1139   uint8_t srcR = static_cast<uint8_t>(src[0] * scaleFactor);
1140   uint8_t srcG = static_cast<uint8_t>(src[1] * scaleFactor);
1141   uint8_t srcB = static_cast<uint8_t>(src[2] * scaleFactor);
1142   *dst = uint16_t(((srcR & 0xF0) << 8) | ((srcG & 0xF0) << 4) | (srcB & 0xF0) |
1143                   (src[3] >> 4));
1144 }
1145
1146 // FIXME: this routine is lossy and must be removed.
1147 template <>
1148 MOZ_ALWAYS_INLINE void
1149 pack<WebGLTexelFormat::RGBA4444, WebGLTexelPremultiplicationOp::Unpremultiply,
1150      uint8_t, uint16_t>(const uint8_t* __restrict src,
1151                         uint16_t* __restrict dst) {
1152   float scaleFactor = src[3] ? 255.0f / src[3] : 1.0f;
1153   uint8_t srcR = static_cast<uint8_t>(src[0] * scaleFactor);
1154   uint8_t srcG = static_cast<uint8_t>(src[1] * scaleFactor);
1155   uint8_t srcB = static_cast<uint8_t>(src[2] * scaleFactor);
1156   *dst = uint16_t(((srcR & 0xF0) << 8) | ((srcG & 0xF0) << 4) | (srcB & 0xF0) |
1157                   (src[3] >> 4));
1158 }
1159
1160 template <>
1161 MOZ_ALWAYS_INLINE void
1162 pack<WebGLTexelFormat::RGBA5551, WebGLTexelPremultiplicationOp::None, uint8_t,
1163      uint16_t>(const uint8_t* __restrict src, uint16_t* __restrict dst) {
1164   *dst = uint16_t(((src[0] & 0xF8) << 8) | ((src[1] & 0xF8) << 3) |
1165                   ((src[2] & 0xF8) >> 2) | (src[3] >> 7));
1166 }
1167
1168 template <>
1169 MOZ_ALWAYS_INLINE void
1170 pack<WebGLTexelFormat::RGBA5551, WebGLTexelPremultiplicationOp::Premultiply,
1171      uint8_t, uint16_t>(const uint8_t* __restrict src,
1172                         uint16_t* __restrict dst) {
1173   float scaleFactor = src[3] / 255.0f;
1174   uint8_t srcR = static_cast<uint8_t>(src[0] * scaleFactor);
1175   uint8_t srcG = static_cast<uint8_t>(src[1] * scaleFactor);
1176   uint8_t srcB = static_cast<uint8_t>(src[2] * scaleFactor);
1177   *dst = uint16_t(((srcR & 0xF8) << 8) | ((srcG & 0xF8) << 3) |
1178                   ((srcB & 0xF8) >> 2) | (src[3] >> 7));
1179 }
1180
1181 // FIXME: this routine is lossy and must be removed.
1182 template <>
1183 MOZ_ALWAYS_INLINE void
1184 pack<WebGLTexelFormat::RGBA5551, WebGLTexelPremultiplicationOp::Unpremultiply,
1185      uint8_t, uint16_t>(const uint8_t* __restrict src,
1186                         uint16_t* __restrict dst) {
1187   float scaleFactor = src[3] ? 255.0f / src[3] : 1.0f;
1188   uint8_t srcR = static_cast<uint8_t>(src[0] * scaleFactor);
1189   uint8_t srcG = static_cast<uint8_t>(src[1] * scaleFactor);
1190   uint8_t srcB = static_cast<uint8_t>(src[2] * scaleFactor);
1191   *dst = uint16_t(((srcR & 0xF8) << 8) | ((srcG & 0xF8) << 3) |
1192                   ((srcB & 0xF8) >> 2) | (src[3] >> 7));
1193 }
1194
1195 template <>
1196 MOZ_ALWAYS_INLINE void
1197 pack<WebGLTexelFormat::RGBA8, WebGLTexelPremultiplicationOp::None, uint8_t,
1198      uint8_t>(const uint8_t* __restrict src, uint8_t* __restrict dst) {
1199   dst[0] = src[0];
1200   dst[1] = src[1];
1201   dst[2] = src[2];
1202   dst[3] = src[3];
1203 }
1204
1205 template <>
1206 MOZ_ALWAYS_INLINE void
1207 pack<WebGLTexelFormat::RGBA8, WebGLTexelPremultiplicationOp::Premultiply,
1208      uint8_t, uint8_t>(const uint8_t* __restrict src, uint8_t* __restrict dst) {
1209   float scaleFactor = src[3] / 255.0f;
1210   uint8_t srcR = static_cast<uint8_t>(src[0] * scaleFactor);
1211   uint8_t srcG = static_cast<uint8_t>(src[1] * scaleFactor);
1212   uint8_t srcB = static_cast<uint8_t>(src[2] * scaleFactor);
1213   dst[0] = srcR;
1214   dst[1] = srcG;
1215   dst[2] = srcB;
1216   dst[3] = src[3];
1217 }
1218
1219 // FIXME: this routine is lossy and must be removed.
1220 template <>
1221 MOZ_ALWAYS_INLINE void
1222 pack<WebGLTexelFormat::RGBA8, WebGLTexelPremultiplicationOp::Unpremultiply,
1223      uint8_t, uint8_t>(const uint8_t* __restrict src, uint8_t* __restrict dst) {
1224   float scaleFactor = src[3] ? 255.0f / src[3] : 1.0f;
1225   uint8_t srcR = static_cast<uint8_t>(src[0] * scaleFactor);
1226   uint8_t srcG = static_cast<uint8_t>(src[1] * scaleFactor);
1227   uint8_t srcB = static_cast<uint8_t>(src[2] * scaleFactor);
1228   dst[0] = srcR;
1229   dst[1] = srcG;
1230   dst[2] = srcB;
1231   dst[3] = src[3];
1232 }
1233
1234 template <>
1235 MOZ_ALWAYS_INLINE void
1236 pack<WebGLTexelFormat::BGRA8, WebGLTexelPremultiplicationOp::None, uint8_t,
1237      uint8_t>(const uint8_t* __restrict src, uint8_t* __restrict dst) {
1238   dst[0] = src[2];
1239   dst[1] = src[1];
1240   dst[2] = src[0];
1241   dst[3] = src[3];
1242 }
1243
1244 template <>
1245 MOZ_ALWAYS_INLINE void
1246 pack<WebGLTexelFormat::BGRA8, WebGLTexelPremultiplicationOp::Premultiply,
1247      uint8_t, uint8_t>(const uint8_t* __restrict src, uint8_t* __restrict dst) {
1248   float scaleFactor = src[3] / 255.0f;
1249   uint8_t srcR = static_cast<uint8_t>(src[0] * scaleFactor);
1250   uint8_t srcG = static_cast<uint8_t>(src[1] * scaleFactor);
1251   uint8_t srcB = static_cast<uint8_t>(src[2] * scaleFactor);
1252   dst[0] = srcB;
1253   dst[1] = srcG;
1254   dst[2] = srcR;
1255   dst[3] = src[3];
1256 }
1257
1258 // FIXME: this routine is lossy and must be removed.
1259 template <>
1260 MOZ_ALWAYS_INLINE void
1261 pack<WebGLTexelFormat::BGRA8, WebGLTexelPremultiplicationOp::Unpremultiply,
1262      uint8_t, uint8_t>(const uint8_t* __restrict src, uint8_t* __restrict dst) {
1263   float scaleFactor = src[3] ? 255.0f / src[3] : 1.0f;
1264   uint8_t srcR = static_cast<uint8_t>(src[0] * scaleFactor);
1265   uint8_t srcG = static_cast<uint8_t>(src[1] * scaleFactor);
1266   uint8_t srcB = static_cast<uint8_t>(src[2] * scaleFactor);
1267   dst[0] = srcB;
1268   dst[1] = srcG;
1269   dst[2] = srcR;
1270   dst[3] = src[3];
1271 }
1272
1273 template <>
1274 MOZ_ALWAYS_INLINE void
1275 pack<WebGLTexelFormat::RGBA16F, WebGLTexelPremultiplicationOp::None, uint16_t,
1276      uint16_t>(const uint16_t* __restrict src, uint16_t* __restrict dst) {
1277   dst[0] = src[0];
1278   dst[1] = src[1];
1279   dst[2] = src[2];
1280   dst[3] = src[3];
1281 }
1282
1283 template <>
1284 MOZ_ALWAYS_INLINE void
1285 pack<WebGLTexelFormat::RGBA16F, WebGLTexelPremultiplicationOp::Premultiply,
1286      uint16_t, uint16_t>(const uint16_t* __restrict src,
1287                          uint16_t* __restrict dst) {
1288   float scaleFactor = unpackFromFloat16(src[3]);
1289   dst[0] = packToFloat16(unpackFromFloat16(src[0]) * scaleFactor);
1290   dst[1] = packToFloat16(unpackFromFloat16(src[1]) * scaleFactor);
1291   dst[2] = packToFloat16(unpackFromFloat16(src[2]) * scaleFactor);
1292   dst[3] = src[3];
1293 }
1294
1295 template <>
1296 MOZ_ALWAYS_INLINE void
1297 pack<WebGLTexelFormat::RGBA16F, WebGLTexelPremultiplicationOp::Unpremultiply,
1298      uint16_t, uint16_t>(const uint16_t* __restrict src,
1299                          uint16_t* __restrict dst) {
1300   float unpackedAlpha = unpackFromFloat16(src[3]);
1301   float scaleFactor = unpackedAlpha ? 1.0f / unpackedAlpha : 1.0f;
1302   dst[0] = packToFloat16(unpackFromFloat16(src[0]) * scaleFactor);
1303   dst[1] = packToFloat16(unpackFromFloat16(src[1]) * scaleFactor);
1304   dst[2] = packToFloat16(unpackFromFloat16(src[2]) * scaleFactor);
1305   dst[3] = src[3];
1306 }
1307
1308 template <>
1309 MOZ_ALWAYS_INLINE void pack<WebGLTexelFormat::RGBA32F,
1310                             WebGLTexelPremultiplicationOp::None, float, float>(
1311     const float* __restrict src, float* __restrict dst) {
1312   dst[0] = src[0];
1313   dst[1] = src[1];
1314   dst[2] = src[2];
1315   dst[3] = src[3];
1316 }
1317
1318 template <>
1319 MOZ_ALWAYS_INLINE void
1320 pack<WebGLTexelFormat::RGBA32F, WebGLTexelPremultiplicationOp::Premultiply,
1321      float, float>(const float* __restrict src, float* __restrict dst) {
1322   float scaleFactor = src[3];
1323   dst[0] = src[0] * scaleFactor;
1324   dst[1] = src[1] * scaleFactor;
1325   dst[2] = src[2] * scaleFactor;
1326   dst[3] = src[3];
1327 }
1328
1329 template <>
1330 MOZ_ALWAYS_INLINE void
1331 pack<WebGLTexelFormat::RGBA32F, WebGLTexelPremultiplicationOp::Unpremultiply,
1332      float, float>(const float* __restrict src, float* __restrict dst) {
1333   float scaleFactor = src[3] ? 1.0f / src[3] : 1.0f;
1334   dst[0] = src[0] * scaleFactor;
1335   dst[1] = src[1] * scaleFactor;
1336   dst[2] = src[2] * scaleFactor;
1337   dst[3] = src[3];
1338 }
1339
1340 /****** END CODE SHARED WITH WEBKIT ******/
1341
1342 template <typename SrcType, typename DstType>
1343 MOZ_ALWAYS_INLINE void convertType(const SrcType* __restrict src,
1344                                    DstType* __restrict dst) {
1345   MOZ_ASSERT(false, "Unimplemented texture format conversion");
1346 }
1347
1348 template <>
1349 MOZ_ALWAYS_INLINE void convertType<uint8_t, uint8_t>(
1350     const uint8_t* __restrict src, uint8_t* __restrict dst) {
1351   dst[0] = src[0];
1352   dst[1] = src[1];
1353   dst[2] = src[2];
1354   dst[3] = src[3];
1355 }
1356
1357 template <>
1358 MOZ_ALWAYS_INLINE void convertType<uint16_t, uint16_t>(
1359     const uint16_t* __restrict src, uint16_t* __restrict dst) {
1360   dst[0] = src[0];
1361   dst[1] = src[1];
1362   dst[2] = src[2];
1363   dst[3] = src[3];
1364 }
1365
1366 template <>
1367 MOZ_ALWAYS_INLINE void convertType<float, float>(const float* __restrict src,
1368                                                  float* __restrict dst) {
1369   dst[0] = src[0];
1370   dst[1] = src[1];
1371   dst[2] = src[2];
1372   dst[3] = src[3];
1373 }
1374
1375 template <>
1376 MOZ_ALWAYS_INLINE void convertType<uint8_t, float>(
1377     const uint8_t* __restrict src, float* __restrict dst) {
1378   const float scaleFactor = 1.f / 255.0f;
1379   dst[0] = src[0] * scaleFactor;
1380   dst[1] = src[1] * scaleFactor;
1381   dst[2] = src[2] * scaleFactor;
1382   dst[3] = src[3] * scaleFactor;
1383 }
1384
1385 template <>
1386 MOZ_ALWAYS_INLINE void convertType<uint8_t, uint16_t>(
1387     const uint8_t* __restrict src, uint16_t* __restrict dst) {
1388   const float scaleFactor = 1.f / 255.0f;
1389   dst[0] = packToFloat16(src[0] * scaleFactor);
1390   dst[1] = packToFloat16(src[1] * scaleFactor);
1391   dst[2] = packToFloat16(src[2] * scaleFactor);
1392   dst[3] = packToFloat16(src[3] * scaleFactor);
1393 }
1394
1395 }  // end namespace WebGLTexelConversions
1396
1397 }  // end namespace mozilla
1398
1399 #endif  // WEBGLTEXELCONVERSIONS_H_