Source code
Revision control
Copy as Markdown
Other Tools
/* -*- Mode: C++; tab-width: 8; indent-tabs-mode: nil; c-basic-offset: 2 -*- */
/* vim: set ts=8 sts=2 et sw=2 tw=80: */
/* This Source Code Form is subject to the terms of the Mozilla Public
* License, v. 2.0. If a copy of the MPL was not distributed with this
#ifndef MOZILLA_GFX_TYPES_H_
#define MOZILLA_GFX_TYPES_H_
#include "mozilla/DefineEnum.h" // for MOZ_DEFINE_ENUM_CLASS_WITH_BASE
#include "mozilla/EndianUtils.h"
#include "mozilla/EnumeratedRange.h"
#include "mozilla/MacroArgs.h" // for MOZ_CONCAT
#include "mozilla/TypedEnumBits.h"
#include <iosfwd> // for ostream
#include <stddef.h>
#include <stdint.h>
#include <optional>
namespace mozilla {
namespace gfx {
typedef float Float;
typedef double Double;
enum class SurfaceType : int8_t {
DATA, /* Data surface - bitmap in memory */
D2D1_BITMAP, /* Surface wrapping a ID2D1Bitmap */
D2D1_DRAWTARGET, /* Surface made from a D2D draw target */
CAIRO, /* Surface wrapping a cairo surface */
CAIRO_IMAGE, /* Data surface wrapping a cairo image surface */
COREGRAPHICS_IMAGE, /* Surface wrapping a CoreGraphics Image */
COREGRAPHICS_CGCONTEXT, /* Surface wrapping a CG context */
SKIA, /* Surface wrapping a Skia bitmap */
D2D1_1_IMAGE, /* A D2D 1.1 ID2D1Image SourceSurface */
RECORDING, /* Surface used for recording */
DATA_SHARED, /* Data surface using shared memory */
DATA_RECYCLING_SHARED, /* Data surface using shared memory */
OFFSET, /* Offset */
DATA_ALIGNED, /* Data surface using aligned heap memory */
DATA_SHARED_WRAPPER, /* Shared memory mapped in from another process */
BLOB_IMAGE, /* Recorded blob image */
DATA_MAPPED, /* Data surface wrapping a ScopedMap */
WEBGL, /* Surface wrapping a DrawTargetWebgl texture */
};
enum class SurfaceFormat : int8_t {
// The following values are named to reflect layout of colors in memory, from
// lowest byte to highest byte. The 32-bit value layout depends on machine
// endianness.
// in-memory 32-bit LE value 32-bit BE value
B8G8R8A8, // [BB, GG, RR, AA] 0xAARRGGBB 0xBBGGRRAA
B8G8R8X8, // [BB, GG, RR, 00] 0x00RRGGBB 0xBBGGRR00
R8G8B8A8, // [RR, GG, BB, AA] 0xAABBGGRR 0xRRGGBBAA
R8G8B8X8, // [RR, GG, BB, 00] 0x00BBGGRR 0xRRGGBB00
A8R8G8B8, // [AA, RR, GG, BB] 0xBBGGRRAA 0xAARRGGBB
X8R8G8B8, // [00, RR, GG, BB] 0xBBGGRR00 0x00RRGGBB
R8G8B8,
B8G8R8,
// The _UINT16 suffix here indicates that the name reflects the layout when
// viewed as a uint16_t value. In memory these values are stored using native
// endianness.
R5G6B5_UINT16, // 0bRRRRRGGGGGGBBBBB
// This one is a single-byte, so endianness isn't an issue.
A8,
A16,
R8G8,
R16G16,
// These ones are their own special cases.
YUV420, // Sometimes called YU12. 3 planes of 8 bit Y, then Cb, then Cr.
// 4:2:0 chroma subsampling.
YUV422P10, // 3 planes like YUV420, but with 4:2:2 chroma subampling and
// 16 bit plane values where the 6 least significant bits are 0.
NV12, // 2 planes. YUV 4:2:0 image with a plane of 8 bit Y samples
// followed by an interleaved U/V plane containing 8 bit 2x2
// subsampled colour difference samples.
P016, // Similar to NV12, but with 16 bits plane values
P010, // Identical to P016 but the 6 least significant bits are 0.
// With DXGI in theory entirely compatible, however practice has
// shown that it's not the case.
NV16, // Similar to NV12, but with 4:2:2 chroma subsampling. Technically
// 8 bit, but we only use it for 10 bit, and it's really only here
// to support the macOS bi-planar 422 formats.
YUY2, // Sometimes called YUYV. Single plane / packed YUV 4:2:2 8 bit
// samples interleaved as Y`0 Cb Y`1 Cr. Since 4 pixels require
// 64 bits, this can also be considered a 16bpp format, but each
// component is only 8 bits. We sometimes pack RGBA data into
// this format.
HSV,
Lab,
Depth,
// This represents the unknown format.
UNKNOWN, // TODO: Replace uses with Maybe<SurfaceFormat>.
// The following values are endian-independent synonyms. The _UINT32 suffix
// indicates that the name reflects the layout when viewed as a uint32_t
// value.
#if MOZ_LITTLE_ENDIAN()
A8R8G8B8_UINT32 = B8G8R8A8, // 0xAARRGGBB
X8R8G8B8_UINT32 = B8G8R8X8, // 0x00RRGGBB
#elif MOZ_BIG_ENDIAN()
A8R8G8B8_UINT32 = A8R8G8B8, // 0xAARRGGBB
X8R8G8B8_UINT32 = X8R8G8B8, // 0x00RRGGBB
#else
# error "bad endianness"
#endif
// The following values are OS and endian-independent synonyms.
//
// can make this use R8B8G8A8 and R8B8G8X8 for non-Windows platforms.
OS_RGBA = A8R8G8B8_UINT32,
OS_RGBX = X8R8G8B8_UINT32
};
struct SurfaceFormatInfo {
bool hasColor;
bool hasAlpha;
bool isYuv;
std::optional<uint8_t> bytesPerPixel;
};
inline std::optional<SurfaceFormatInfo> Info(const SurfaceFormat aFormat) {
auto info = SurfaceFormatInfo{};
switch (aFormat) {
case SurfaceFormat::B8G8R8A8:
case SurfaceFormat::R8G8B8A8:
case SurfaceFormat::A8R8G8B8:
info.hasColor = true;
info.hasAlpha = true;
break;
case SurfaceFormat::B8G8R8X8:
case SurfaceFormat::R8G8B8X8:
case SurfaceFormat::X8R8G8B8:
case SurfaceFormat::R8G8B8:
case SurfaceFormat::B8G8R8:
case SurfaceFormat::R5G6B5_UINT16:
case SurfaceFormat::R8G8:
case SurfaceFormat::R16G16:
case SurfaceFormat::HSV:
case SurfaceFormat::Lab:
info.hasColor = true;
info.hasAlpha = false;
break;
case SurfaceFormat::A8:
case SurfaceFormat::A16:
info.hasColor = false;
info.hasAlpha = true;
break;
case SurfaceFormat::YUV420:
case SurfaceFormat::YUV422P10:
case SurfaceFormat::NV12:
case SurfaceFormat::P016:
case SurfaceFormat::P010:
case SurfaceFormat::NV16:
case SurfaceFormat::YUY2:
info.hasColor = true;
info.hasAlpha = false;
info.isYuv = true;
break;
case SurfaceFormat::Depth:
info.hasColor = false;
info.hasAlpha = false;
info.isYuv = false;
break;
case SurfaceFormat::UNKNOWN:
break;
}
// -
// bytesPerPixel
switch (aFormat) {
case SurfaceFormat::B8G8R8A8:
case SurfaceFormat::R8G8B8A8:
case SurfaceFormat::A8R8G8B8:
case SurfaceFormat::B8G8R8X8:
case SurfaceFormat::R8G8B8X8:
case SurfaceFormat::X8R8G8B8:
case SurfaceFormat::R16G16:
info.bytesPerPixel = 4;
break;
case SurfaceFormat::R8G8B8:
case SurfaceFormat::B8G8R8:
info.bytesPerPixel = 3;
break;
case SurfaceFormat::R5G6B5_UINT16:
case SurfaceFormat::R8G8:
case SurfaceFormat::A16:
case SurfaceFormat::Depth: // uint16_t
info.bytesPerPixel = 2;
break;
case SurfaceFormat::A8:
info.bytesPerPixel = 1;
break;
case SurfaceFormat::HSV:
case SurfaceFormat::Lab:
info.bytesPerPixel = 3 * sizeof(float);
break;
case SurfaceFormat::YUV420:
case SurfaceFormat::YUV422P10:
case SurfaceFormat::NV12:
case SurfaceFormat::P016:
case SurfaceFormat::P010:
case SurfaceFormat::NV16:
case SurfaceFormat::YUY2:
case SurfaceFormat::UNKNOWN:
break; // No bytesPerPixel per se.
}
// -
if (aFormat == SurfaceFormat::UNKNOWN) {
return {};
}
return info;
}
std::ostream& operator<<(std::ostream& aOut, const SurfaceFormat& aFormat);
// Represents the bit-shifts required to access color channels when the layout
// is viewed as a uint32_t value.
enum class SurfaceFormatBit : uint32_t {
#if MOZ_LITTLE_ENDIAN()
R8G8B8A8_R = 0,
R8G8B8A8_G = 8,
R8G8B8A8_B = 16,
R8G8B8A8_A = 24,
#elif MOZ_BIG_ENDIAN()
R8G8B8A8_A = 0,
R8G8B8A8_B = 8,
R8G8B8A8_G = 16,
R8G8B8A8_R = 24,
#else
# error "bad endianness"
#endif
// The following values are endian-independent for A8R8G8B8_UINT32.
A8R8G8B8_UINT32_B = 0,
A8R8G8B8_UINT32_G = 8,
A8R8G8B8_UINT32_R = 16,
A8R8G8B8_UINT32_A = 24,
// The following values are OS and endian-independent.
//
// can make this use R8G8B8A8_X for non-Windows platforms.
OS_R = A8R8G8B8_UINT32_R,
OS_G = A8R8G8B8_UINT32_G,
OS_B = A8R8G8B8_UINT32_B,
OS_A = A8R8G8B8_UINT32_A,
};
inline uint32_t operator<<(uint8_t a, SurfaceFormatBit b) {
return a << static_cast<uint32_t>(b);
}
inline uint32_t operator>>(uint32_t a, SurfaceFormatBit b) {
return a >> static_cast<uint32_t>(b);
}
static inline int BytesPerPixel(SurfaceFormat aFormat) {
// TODO: return Info(aFormat).value().bytesPerPixel.value();
switch (aFormat) {
case SurfaceFormat::A8:
return 1;
case SurfaceFormat::R5G6B5_UINT16:
case SurfaceFormat::A16:
return 2;
case SurfaceFormat::R8G8B8:
case SurfaceFormat::B8G8R8:
return 3;
case SurfaceFormat::HSV:
case SurfaceFormat::Lab:
return 3 * sizeof(float);
case SurfaceFormat::Depth:
return sizeof(uint16_t);
default:
return 4;
}
}
inline bool IsOpaque(SurfaceFormat aFormat) {
// TODO: return Info(aFormat).value().hasAlpha;
switch (aFormat) {
case SurfaceFormat::B8G8R8X8:
case SurfaceFormat::R8G8B8X8:
case SurfaceFormat::X8R8G8B8:
case SurfaceFormat::R5G6B5_UINT16:
case SurfaceFormat::R8G8B8:
case SurfaceFormat::B8G8R8:
case SurfaceFormat::R8G8:
case SurfaceFormat::HSV:
case SurfaceFormat::Lab:
case SurfaceFormat::Depth:
case SurfaceFormat::YUV420:
case SurfaceFormat::NV12:
case SurfaceFormat::P010:
case SurfaceFormat::P016:
case SurfaceFormat::YUY2:
return true;
default:
return false;
}
}
// These are standardized Coding-independent Code Points
// See [Rec. ITU-T H.273
//
// We deliberately use an unscoped enum with fixed uint8_t representation since
// all possible values [0, 255] are legal, but it's unwieldy to declare 200+
// "RESERVED" enumeration values. Having a fixed underlying type avoids any
// potential UB and avoids the need for a cast when passing these values across
// FFI to functions like qcms_profile_create_cicp.
namespace CICP {
enum ColourPrimaries : uint8_t {
CP_RESERVED_MIN = 0, // 0, 3, [13, 21], [23, 255] are all reserved
CP_BT709 = 1,
CP_UNSPECIFIED = 2,
CP_BT470M = 4,
CP_BT470BG = 5,
CP_BT601 = 6,
CP_SMPTE240 = 7,
CP_GENERIC_FILM = 8,
CP_BT2020 = 9,
CP_XYZ = 10,
CP_SMPTE431 = 11,
CP_SMPTE432 = 12,
CP_EBU3213 = 22,
};
inline bool IsReserved(ColourPrimaries aIn) {
switch (aIn) {
case CP_BT709:
case CP_UNSPECIFIED:
case CP_BT470M:
case CP_BT470BG:
case CP_BT601:
case CP_SMPTE240:
case CP_GENERIC_FILM:
case CP_BT2020:
case CP_XYZ:
case CP_SMPTE431:
case CP_SMPTE432:
case CP_EBU3213:
return false;
default:
return true;
}
}
enum TransferCharacteristics : uint8_t {
TC_RESERVED_MIN = 0, // 0, 3, [19, 255] are all reserved
TC_BT709 = 1,
TC_UNSPECIFIED = 2,
TC_BT470M = 4,
TC_BT470BG = 5,
TC_BT601 = 6,
TC_SMPTE240 = 7,
TC_LINEAR = 8,
TC_LOG_100 = 9,
TC_LOG_100_SQRT10 = 10,
TC_IEC61966 = 11,
TC_BT_1361 = 12,
TC_SRGB = 13,
TC_BT2020_10BIT = 14,
TC_BT2020_12BIT = 15,
TC_SMPTE2084 = 16,
TC_SMPTE428 = 17,
TC_HLG = 18,
};
inline bool IsReserved(TransferCharacteristics aIn) {
switch (aIn) {
case TC_BT709:
case TC_UNSPECIFIED:
case TC_BT470M:
case TC_BT470BG:
case TC_BT601:
case TC_SMPTE240:
case TC_LINEAR:
case TC_LOG_100:
case TC_LOG_100_SQRT10:
case TC_IEC61966:
case TC_BT_1361:
case TC_SRGB:
case TC_BT2020_10BIT:
case TC_BT2020_12BIT:
case TC_SMPTE2084:
case TC_SMPTE428:
case TC_HLG:
return false;
default:
return true;
}
}
enum MatrixCoefficients : uint8_t {
MC_IDENTITY = 0,
MC_BT709 = 1,
MC_UNSPECIFIED = 2,
MC_RESERVED_MIN = 3, // 3, [15, 255] are all reserved
MC_FCC = 4,
MC_BT470BG = 5,
MC_BT601 = 6,
MC_SMPTE240 = 7,
MC_YCGCO = 8,
MC_BT2020_NCL = 9,
MC_BT2020_CL = 10,
MC_SMPTE2085 = 11,
MC_CHROMAT_NCL = 12,
MC_CHROMAT_CL = 13,
MC_ICTCP = 14,
};
inline bool IsReserved(MatrixCoefficients aIn) {
switch (aIn) {
case MC_IDENTITY:
case MC_BT709:
case MC_UNSPECIFIED:
case MC_RESERVED_MIN:
case MC_FCC:
case MC_BT470BG:
case MC_BT601:
case MC_SMPTE240:
case MC_YCGCO:
case MC_BT2020_NCL:
case MC_BT2020_CL:
case MC_SMPTE2085:
case MC_CHROMAT_NCL:
case MC_CHROMAT_CL:
case MC_ICTCP:
return false;
default:
return true;
}
}
} // namespace CICP
// The matrix coeffiecients used for YUV to RGB conversion.
enum class YUVColorSpace : uint8_t {
BT601,
BT709,
BT2020,
Identity, // Todo: s/YUVColorSpace/ColorSpace/, s/Identity/SRGB/
Default = BT709,
_First = BT601,
_Last = Identity,
};
enum class ColorDepth : uint8_t {
COLOR_8,
COLOR_10,
COLOR_12,
COLOR_16,
_First = COLOR_8,
_Last = COLOR_16,
};
enum class TransferFunction : uint8_t {
BT709,
SRGB,
PQ,
HLG,
_First = BT709,
_Last = HLG,
Default = BT709,
};
enum class ColorRange : uint8_t {
LIMITED,
FULL,
_First = LIMITED,
_Last = FULL,
};
// Really "YcbcrColorColorSpace"
enum class YUVRangedColorSpace : uint8_t {
BT601_Narrow = 0,
BT601_Full,
BT709_Narrow,
BT709_Full,
BT2020_Narrow,
BT2020_Full,
GbrIdentity,
_First = BT601_Narrow,
_Last = GbrIdentity,
Default = BT709_Narrow,
};
// I can either come up with a longer "very clever" name that doesn't conflict
// with FilterSupport.h, embrace and expand FilterSupport, or rename the old
// one.
// Some times Worse Is Better.
enum class ColorSpace2 : uint8_t {
Display,
UNKNOWN = Display, // We feel sufficiently bad about this TODO.
SRGB,
DISPLAY_P3,
BT601_525, // aka smpte170m NTSC
BT709, // Same gamut as SRGB, but different gamma.
BT601_625 =
BT709, // aka bt470bg PAL. Basically BT709, just Xg is 0.290 not 0.300.
BT2020,
_First = Display,
_Last = BT2020,
};
inline ColorSpace2 ToColorSpace2(const YUVColorSpace in) {
switch (in) {
case YUVColorSpace::BT601:
return ColorSpace2::BT601_525;
case YUVColorSpace::BT709:
return ColorSpace2::BT709;
case YUVColorSpace::BT2020:
return ColorSpace2::BT2020;
case YUVColorSpace::Identity:
return ColorSpace2::SRGB;
}
MOZ_ASSERT_UNREACHABLE();
}
inline YUVColorSpace ToYUVColorSpace(const ColorSpace2 in) {
switch (in) {
case ColorSpace2::BT601_525:
return YUVColorSpace::BT601;
case ColorSpace2::BT709:
return YUVColorSpace::BT709;
case ColorSpace2::BT2020:
return YUVColorSpace::BT2020;
case ColorSpace2::SRGB:
return YUVColorSpace::Identity;
case ColorSpace2::UNKNOWN:
case ColorSpace2::DISPLAY_P3:
MOZ_CRASH("Bad ColorSpace2 for ToYUVColorSpace");
}
MOZ_ASSERT_UNREACHABLE();
}
struct FromYUVRangedColorSpaceT final {
const YUVColorSpace space;
const ColorRange range;
};
inline FromYUVRangedColorSpaceT FromYUVRangedColorSpace(
const YUVRangedColorSpace s) {
switch (s) {
case YUVRangedColorSpace::BT601_Narrow:
return {YUVColorSpace::BT601, ColorRange::LIMITED};
case YUVRangedColorSpace::BT601_Full:
return {YUVColorSpace::BT601, ColorRange::FULL};
case YUVRangedColorSpace::BT709_Narrow:
return {YUVColorSpace::BT709, ColorRange::LIMITED};
case YUVRangedColorSpace::BT709_Full:
return {YUVColorSpace::BT709, ColorRange::FULL};
case YUVRangedColorSpace::BT2020_Narrow:
return {YUVColorSpace::BT2020, ColorRange::LIMITED};
case YUVRangedColorSpace::BT2020_Full:
return {YUVColorSpace::BT2020, ColorRange::FULL};
case YUVRangedColorSpace::GbrIdentity:
return {YUVColorSpace::Identity, ColorRange::FULL};
}
MOZ_CRASH("bad YUVRangedColorSpace");
}
// Todo: This should go in the CPP.
inline YUVRangedColorSpace ToYUVRangedColorSpace(const YUVColorSpace space,
const ColorRange range) {
bool narrow;
switch (range) {
case ColorRange::FULL:
narrow = false;
break;
case ColorRange::LIMITED:
narrow = true;
break;
}
switch (space) {
case YUVColorSpace::Identity:
MOZ_ASSERT(range == ColorRange::FULL);
return YUVRangedColorSpace::GbrIdentity;
case YUVColorSpace::BT601:
return narrow ? YUVRangedColorSpace::BT601_Narrow
: YUVRangedColorSpace::BT601_Full;
case YUVColorSpace::BT709:
return narrow ? YUVRangedColorSpace::BT709_Narrow
: YUVRangedColorSpace::BT709_Full;
case YUVColorSpace::BT2020:
return narrow ? YUVRangedColorSpace::BT2020_Narrow
: YUVRangedColorSpace::BT2020_Full;
}
MOZ_CRASH("bad YUVColorSpace");
}
template <typename DescriptorT>
inline YUVRangedColorSpace GetYUVRangedColorSpace(const DescriptorT& d) {
return ToYUVRangedColorSpace(d.yUVColorSpace(), d.colorRange());
}
static inline SurfaceFormat SurfaceFormatForColorDepth(ColorDepth aColorDepth) {
SurfaceFormat format = SurfaceFormat::A8;
switch (aColorDepth) {
case ColorDepth::COLOR_8:
break;
case ColorDepth::COLOR_10:
case ColorDepth::COLOR_12:
case ColorDepth::COLOR_16:
format = SurfaceFormat::A16;
break;
}
return format;
}
static inline uint8_t BitDepthForColorDepth(ColorDepth aColorDepth) {
uint8_t depth = 8;
switch (aColorDepth) {
case ColorDepth::COLOR_8:
break;
case ColorDepth::COLOR_10:
depth = 10;
break;
case ColorDepth::COLOR_12:
depth = 12;
break;
case ColorDepth::COLOR_16:
depth = 16;
break;
}
return depth;
}
static inline ColorDepth ColorDepthForBitDepth(uint8_t aBitDepth) {
ColorDepth depth = ColorDepth::COLOR_8;
switch (aBitDepth) {
case 8:
break;
case 10:
depth = ColorDepth::COLOR_10;
break;
case 12:
depth = ColorDepth::COLOR_12;
break;
case 16:
depth = ColorDepth::COLOR_16;
break;
}
return depth;
}
// 10 and 12 bits color depth image are using 16 bits integers for storage
// As such we need to rescale the value from 10 or 12 bits to 16.
static inline uint32_t RescalingFactorForColorDepth(ColorDepth aColorDepth) {
uint32_t factor = 1;
switch (aColorDepth) {
case ColorDepth::COLOR_8:
break;
case ColorDepth::COLOR_10:
factor = 64;
break;
case ColorDepth::COLOR_12:
factor = 16;
break;
case ColorDepth::COLOR_16:
break;
}
return factor;
}
enum class ChromaSubsampling : uint8_t {
FULL,
HALF_WIDTH,
HALF_WIDTH_AND_HEIGHT,
_First = FULL,
_Last = HALF_WIDTH_AND_HEIGHT,
};
template <typename T>
static inline T ChromaSize(const T& aYSize, ChromaSubsampling aSubsampling) {
switch (aSubsampling) {
case ChromaSubsampling::FULL:
return aYSize;
case ChromaSubsampling::HALF_WIDTH:
return T((aYSize.width + 1) / 2, aYSize.height);
case ChromaSubsampling::HALF_WIDTH_AND_HEIGHT:
return T((aYSize.width + 1) / 2, (aYSize.height + 1) / 2);
}
MOZ_CRASH("bad ChromaSubsampling");
}
enum class FilterType : int8_t {
BLEND = 0,
TRANSFORM,
MORPHOLOGY,
COLOR_MATRIX,
FLOOD,
TILE,
TABLE_TRANSFER,
DISCRETE_TRANSFER,
LINEAR_TRANSFER,
GAMMA_TRANSFER,
CONVOLVE_MATRIX,
DISPLACEMENT_MAP,
TURBULENCE,
ARITHMETIC_COMBINE,
COMPOSITE,
DIRECTIONAL_BLUR,
GAUSSIAN_BLUR,
POINT_DIFFUSE,
POINT_SPECULAR,
SPOT_DIFFUSE,
SPOT_SPECULAR,
DISTANT_DIFFUSE,
DISTANT_SPECULAR,
CROP,
PREMULTIPLY,
UNPREMULTIPLY,
OPACITY
};
enum class DrawTargetType : int8_t {
SOFTWARE_RASTER = 0,
HARDWARE_RASTER,
VECTOR
};
enum class BackendType : int8_t {
NONE = 0,
DIRECT2D, // Used for version independent D2D objects.
CAIRO,
SKIA,
RECORDING,
DIRECT2D1_1,
WEBRENDER_TEXT,
WEBGL,
// Add new entries above this line.
BACKEND_LAST
};
enum class RecorderType : int8_t {
UNKNOWN,
PRIVATE,
MEMORY,
CANVAS,
PRFILEDESC,
WEBRENDER
};
enum class FontType : int8_t {
DWRITE,
GDI,
MAC,
FONTCONFIG,
FREETYPE,
UNKNOWN
};
enum class NativeSurfaceType : int8_t {
D3D10_TEXTURE,
CAIRO_CONTEXT,
CGCONTEXT,
CGCONTEXT_ACCELERATED,
OPENGL_TEXTURE,
WEBGL_CONTEXT
};
enum class FontStyle : int8_t { NORMAL, ITALIC, BOLD, BOLD_ITALIC };
enum class FontHinting : int8_t { NONE, LIGHT, NORMAL, FULL };
enum class CompositionOp : int8_t {
OP_CLEAR,
OP_OVER,
OP_ADD,
OP_ATOP,
OP_OUT,
OP_IN,
OP_SOURCE,
OP_DEST_IN,
OP_DEST_OUT,
OP_DEST_OVER,
OP_DEST_ATOP,
OP_XOR,
OP_MULTIPLY,
OP_SCREEN,
OP_OVERLAY,
OP_DARKEN,
OP_LIGHTEN,
OP_COLOR_DODGE,
OP_COLOR_BURN,
OP_HARD_LIGHT,
OP_SOFT_LIGHT,
OP_DIFFERENCE,
OP_EXCLUSION,
OP_HUE,
OP_SATURATION,
OP_COLOR,
OP_LUMINOSITY,
OP_COUNT
};
enum class Axis : int8_t { X_AXIS, Y_AXIS, BOTH };
enum class ExtendMode : int8_t {
CLAMP, // Do not repeat
REPEAT, // Repeat in both axis
REPEAT_X, // Only X axis
REPEAT_Y, // Only Y axis
REFLECT // Mirror the image
};
enum class FillRule : int8_t { FILL_WINDING, FILL_EVEN_ODD };
enum class AntialiasMode : int8_t { NONE, GRAY, SUBPIXEL, DEFAULT };
enum class SamplingFilter : int8_t {
GOOD,
LINEAR,
POINT,
SENTINEL // one past the last valid value
};
std::ostream& operator<<(std::ostream& aOut, const SamplingFilter& aFilter);
// clang-format off
MOZ_DEFINE_ENUM_CLASS_WITH_BASE(PatternType, int8_t, (
COLOR,
SURFACE,
LINEAR_GRADIENT,
RADIAL_GRADIENT,
CONIC_GRADIENT
));
// clang-format on
enum class JoinStyle : int8_t {
BEVEL,
ROUND,
MITER, //!< Mitered if within the miter limit, else, if the backed supports
//!< it (D2D), the miter is clamped. If the backend does not support
//!< miter clamping the behavior is as for MITER_OR_BEVEL.
MITER_OR_BEVEL //!< Mitered if within the miter limit, else beveled.
};
enum class CapStyle : int8_t { BUTT, ROUND, SQUARE };
enum class SamplingBounds : int8_t { UNBOUNDED, BOUNDED };
// Moz2d version for SVG mask types
enum class LuminanceType : int8_t {
LUMINANCE,
LINEARRGB,
};
/* Color is stored in non-premultiplied form in sRGB color space */
struct sRGBColor {
public:
constexpr sRGBColor() : r(0.0f), g(0.0f), b(0.0f), a(0.0f) {}
constexpr sRGBColor(Float aR, Float aG, Float aB, Float aA)
: r(aR), g(aG), b(aB), a(aA) {}
constexpr sRGBColor(Float aR, Float aG, Float aB)
: r(aR), g(aG), b(aB), a(1.0f) {}
static constexpr sRGBColor White(float aA) {
return sRGBColor(1.f, 1.f, 1.f, aA);
}
static constexpr sRGBColor Black(float aA) {
return sRGBColor(0.f, 0.f, 0.f, aA);
}
static constexpr sRGBColor OpaqueWhite() { return White(1.f); }
static constexpr sRGBColor OpaqueBlack() { return Black(1.f); }
static constexpr sRGBColor FromU8(uint8_t aR, uint8_t aG, uint8_t aB,
uint8_t aA) {
return sRGBColor(float(aR) / 255.f, float(aG) / 255.f, float(aB) / 255.f,
float(aA) / 255.f);
}
static constexpr sRGBColor FromABGR(uint32_t aColor) {
return sRGBColor(((aColor >> 0) & 0xff) * (1.0f / 255.0f),
((aColor >> 8) & 0xff) * (1.0f / 255.0f),
((aColor >> 16) & 0xff) * (1.0f / 255.0f),
((aColor >> 24) & 0xff) * (1.0f / 255.0f));
}
// The "Unusual" prefix is to avoid unintentionally using this function when
// FromABGR(), which is much more common, is needed.
static constexpr sRGBColor UnusualFromARGB(uint32_t aColor) {
return sRGBColor(((aColor >> 16) & 0xff) * (1.0f / 255.0f),
((aColor >> 8) & 0xff) * (1.0f / 255.0f),
((aColor >> 0) & 0xff) * (1.0f / 255.0f),
((aColor >> 24) & 0xff) * (1.0f / 255.0f));
}
constexpr uint32_t ToABGR() const {
return uint32_t(r * 255.0f) | uint32_t(g * 255.0f) << 8 |
uint32_t(b * 255.0f) << 16 | uint32_t(a * 255.0f) << 24;
}
constexpr sRGBColor Premultiplied() const {
return sRGBColor(r * a, g * a, b * a, a);
}
constexpr sRGBColor Unpremultiplied() const {
return a > 0.f ? sRGBColor(r / a, g / a, b / a, a) : *this;
}
// The "Unusual" prefix is to avoid unintentionally using this function when
// ToABGR(), which is much more common, is needed.
uint32_t UnusualToARGB() const {
return uint32_t(b * 255.0f) | uint32_t(g * 255.0f) << 8 |
uint32_t(r * 255.0f) << 16 | uint32_t(a * 255.0f) << 24;
}
bool operator==(const sRGBColor& aColor) const {
return r == aColor.r && g == aColor.g && b == aColor.b && a == aColor.a;
}
bool operator!=(const sRGBColor& aColor) const { return !(*this == aColor); }
Float r, g, b, a;
};
/* Color is stored in non-premultiplied form in device color space */
struct DeviceColor {
public:
constexpr DeviceColor() : r(0.0f), g(0.0f), b(0.0f), a(0.0f) {}
constexpr DeviceColor(Float aR, Float aG, Float aB, Float aA)
: r(aR), g(aG), b(aB), a(aA) {}
constexpr DeviceColor(Float aR, Float aG, Float aB)
: r(aR), g(aG), b(aB), a(1.0f) {}
/* The following Mask* variants are helpers used to make it clear when a
* particular color is being used for masking purposes. These masks should
* never be colored managed. */
static DeviceColor Mask(float aC, float aA) {
return DeviceColor(aC, aC, aC, aA);
}
static DeviceColor MaskWhite(float aA) { return Mask(1.f, aA); }
static DeviceColor MaskBlack(float aA) { return Mask(0.f, aA); }
static DeviceColor MaskOpaqueWhite() { return MaskWhite(1.f); }
static DeviceColor MaskOpaqueBlack() { return MaskBlack(1.f); }
static DeviceColor FromU8(uint8_t aR, uint8_t aG, uint8_t aB, uint8_t aA) {
return DeviceColor(float(aR) / 255.f, float(aG) / 255.f, float(aB) / 255.f,
float(aA) / 255.f);
}
static DeviceColor FromABGR(uint32_t aColor) {
DeviceColor newColor(((aColor >> 0) & 0xff) * (1.0f / 255.0f),
((aColor >> 8) & 0xff) * (1.0f / 255.0f),
((aColor >> 16) & 0xff) * (1.0f / 255.0f),
((aColor >> 24) & 0xff) * (1.0f / 255.0f));
return newColor;
}
// The "Unusual" prefix is to avoid unintentionally using this function when
// FromABGR(), which is much more common, is needed.
static DeviceColor UnusualFromARGB(uint32_t aColor) {
DeviceColor newColor(((aColor >> 16) & 0xff) * (1.0f / 255.0f),
((aColor >> 8) & 0xff) * (1.0f / 255.0f),
((aColor >> 0) & 0xff) * (1.0f / 255.0f),
((aColor >> 24) & 0xff) * (1.0f / 255.0f));
return newColor;
}
uint32_t ToABGR() const {
return uint32_t(r * 255.0f) | uint32_t(g * 255.0f) << 8 |
uint32_t(b * 255.0f) << 16 | uint32_t(a * 255.0f) << 24;
}
// The "Unusual" prefix is to avoid unintentionally using this function when
// ToABGR(), which is much more common, is needed.
uint32_t UnusualToARGB() const {
return uint32_t(b * 255.0f) | uint32_t(g * 255.0f) << 8 |
uint32_t(r * 255.0f) << 16 | uint32_t(a * 255.0f) << 24;
}
bool operator==(const DeviceColor& aColor) const {
return r == aColor.r && g == aColor.g && b == aColor.b && a == aColor.a;
}
bool operator!=(const DeviceColor& aColor) const {
return !(*this == aColor);
}
friend std::ostream& operator<<(std::ostream& aOut,
const DeviceColor& aColor);
Float r, g, b, a;
};
struct GradientStop {
bool operator<(const GradientStop& aOther) const {
return offset < aOther.offset;
}
Float offset;
DeviceColor color;
};
enum class JobStatus { Complete, Wait, Yield, Error };
enum class DeviceResetReason {
OK = 0, // No reset.
HUNG, // Windows specific, guilty device reset.
REMOVED, // Windows specific, device removed or driver upgraded.
RESET, // Guilty device reset.
DRIVER_ERROR, // Innocent device reset.
INVALID_CALL, // Windows specific, guilty device reset.
OUT_OF_MEMORY,
FORCED_RESET, // Simulated device reset.
OTHER, // Unrecognized reason for device reset.
NVIDIA_VIDEO, // Linux specific, NVIDIA video memory was reset.
UNKNOWN, // GL specific, unknown if guilty or innocent.
_First = OK,
_Last = UNKNOWN,
};
enum class DeviceResetDetectPlace {
WR_BEGIN_FRAME = 0,
WR_WAIT_FOR_GPU,
WR_POST_UPDATE,
WR_SYNC_OBJRCT,
WR_SIMULATE,
WIDGET,
CANVAS_TRANSLATOR,
_First = WR_BEGIN_FRAME,
_Last = CANVAS_TRANSLATOR,
};
enum class ForcedDeviceResetReason {
OPENSHAREDHANDLE = 0,
COMPOSITOR_UPDATED,
};
} // namespace gfx
} // namespace mozilla
// XXX: temporary
typedef mozilla::gfx::SurfaceFormat gfxImageFormat;
#if defined(XP_WIN) && defined(MOZ_GFX)
# ifdef GFX2D_INTERNAL
# define GFX2D_API __declspec(dllexport)
# else
# define GFX2D_API __declspec(dllimport)
# endif
#else
# define GFX2D_API
#endif
namespace mozilla {
// Side constants for use in various places.
enum Side : uint8_t { eSideTop, eSideRight, eSideBottom, eSideLeft };
std::ostream& operator<<(std::ostream&, const mozilla::Side&);
constexpr auto AllPhysicalSides() {
return mozilla::MakeInclusiveEnumeratedRange(eSideTop, eSideLeft);
}
enum class SideBits {
eNone = 0,
eTop = 1 << eSideTop,
eRight = 1 << eSideRight,
eBottom = 1 << eSideBottom,
eLeft = 1 << eSideLeft,
eTopBottom = SideBits::eTop | SideBits::eBottom,
eLeftRight = SideBits::eLeft | SideBits::eRight,
eAll = SideBits::eTopBottom | SideBits::eLeftRight
};
MOZ_MAKE_ENUM_CLASS_BITWISE_OPERATORS(SideBits)
inline constexpr SideBits SideToSideBit(mozilla::Side aSide) {
return SideBits(1 << aSide);
}
enum Corner : uint8_t {
// This order is important!
eCornerTopLeft = 0,
eCornerTopRight = 1,
eCornerBottomRight = 2,
eCornerBottomLeft = 3
};
// RectCornerRadii::radii depends on this value. It is not being added to
// Corner because we want to lift the responsibility to handle it in the
// switch-case.
constexpr int eCornerCount = 4;
constexpr auto AllPhysicalCorners() {
return mozilla::MakeInclusiveEnumeratedRange(eCornerTopLeft,
eCornerBottomLeft);
}
// Indices into "half corner" arrays (nsStyleCorners e.g.)
enum HalfCorner : uint8_t {
// This order is important!
eCornerTopLeftX = 0,
eCornerTopLeftY = 1,
eCornerTopRightX = 2,
eCornerTopRightY = 3,
eCornerBottomRightX = 4,
eCornerBottomRightY = 5,
eCornerBottomLeftX = 6,
eCornerBottomLeftY = 7
};
constexpr auto AllPhysicalHalfCorners() {
return mozilla::MakeInclusiveEnumeratedRange(eCornerTopLeftX,
eCornerBottomLeftY);
}
// The result of these conversion functions are exhaustively checked in
// nsFrame.cpp, which also serves as usage examples.
constexpr bool HalfCornerIsX(HalfCorner aHalfCorner) {
return !(aHalfCorner % 2);
}
constexpr Corner HalfToFullCorner(HalfCorner aHalfCorner) {
return Corner(aHalfCorner / 2);
}
constexpr HalfCorner FullToHalfCorner(Corner aCorner, bool aIsVertical) {
return HalfCorner(aCorner * 2 + aIsVertical);
}
constexpr bool SideIsVertical(mozilla::Side aSide) { return aSide % 2; }
// @param aIsSecond when true, return the clockwise second of the two
// corners associated with aSide. For example, with aSide = eSideBottom the
// result is eCornerBottomRight when aIsSecond is false, and
// eCornerBottomLeft when aIsSecond is true.
constexpr Corner SideToFullCorner(mozilla::Side aSide, bool aIsSecond) {
return Corner((aSide + aIsSecond) % 4);
}
// @param aIsSecond see SideToFullCorner.
// @param aIsParallel return the half-corner that is parallel with aSide
// when aIsParallel is true. For example with aSide=eSideTop, aIsSecond=true
// the result is eCornerTopRightX when aIsParallel is true, and
// eCornerTopRightY when aIsParallel is false (because "X" is parallel with
// eSideTop/eSideBottom, similarly "Y" is parallel with
// eSideLeft/eSideRight)
constexpr HalfCorner SideToHalfCorner(mozilla::Side aSide, bool aIsSecond,
bool aIsParallel) {
return HalfCorner(((aSide + aIsSecond) * 2 + (aSide + !aIsParallel) % 2) % 8);
}
} // namespace mozilla
#endif /* MOZILLA_GFX_TYPES_H_ */