63cc29304a
This reverts commit e602f39581
.
Reason for revert: Breaks PDFIUM
Original change's description:
> move GrColor.h to private, IWYU
>
> Bug: skia:
> Change-Id: I0f0dabd7cc54cb7786f53bd6da0c0c012375037e
> Reviewed-on: https://skia-review.googlesource.com/104160
> Commit-Queue: Mike Reed <reed@google.com>
> Reviewed-by: Brian Salomon <bsalomon@google.com>
TBR=bsalomon@google.com,reed@google.com
Change-Id: Ifaa50f8771fa1ca8bc152270efdb1fe27f7210f2
No-Presubmit: true
No-Tree-Checks: true
No-Try: true
Bug: skia:
Reviewed-on: https://skia-review.googlesource.com/104440
Reviewed-by: Brian Salomon <bsalomon@google.com>
Commit-Queue: Brian Salomon <bsalomon@google.com>
295 lines
9.2 KiB
C
295 lines
9.2 KiB
C
|
|
/*
|
|
* Copyright 2010 Google Inc.
|
|
*
|
|
* Use of this source code is governed by a BSD-style license that can be
|
|
* found in the LICENSE file.
|
|
*/
|
|
|
|
|
|
|
|
#ifndef GrColor_DEFINED
|
|
#define GrColor_DEFINED
|
|
|
|
#include "GrTypes.h"
|
|
#include "SkColor.h"
|
|
#include "SkColorPriv.h"
|
|
#include "SkUnPreMultiply.h"
|
|
|
|
/**
|
|
* GrColor is 4 bytes for R, G, B, A, in a specific order defined below. Whether the color is
|
|
* premultiplied or not depends on the context in which it is being used.
|
|
*/
|
|
typedef uint32_t GrColor;
|
|
|
|
// shift amount to assign a component to a GrColor int
|
|
// These shift values are chosen for compatibility with GL attrib arrays
|
|
// ES doesn't allow BGRA vertex attrib order so if they were not in this order
|
|
// we'd have to swizzle in shaders.
|
|
#ifdef SK_CPU_BENDIAN
|
|
#define GrColor_SHIFT_R 24
|
|
#define GrColor_SHIFT_G 16
|
|
#define GrColor_SHIFT_B 8
|
|
#define GrColor_SHIFT_A 0
|
|
#else
|
|
#define GrColor_SHIFT_R 0
|
|
#define GrColor_SHIFT_G 8
|
|
#define GrColor_SHIFT_B 16
|
|
#define GrColor_SHIFT_A 24
|
|
#endif
|
|
|
|
/**
|
|
* Pack 4 components (RGBA) into a GrColor int
|
|
*/
|
|
static inline GrColor GrColorPackRGBA(unsigned r, unsigned g, unsigned b, unsigned a) {
|
|
SkASSERT((uint8_t)r == r);
|
|
SkASSERT((uint8_t)g == g);
|
|
SkASSERT((uint8_t)b == b);
|
|
SkASSERT((uint8_t)a == a);
|
|
return (r << GrColor_SHIFT_R) |
|
|
(g << GrColor_SHIFT_G) |
|
|
(b << GrColor_SHIFT_B) |
|
|
(a << GrColor_SHIFT_A);
|
|
}
|
|
|
|
/**
|
|
* Packs a color with an alpha channel replicated across all four channels.
|
|
*/
|
|
static inline GrColor GrColorPackA4(unsigned a) {
|
|
SkASSERT((uint8_t)a == a);
|
|
return (a << GrColor_SHIFT_R) |
|
|
(a << GrColor_SHIFT_G) |
|
|
(a << GrColor_SHIFT_B) |
|
|
(a << GrColor_SHIFT_A);
|
|
}
|
|
|
|
// extract a component (byte) from a GrColor int
|
|
|
|
#define GrColorUnpackR(color) (((color) >> GrColor_SHIFT_R) & 0xFF)
|
|
#define GrColorUnpackG(color) (((color) >> GrColor_SHIFT_G) & 0xFF)
|
|
#define GrColorUnpackB(color) (((color) >> GrColor_SHIFT_B) & 0xFF)
|
|
#define GrColorUnpackA(color) (((color) >> GrColor_SHIFT_A) & 0xFF)
|
|
|
|
/**
|
|
* Since premultiplied means that alpha >= color, we construct a color with
|
|
* each component==255 and alpha == 0 to be "illegal"
|
|
*/
|
|
#define GrColor_ILLEGAL (~(0xFF << GrColor_SHIFT_A))
|
|
|
|
#define GrColor_WHITE 0xFFFFFFFF
|
|
#define GrColor_TRANSPARENT_BLACK 0x0
|
|
|
|
/**
|
|
* Assert in debug builds that a GrColor is premultiplied.
|
|
*/
|
|
static inline void GrColorIsPMAssert(GrColor SkDEBUGCODE(c)) {
|
|
#ifdef SK_DEBUG
|
|
unsigned a = GrColorUnpackA(c);
|
|
unsigned r = GrColorUnpackR(c);
|
|
unsigned g = GrColorUnpackG(c);
|
|
unsigned b = GrColorUnpackB(c);
|
|
|
|
SkASSERT(r <= a);
|
|
SkASSERT(g <= a);
|
|
SkASSERT(b <= a);
|
|
#endif
|
|
}
|
|
|
|
/** Inverts each color channel. */
|
|
static inline GrColor GrInvertColor(GrColor c) {
|
|
U8CPU a = GrColorUnpackA(c);
|
|
U8CPU r = GrColorUnpackR(c);
|
|
U8CPU g = GrColorUnpackG(c);
|
|
U8CPU b = GrColorUnpackB(c);
|
|
return GrColorPackRGBA(0xff - r, 0xff - g, 0xff - b, 0xff - a);
|
|
}
|
|
|
|
static inline GrColor GrColorMul(GrColor c0, GrColor c1) {
|
|
U8CPU r = SkMulDiv255Round(GrColorUnpackR(c0), GrColorUnpackR(c1));
|
|
U8CPU g = SkMulDiv255Round(GrColorUnpackG(c0), GrColorUnpackG(c1));
|
|
U8CPU b = SkMulDiv255Round(GrColorUnpackB(c0), GrColorUnpackB(c1));
|
|
U8CPU a = SkMulDiv255Round(GrColorUnpackA(c0), GrColorUnpackA(c1));
|
|
return GrColorPackRGBA(r, g, b, a);
|
|
}
|
|
|
|
static inline GrColor GrColorSatAdd(GrColor c0, GrColor c1) {
|
|
unsigned r = SkTMin<unsigned>(GrColorUnpackR(c0) + GrColorUnpackR(c1), 0xff);
|
|
unsigned g = SkTMin<unsigned>(GrColorUnpackG(c0) + GrColorUnpackG(c1), 0xff);
|
|
unsigned b = SkTMin<unsigned>(GrColorUnpackB(c0) + GrColorUnpackB(c1), 0xff);
|
|
unsigned a = SkTMin<unsigned>(GrColorUnpackA(c0) + GrColorUnpackA(c1), 0xff);
|
|
return GrColorPackRGBA(r, g, b, a);
|
|
}
|
|
|
|
/** Converts a GrColor to an rgba array of GrGLfloat */
|
|
static inline void GrColorToRGBAFloat(GrColor color, float rgba[4]) {
|
|
static const float ONE_OVER_255 = 1.f / 255.f;
|
|
rgba[0] = GrColorUnpackR(color) * ONE_OVER_255;
|
|
rgba[1] = GrColorUnpackG(color) * ONE_OVER_255;
|
|
rgba[2] = GrColorUnpackB(color) * ONE_OVER_255;
|
|
rgba[3] = GrColorUnpackA(color) * ONE_OVER_255;
|
|
}
|
|
|
|
/** Normalizes and coverts an uint8_t to a float. [0, 255] -> [0.0, 1.0] */
|
|
static inline float GrNormalizeByteToFloat(uint8_t value) {
|
|
static const float ONE_OVER_255 = 1.f / 255.f;
|
|
return value * ONE_OVER_255;
|
|
}
|
|
|
|
/** Determines whether the color is opaque or not. */
|
|
static inline bool GrColorIsOpaque(GrColor color) {
|
|
return (color & (0xFFU << GrColor_SHIFT_A)) == (0xFFU << GrColor_SHIFT_A);
|
|
}
|
|
|
|
static inline GrColor GrPremulColor(GrColor color) {
|
|
unsigned r = GrColorUnpackR(color);
|
|
unsigned g = GrColorUnpackG(color);
|
|
unsigned b = GrColorUnpackB(color);
|
|
unsigned a = GrColorUnpackA(color);
|
|
return GrColorPackRGBA(SkMulDiv255Round(r, a),
|
|
SkMulDiv255Round(g, a),
|
|
SkMulDiv255Round(b, a),
|
|
a);
|
|
}
|
|
|
|
/** Returns an unpremuled version of the GrColor. */
|
|
static inline GrColor GrUnpremulColor(GrColor color) {
|
|
GrColorIsPMAssert(color);
|
|
unsigned r = GrColorUnpackR(color);
|
|
unsigned g = GrColorUnpackG(color);
|
|
unsigned b = GrColorUnpackB(color);
|
|
unsigned a = GrColorUnpackA(color);
|
|
SkPMColor colorPM = SkPackARGB32(a, r, g, b);
|
|
SkColor colorUPM = SkUnPreMultiply::PMColorToColor(colorPM);
|
|
|
|
r = SkColorGetR(colorUPM);
|
|
g = SkColorGetG(colorUPM);
|
|
b = SkColorGetB(colorUPM);
|
|
a = SkColorGetA(colorUPM);
|
|
|
|
return GrColorPackRGBA(r, g, b, a);
|
|
}
|
|
|
|
|
|
/**
|
|
* Similarly, GrColor4f is 4 floats for R, G, B, A, in that order. And like GrColor, whether
|
|
* the color is premultiplied or not depends on the context.
|
|
*/
|
|
struct GrColor4f {
|
|
float fRGBA[4];
|
|
|
|
GrColor4f() {}
|
|
GrColor4f(float r, float g, float b, float a) {
|
|
fRGBA[0] = r;
|
|
fRGBA[1] = g;
|
|
fRGBA[2] = b;
|
|
fRGBA[3] = a;
|
|
}
|
|
|
|
enum Illegal_Constructor {
|
|
kIllegalConstructor
|
|
};
|
|
GrColor4f(Illegal_Constructor) {
|
|
fRGBA[0] = SK_FloatNaN;
|
|
fRGBA[1] = SK_FloatNaN;
|
|
fRGBA[2] = SK_FloatNaN;
|
|
fRGBA[3] = SK_FloatNaN;
|
|
}
|
|
|
|
static GrColor4f OpaqueWhite() {
|
|
return GrColor4f(1.0f, 1.0f, 1.0f, 1.0f);
|
|
}
|
|
|
|
static GrColor4f TransparentBlack() {
|
|
return GrColor4f(0.0f, 0.0f, 0.0f, 0.0f);
|
|
}
|
|
|
|
static GrColor4f FromGrColor(GrColor color) {
|
|
GrColor4f result;
|
|
GrColorToRGBAFloat(color, result.fRGBA);
|
|
return result;
|
|
}
|
|
|
|
static GrColor4f FromSkColor4f(const SkColor4f& color) {
|
|
return GrColor4f(color.fR, color.fG, color.fB, color.fA);
|
|
}
|
|
|
|
GrColor4f modulate(const GrColor4f& x) const {
|
|
return GrColor4f(fRGBA[0] * x.fRGBA[0],
|
|
fRGBA[1] * x.fRGBA[1],
|
|
fRGBA[2] * x.fRGBA[2],
|
|
fRGBA[3] * x.fRGBA[3]);
|
|
}
|
|
|
|
GrColor4f mulByScalar(float x) const {
|
|
return GrColor4f(fRGBA[0] * x, fRGBA[1] * x, fRGBA[2] * x, fRGBA[3] * x);
|
|
}
|
|
|
|
bool operator==(const GrColor4f& other) const {
|
|
return
|
|
fRGBA[0] == other.fRGBA[0] &&
|
|
fRGBA[1] == other.fRGBA[1] &&
|
|
fRGBA[2] == other.fRGBA[2] &&
|
|
fRGBA[3] == other.fRGBA[3];
|
|
}
|
|
bool operator!=(const GrColor4f& other) const {
|
|
return !(*this == other);
|
|
}
|
|
|
|
GrColor toGrColor() const {
|
|
return GrColorPackRGBA(
|
|
SkTPin<unsigned>(static_cast<unsigned>(fRGBA[0] * 255.0f + 0.5f), 0, 255),
|
|
SkTPin<unsigned>(static_cast<unsigned>(fRGBA[1] * 255.0f + 0.5f), 0, 255),
|
|
SkTPin<unsigned>(static_cast<unsigned>(fRGBA[2] * 255.0f + 0.5f), 0, 255),
|
|
SkTPin<unsigned>(static_cast<unsigned>(fRGBA[3] * 255.0f + 0.5f), 0, 255));
|
|
}
|
|
|
|
SkColor4f toSkColor4f() const {
|
|
return SkColor4f { fRGBA[0], fRGBA[1], fRGBA[2], fRGBA[3] };
|
|
}
|
|
|
|
GrColor4f opaque() const {
|
|
return GrColor4f(fRGBA[0], fRGBA[1], fRGBA[2], 1.0f);
|
|
}
|
|
|
|
bool isOpaque() const {
|
|
return fRGBA[3] >= 1.f; // just in case precision causes a superopaque value.
|
|
}
|
|
|
|
GrColor4f premul() const {
|
|
float a = fRGBA[3];
|
|
return GrColor4f(fRGBA[0] * a, fRGBA[1] * a, fRGBA[2] * a, a);
|
|
}
|
|
|
|
GrColor4f unpremul() const {
|
|
float a = fRGBA[3];
|
|
if (a <= 0.0f) {
|
|
return GrColor4f(0.0f, 0.0f, 0.0f, 0.0f);
|
|
}
|
|
float invAlpha = 1.0f / a;
|
|
return GrColor4f(fRGBA[0] * invAlpha, fRGBA[1] * invAlpha, fRGBA[2] * invAlpha, a);
|
|
}
|
|
};
|
|
|
|
/**
|
|
* Flags used for bitfields of color components. They are defined so that the bit order reflects the
|
|
* GrColor shift order.
|
|
*/
|
|
enum GrColorComponentFlags {
|
|
kR_GrColorComponentFlag = 1 << (GrColor_SHIFT_R / 8),
|
|
kG_GrColorComponentFlag = 1 << (GrColor_SHIFT_G / 8),
|
|
kB_GrColorComponentFlag = 1 << (GrColor_SHIFT_B / 8),
|
|
kA_GrColorComponentFlag = 1 << (GrColor_SHIFT_A / 8),
|
|
|
|
kNone_GrColorComponentFlags = 0,
|
|
|
|
kRGB_GrColorComponentFlags = (kR_GrColorComponentFlag | kG_GrColorComponentFlag |
|
|
kB_GrColorComponentFlag),
|
|
|
|
kRGBA_GrColorComponentFlags = (kR_GrColorComponentFlag | kG_GrColorComponentFlag |
|
|
kB_GrColorComponentFlag | kA_GrColorComponentFlag)
|
|
};
|
|
|
|
GR_MAKE_BITFIELD_OPS(GrColorComponentFlags)
|
|
|
|
#endif
|