android_frameworks_base/libs/hwui/ProgramCache.cpp

866 lines
34 KiB
C++
Raw Normal View History

/*
* Copyright (C) 2010 The Android Open Source Project
*
* Licensed under the Apache License, Version 2.0 (the "License");
* you may not use this file except in compliance with the License.
* You may obtain a copy of the License at
*
* http://www.apache.org/licenses/LICENSE-2.0
*
* Unless required by applicable law or agreed to in writing, software
* distributed under the License is distributed on an "AS IS" BASIS,
* WITHOUT WARRANTIES OR CONDITIONS OF ANY KIND, either express or implied.
* See the License for the specific language governing permissions and
* limitations under the License.
*/
#include <utils/String8.h>
#include "Caches.h"
#include "ProgramCache.h"
Linear blending, step 1 NOTE: Linear blending is currently disabled in this CL as the feature is still a work in progress Android currently performs all blending (any kind of linear math on colors really) on gamma-encoded colors. Since Android assumes that the default color space is sRGB, all bitmaps and colors are encoded with the sRGB Opto-Electronic Conversion Function (OECF, which can be approximated with a power function). Since the power curve is not linear, our linear math is incorrect. The result is that we generate colors that tend to be too dark; this affects blending but also anti-aliasing, gradients, blurs, etc. The solution is to convert gamma-encoded colors back to linear space before doing any math on them, using the sRGB Electo-Optical Conversion Function (EOCF). This is achieved in different ways in different parts of the pipeline: - Using hardware conversions when sampling from OpenGL textures or writing into OpenGL frame buffers - Using software conversion functions, to translate app-supplied colors to and from sRGB - Using Skia's color spaces Any type of processing on colors must roughly ollow these steps: [sRGB input]->EOCF->[linear data]->[processing]->OECF->[sRGB output] For the sRGB color space, the conversion functions are defined as follows: OECF(linear) := linear <= 0.0031308 ? linear * 12.92 : (pow(linear, 1/2.4) * 1.055) - 0.055 EOCF(srgb) := srgb <= 0.04045 ? srgb / 12.92 : pow((srgb + 0.055) / 1.055, 2.4) The EOCF is simply the reciprocal of the OECF. While it is highly recommended to use the exact sRGB conversion functions everywhere possible, it is sometimes useful or beneficial to rely on approximations: - pow(x,2.2) and pow(x,1/2.2) - x^2 and sqrt(x) The latter is particularly useful in fragment shaders (for instance to apply dithering in sRGB space), especially if the sqrt() can be replaced with an inversesqrt(). Here is a fairly exhaustive list of modifications implemented in this CL: - Set TARGET_ENABLE_LINEAR_BLENDING := false in BoardConfig.mk to disable linear blending. This is only for GLES 2.0 GPUs with no hardware sRGB support. This flag is currently assumed to be false (see note above) - sRGB writes are disabled when entering a functor (WebView). This will need to be fixed at some point - Skia bitmaps are created with the sRGB color space - Bitmaps using a 565 config are expanded to 888 - Linear blending is disabled when entering a functor - External textures are not properly sampled (see below) - Gradients are interpolated in linear space - Texture-based dithering was replaced with analytical dithering - Dithering is done in the quantization color space, which is why we must do EOCF(OECF(color)+dither) - Text is now gamma corrected differently depending on the luminance of the source pixel. The asumption is that a bright pixel will be blended on a dark background and the other way around. The source alpha is gamma corrected to thicken dark on bright and thin bright on dark to match the intended design of fonts. This also matches the behavior of popular design/drawing applications - Removed the asset atlas. It did not contain anything useful and could not be sampled in sRGB without a yet-to-be-defined GL extension - The last column of color matrices is converted to linear space because its value are added to linear colors Missing features: - Resource qualifier? - Regeneration of goldeng images for automated tests - Handle alpha8/grey8 properly - Disable sRGB write for layers with external textures Test: Manual testing while work in progress Bug: 29940137 Change-Id: I6a07b15ab49b554377cd33a36b6d9971a15e9a0b
2016-09-28 17:34:42 -07:00
#include "Properties.h"
namespace android {
namespace uirenderer {
///////////////////////////////////////////////////////////////////////////////
// Defines
///////////////////////////////////////////////////////////////////////////////
#define MODULATE_OP_NO_MODULATE 0
#define MODULATE_OP_MODULATE 1
#define MODULATE_OP_MODULATE_A8 2
#define STR(x) STR1(x)
#define STR1(x) #x
///////////////////////////////////////////////////////////////////////////////
// Vertex shaders snippets
///////////////////////////////////////////////////////////////////////////////
const char* gVS_Header_Start =
"#version 100\n"
"attribute vec4 position;\n";
const char* gVS_Header_Attributes_TexCoords = "attribute vec2 texCoords;\n";
const char* gVS_Header_Attributes_Colors = "attribute vec4 colors;\n";
const char* gVS_Header_Attributes_VertexAlphaParameters = "attribute float vtxAlpha;\n";
const char* gVS_Header_Uniforms_TextureTransform = "uniform mat4 mainTextureTransform;\n";
const char* gVS_Header_Uniforms =
"uniform mat4 projection;\n"
"uniform mat4 transform;\n";
const char* gVS_Header_Uniforms_HasGradient = "uniform mat4 screenSpace;\n";
const char* gVS_Header_Uniforms_HasBitmap =
"uniform mat4 textureTransform;\n"
"uniform mediump vec2 textureDimension;\n";
const char* gVS_Header_Uniforms_HasRoundRectClip =
"uniform mat4 roundRectInvTransform;\n"
"uniform mediump vec4 roundRectInnerRectLTWH;\n"
"uniform mediump float roundRectRadius;\n";
const char* gVS_Header_Varyings_HasTexture = "varying vec2 outTexCoords;\n";
const char* gVS_Header_Varyings_HasColors = "varying vec4 outColors;\n";
const char* gVS_Header_Varyings_HasVertexAlpha = "varying float alpha;\n";
const char* gVS_Header_Varyings_HasBitmap = "varying highp vec2 outBitmapTexCoords;\n";
const char* gVS_Header_Varyings_HasGradient[6] = {
// Linear
"varying highp vec2 linear;\n", "varying float linear;\n",
// Circular
"varying highp vec2 circular;\n", "varying highp vec2 circular;\n",
// Sweep
"varying highp vec2 sweep;\n", "varying highp vec2 sweep;\n",
};
const char* gVS_Header_Varyings_HasRoundRectClip = "varying mediump vec2 roundRectPos;\n";
const char* gVS_Main = "\nvoid main(void) {\n";
const char* gVS_Main_OutTexCoords = " outTexCoords = texCoords;\n";
const char* gVS_Main_OutColors = " outColors = colors;\n";
const char* gVS_Main_OutTransformedTexCoords =
" outTexCoords = (mainTextureTransform * vec4(texCoords, 0.0, 1.0)).xy;\n";
const char* gVS_Main_OutGradient[6] = {
// Linear
Linear blending, step 1 NOTE: Linear blending is currently disabled in this CL as the feature is still a work in progress Android currently performs all blending (any kind of linear math on colors really) on gamma-encoded colors. Since Android assumes that the default color space is sRGB, all bitmaps and colors are encoded with the sRGB Opto-Electronic Conversion Function (OECF, which can be approximated with a power function). Since the power curve is not linear, our linear math is incorrect. The result is that we generate colors that tend to be too dark; this affects blending but also anti-aliasing, gradients, blurs, etc. The solution is to convert gamma-encoded colors back to linear space before doing any math on them, using the sRGB Electo-Optical Conversion Function (EOCF). This is achieved in different ways in different parts of the pipeline: - Using hardware conversions when sampling from OpenGL textures or writing into OpenGL frame buffers - Using software conversion functions, to translate app-supplied colors to and from sRGB - Using Skia's color spaces Any type of processing on colors must roughly ollow these steps: [sRGB input]->EOCF->[linear data]->[processing]->OECF->[sRGB output] For the sRGB color space, the conversion functions are defined as follows: OECF(linear) := linear <= 0.0031308 ? linear * 12.92 : (pow(linear, 1/2.4) * 1.055) - 0.055 EOCF(srgb) := srgb <= 0.04045 ? srgb / 12.92 : pow((srgb + 0.055) / 1.055, 2.4) The EOCF is simply the reciprocal of the OECF. While it is highly recommended to use the exact sRGB conversion functions everywhere possible, it is sometimes useful or beneficial to rely on approximations: - pow(x,2.2) and pow(x,1/2.2) - x^2 and sqrt(x) The latter is particularly useful in fragment shaders (for instance to apply dithering in sRGB space), especially if the sqrt() can be replaced with an inversesqrt(). Here is a fairly exhaustive list of modifications implemented in this CL: - Set TARGET_ENABLE_LINEAR_BLENDING := false in BoardConfig.mk to disable linear blending. This is only for GLES 2.0 GPUs with no hardware sRGB support. This flag is currently assumed to be false (see note above) - sRGB writes are disabled when entering a functor (WebView). This will need to be fixed at some point - Skia bitmaps are created with the sRGB color space - Bitmaps using a 565 config are expanded to 888 - Linear blending is disabled when entering a functor - External textures are not properly sampled (see below) - Gradients are interpolated in linear space - Texture-based dithering was replaced with analytical dithering - Dithering is done in the quantization color space, which is why we must do EOCF(OECF(color)+dither) - Text is now gamma corrected differently depending on the luminance of the source pixel. The asumption is that a bright pixel will be blended on a dark background and the other way around. The source alpha is gamma corrected to thicken dark on bright and thin bright on dark to match the intended design of fonts. This also matches the behavior of popular design/drawing applications - Removed the asset atlas. It did not contain anything useful and could not be sampled in sRGB without a yet-to-be-defined GL extension - The last column of color matrices is converted to linear space because its value are added to linear colors Missing features: - Resource qualifier? - Regeneration of goldeng images for automated tests - Handle alpha8/grey8 properly - Disable sRGB write for layers with external textures Test: Manual testing while work in progress Bug: 29940137 Change-Id: I6a07b15ab49b554377cd33a36b6d9971a15e9a0b
2016-09-28 17:34:42 -07:00
" linear = vec2((screenSpace * position).x, 0.5);\n",
" linear = (screenSpace * position).x;\n",
// Circular
Linear blending, step 1 NOTE: Linear blending is currently disabled in this CL as the feature is still a work in progress Android currently performs all blending (any kind of linear math on colors really) on gamma-encoded colors. Since Android assumes that the default color space is sRGB, all bitmaps and colors are encoded with the sRGB Opto-Electronic Conversion Function (OECF, which can be approximated with a power function). Since the power curve is not linear, our linear math is incorrect. The result is that we generate colors that tend to be too dark; this affects blending but also anti-aliasing, gradients, blurs, etc. The solution is to convert gamma-encoded colors back to linear space before doing any math on them, using the sRGB Electo-Optical Conversion Function (EOCF). This is achieved in different ways in different parts of the pipeline: - Using hardware conversions when sampling from OpenGL textures or writing into OpenGL frame buffers - Using software conversion functions, to translate app-supplied colors to and from sRGB - Using Skia's color spaces Any type of processing on colors must roughly ollow these steps: [sRGB input]->EOCF->[linear data]->[processing]->OECF->[sRGB output] For the sRGB color space, the conversion functions are defined as follows: OECF(linear) := linear <= 0.0031308 ? linear * 12.92 : (pow(linear, 1/2.4) * 1.055) - 0.055 EOCF(srgb) := srgb <= 0.04045 ? srgb / 12.92 : pow((srgb + 0.055) / 1.055, 2.4) The EOCF is simply the reciprocal of the OECF. While it is highly recommended to use the exact sRGB conversion functions everywhere possible, it is sometimes useful or beneficial to rely on approximations: - pow(x,2.2) and pow(x,1/2.2) - x^2 and sqrt(x) The latter is particularly useful in fragment shaders (for instance to apply dithering in sRGB space), especially if the sqrt() can be replaced with an inversesqrt(). Here is a fairly exhaustive list of modifications implemented in this CL: - Set TARGET_ENABLE_LINEAR_BLENDING := false in BoardConfig.mk to disable linear blending. This is only for GLES 2.0 GPUs with no hardware sRGB support. This flag is currently assumed to be false (see note above) - sRGB writes are disabled when entering a functor (WebView). This will need to be fixed at some point - Skia bitmaps are created with the sRGB color space - Bitmaps using a 565 config are expanded to 888 - Linear blending is disabled when entering a functor - External textures are not properly sampled (see below) - Gradients are interpolated in linear space - Texture-based dithering was replaced with analytical dithering - Dithering is done in the quantization color space, which is why we must do EOCF(OECF(color)+dither) - Text is now gamma corrected differently depending on the luminance of the source pixel. The asumption is that a bright pixel will be blended on a dark background and the other way around. The source alpha is gamma corrected to thicken dark on bright and thin bright on dark to match the intended design of fonts. This also matches the behavior of popular design/drawing applications - Removed the asset atlas. It did not contain anything useful and could not be sampled in sRGB without a yet-to-be-defined GL extension - The last column of color matrices is converted to linear space because its value are added to linear colors Missing features: - Resource qualifier? - Regeneration of goldeng images for automated tests - Handle alpha8/grey8 properly - Disable sRGB write for layers with external textures Test: Manual testing while work in progress Bug: 29940137 Change-Id: I6a07b15ab49b554377cd33a36b6d9971a15e9a0b
2016-09-28 17:34:42 -07:00
" circular = (screenSpace * position).xy;\n",
" circular = (screenSpace * position).xy;\n",
// Sweep
" sweep = (screenSpace * position).xy;\n", " sweep = (screenSpace * position).xy;\n"};
const char* gVS_Main_OutBitmapTexCoords =
" outBitmapTexCoords = (textureTransform * position).xy * textureDimension;\n";
const char* gVS_Main_Position =
" vec4 transformedPosition = projection * transform * position;\n"
" gl_Position = transformedPosition;\n";
const char* gVS_Main_VertexAlpha = " alpha = vtxAlpha;\n";
const char* gVS_Main_HasRoundRectClip =
" roundRectPos = ((roundRectInvTransform * transformedPosition).xy / roundRectRadius) - "
"roundRectInnerRectLTWH.xy;\n";
const char* gVS_Footer = "}\n\n";
///////////////////////////////////////////////////////////////////////////////
// Fragment shaders snippets
///////////////////////////////////////////////////////////////////////////////
const char* gFS_Header_Start = "#version 100\n";
const char* gFS_Header_Extension_FramebufferFetch =
"#extension GL_NV_shader_framebuffer_fetch : enable\n\n";
const char* gFS_Header_Extension_ExternalTexture =
"#extension GL_OES_EGL_image_external : require\n\n";
const char* gFS_Header = "precision mediump float;\n\n";
const char* gFS_Uniforms_Color = "uniform vec4 color;\n";
const char* gFS_Uniforms_TextureSampler = "uniform sampler2D baseSampler;\n";
const char* gFS_Uniforms_ExternalTextureSampler = "uniform samplerExternalOES baseSampler;\n";
const char* gFS_Uniforms_GradientSampler[2] = {
Linear blending, step 1 NOTE: Linear blending is currently disabled in this CL as the feature is still a work in progress Android currently performs all blending (any kind of linear math on colors really) on gamma-encoded colors. Since Android assumes that the default color space is sRGB, all bitmaps and colors are encoded with the sRGB Opto-Electronic Conversion Function (OECF, which can be approximated with a power function). Since the power curve is not linear, our linear math is incorrect. The result is that we generate colors that tend to be too dark; this affects blending but also anti-aliasing, gradients, blurs, etc. The solution is to convert gamma-encoded colors back to linear space before doing any math on them, using the sRGB Electo-Optical Conversion Function (EOCF). This is achieved in different ways in different parts of the pipeline: - Using hardware conversions when sampling from OpenGL textures or writing into OpenGL frame buffers - Using software conversion functions, to translate app-supplied colors to and from sRGB - Using Skia's color spaces Any type of processing on colors must roughly ollow these steps: [sRGB input]->EOCF->[linear data]->[processing]->OECF->[sRGB output] For the sRGB color space, the conversion functions are defined as follows: OECF(linear) := linear <= 0.0031308 ? linear * 12.92 : (pow(linear, 1/2.4) * 1.055) - 0.055 EOCF(srgb) := srgb <= 0.04045 ? srgb / 12.92 : pow((srgb + 0.055) / 1.055, 2.4) The EOCF is simply the reciprocal of the OECF. While it is highly recommended to use the exact sRGB conversion functions everywhere possible, it is sometimes useful or beneficial to rely on approximations: - pow(x,2.2) and pow(x,1/2.2) - x^2 and sqrt(x) The latter is particularly useful in fragment shaders (for instance to apply dithering in sRGB space), especially if the sqrt() can be replaced with an inversesqrt(). Here is a fairly exhaustive list of modifications implemented in this CL: - Set TARGET_ENABLE_LINEAR_BLENDING := false in BoardConfig.mk to disable linear blending. This is only for GLES 2.0 GPUs with no hardware sRGB support. This flag is currently assumed to be false (see note above) - sRGB writes are disabled when entering a functor (WebView). This will need to be fixed at some point - Skia bitmaps are created with the sRGB color space - Bitmaps using a 565 config are expanded to 888 - Linear blending is disabled when entering a functor - External textures are not properly sampled (see below) - Gradients are interpolated in linear space - Texture-based dithering was replaced with analytical dithering - Dithering is done in the quantization color space, which is why we must do EOCF(OECF(color)+dither) - Text is now gamma corrected differently depending on the luminance of the source pixel. The asumption is that a bright pixel will be blended on a dark background and the other way around. The source alpha is gamma corrected to thicken dark on bright and thin bright on dark to match the intended design of fonts. This also matches the behavior of popular design/drawing applications - Removed the asset atlas. It did not contain anything useful and could not be sampled in sRGB without a yet-to-be-defined GL extension - The last column of color matrices is converted to linear space because its value are added to linear colors Missing features: - Resource qualifier? - Regeneration of goldeng images for automated tests - Handle alpha8/grey8 properly - Disable sRGB write for layers with external textures Test: Manual testing while work in progress Bug: 29940137 Change-Id: I6a07b15ab49b554377cd33a36b6d9971a15e9a0b
2016-09-28 17:34:42 -07:00
"uniform vec2 screenSize;\n"
"uniform sampler2D gradientSampler;\n",
Linear blending, step 1 NOTE: Linear blending is currently disabled in this CL as the feature is still a work in progress Android currently performs all blending (any kind of linear math on colors really) on gamma-encoded colors. Since Android assumes that the default color space is sRGB, all bitmaps and colors are encoded with the sRGB Opto-Electronic Conversion Function (OECF, which can be approximated with a power function). Since the power curve is not linear, our linear math is incorrect. The result is that we generate colors that tend to be too dark; this affects blending but also anti-aliasing, gradients, blurs, etc. The solution is to convert gamma-encoded colors back to linear space before doing any math on them, using the sRGB Electo-Optical Conversion Function (EOCF). This is achieved in different ways in different parts of the pipeline: - Using hardware conversions when sampling from OpenGL textures or writing into OpenGL frame buffers - Using software conversion functions, to translate app-supplied colors to and from sRGB - Using Skia's color spaces Any type of processing on colors must roughly ollow these steps: [sRGB input]->EOCF->[linear data]->[processing]->OECF->[sRGB output] For the sRGB color space, the conversion functions are defined as follows: OECF(linear) := linear <= 0.0031308 ? linear * 12.92 : (pow(linear, 1/2.4) * 1.055) - 0.055 EOCF(srgb) := srgb <= 0.04045 ? srgb / 12.92 : pow((srgb + 0.055) / 1.055, 2.4) The EOCF is simply the reciprocal of the OECF. While it is highly recommended to use the exact sRGB conversion functions everywhere possible, it is sometimes useful or beneficial to rely on approximations: - pow(x,2.2) and pow(x,1/2.2) - x^2 and sqrt(x) The latter is particularly useful in fragment shaders (for instance to apply dithering in sRGB space), especially if the sqrt() can be replaced with an inversesqrt(). Here is a fairly exhaustive list of modifications implemented in this CL: - Set TARGET_ENABLE_LINEAR_BLENDING := false in BoardConfig.mk to disable linear blending. This is only for GLES 2.0 GPUs with no hardware sRGB support. This flag is currently assumed to be false (see note above) - sRGB writes are disabled when entering a functor (WebView). This will need to be fixed at some point - Skia bitmaps are created with the sRGB color space - Bitmaps using a 565 config are expanded to 888 - Linear blending is disabled when entering a functor - External textures are not properly sampled (see below) - Gradients are interpolated in linear space - Texture-based dithering was replaced with analytical dithering - Dithering is done in the quantization color space, which is why we must do EOCF(OECF(color)+dither) - Text is now gamma corrected differently depending on the luminance of the source pixel. The asumption is that a bright pixel will be blended on a dark background and the other way around. The source alpha is gamma corrected to thicken dark on bright and thin bright on dark to match the intended design of fonts. This also matches the behavior of popular design/drawing applications - Removed the asset atlas. It did not contain anything useful and could not be sampled in sRGB without a yet-to-be-defined GL extension - The last column of color matrices is converted to linear space because its value are added to linear colors Missing features: - Resource qualifier? - Regeneration of goldeng images for automated tests - Handle alpha8/grey8 properly - Disable sRGB write for layers with external textures Test: Manual testing while work in progress Bug: 29940137 Change-Id: I6a07b15ab49b554377cd33a36b6d9971a15e9a0b
2016-09-28 17:34:42 -07:00
"uniform vec2 screenSize;\n"
"uniform vec4 startColor;\n"
"uniform vec4 endColor;\n"};
const char* gFS_Uniforms_BitmapSampler = "uniform sampler2D bitmapSampler;\n";
const char* gFS_Uniforms_BitmapExternalSampler = "uniform samplerExternalOES bitmapSampler;\n";
const char* gFS_Uniforms_ColorOp[3] = {
// None
"",
// Matrix
"uniform mat4 colorMatrix;\n"
"uniform vec4 colorMatrixVector;\n",
// PorterDuff
"uniform vec4 colorBlend;\n"};
const char* gFS_Uniforms_HasRoundRectClip =
"uniform mediump vec4 roundRectInnerRectLTWH;\n"
"uniform mediump float roundRectRadius;\n";
const char* gFS_Uniforms_ColorSpaceConversion =
// TODO: Should we use a 3D LUT to combine the matrix and transfer functions?
// 32x32x32 fp16 LUTs (for scRGB output) are large and heavy to generate...
"uniform mat3 colorSpaceMatrix;\n";
const char* gFS_Uniforms_TransferFunction[4] = {
// In this order: g, a, b, c, d, e, f
// See ColorSpace::TransferParameters
// We'll use hardware sRGB conversion as much as possible
"", "uniform float transferFunction[7];\n", "uniform float transferFunction[5];\n",
"uniform float transferFunctionGamma;\n"};
const char* gFS_OETF[2] = {
R"__SHADER__(
vec4 OETF(const vec4 linear) {
return linear;
}
)__SHADER__",
// We expect linear data to be scRGB so we mirror the gamma function
R"__SHADER__(
vec4 OETF(const vec4 linear) {
return vec4(sign(linear.rgb) * OETF_sRGB(abs(linear.rgb)), linear.a);
}
)__SHADER__"};
const char* gFS_ColorConvert[3] = {
// Just OETF
R"__SHADER__(
vec4 colorConvert(const vec4 color) {
return OETF(color);
}
)__SHADER__",
// Full color conversion for opaque bitmaps
R"__SHADER__(
vec4 colorConvert(const vec4 color) {
return OETF(vec4(colorSpaceMatrix * EOTF_Parametric(color.rgb), color.a));
}
)__SHADER__",
// Full color conversion for translucent bitmaps
// Note: 0.5/256=0.0019
R"__SHADER__(
vec4 colorConvert(in vec4 color) {
color.rgb /= color.a + 0.0019;
color = OETF(vec4(colorSpaceMatrix * EOTF_Parametric(color.rgb), color.a));
color.rgb *= color.a + 0.0019;
return color;
}
)__SHADER__",
};
const char* gFS_sRGB_TransferFunctions = R"__SHADER__(
float OETF_sRGB(const float linear) {
// IEC 61966-2-1:1999
return linear <= 0.0031308 ? linear * 12.92 : (pow(linear, 1.0 / 2.4) * 1.055) - 0.055;
}
vec3 OETF_sRGB(const vec3 linear) {
return vec3(OETF_sRGB(linear.r), OETF_sRGB(linear.g), OETF_sRGB(linear.b));
}
float EOTF_sRGB(float srgb) {
// IEC 61966-2-1:1999
return srgb <= 0.04045 ? srgb / 12.92 : pow((srgb + 0.055) / 1.055, 2.4);
}
)__SHADER__";
const char* gFS_TransferFunction[4] = {
// Conversion done by the texture unit (sRGB)
R"__SHADER__(
vec3 EOTF_Parametric(const vec3 x) {
return x;
}
)__SHADER__",
// Full transfer function
// TODO: We should probably use a 1D LUT (256x1 with texelFetch() since input is 8 bit)
// TODO: That would cause 3 dependent texture fetches. Is it worth it?
R"__SHADER__(
float EOTF_Parametric(float x) {
return x <= transferFunction[4]
? transferFunction[3] * x + transferFunction[6]
: pow(transferFunction[1] * x + transferFunction[2], transferFunction[0])
+ transferFunction[5];
}
vec3 EOTF_Parametric(const vec3 x) {
return vec3(EOTF_Parametric(x.r), EOTF_Parametric(x.g), EOTF_Parametric(x.b));
}
)__SHADER__",
// Limited transfer function, e = f = 0.0
R"__SHADER__(
float EOTF_Parametric(float x) {
return x <= transferFunction[4]
? transferFunction[3] * x
: pow(transferFunction[1] * x + transferFunction[2], transferFunction[0]);
}
vec3 EOTF_Parametric(const vec3 x) {
return vec3(EOTF_Parametric(x.r), EOTF_Parametric(x.g), EOTF_Parametric(x.b));
}
)__SHADER__",
// Gamma transfer function, e = f = 0.0
R"__SHADER__(
vec3 EOTF_Parametric(const vec3 x) {
return vec3(pow(x.r, transferFunctionGamma),
pow(x.g, transferFunctionGamma),
pow(x.b, transferFunctionGamma));
}
)__SHADER__"};
Linear blending, step 1 NOTE: Linear blending is currently disabled in this CL as the feature is still a work in progress Android currently performs all blending (any kind of linear math on colors really) on gamma-encoded colors. Since Android assumes that the default color space is sRGB, all bitmaps and colors are encoded with the sRGB Opto-Electronic Conversion Function (OECF, which can be approximated with a power function). Since the power curve is not linear, our linear math is incorrect. The result is that we generate colors that tend to be too dark; this affects blending but also anti-aliasing, gradients, blurs, etc. The solution is to convert gamma-encoded colors back to linear space before doing any math on them, using the sRGB Electo-Optical Conversion Function (EOCF). This is achieved in different ways in different parts of the pipeline: - Using hardware conversions when sampling from OpenGL textures or writing into OpenGL frame buffers - Using software conversion functions, to translate app-supplied colors to and from sRGB - Using Skia's color spaces Any type of processing on colors must roughly ollow these steps: [sRGB input]->EOCF->[linear data]->[processing]->OECF->[sRGB output] For the sRGB color space, the conversion functions are defined as follows: OECF(linear) := linear <= 0.0031308 ? linear * 12.92 : (pow(linear, 1/2.4) * 1.055) - 0.055 EOCF(srgb) := srgb <= 0.04045 ? srgb / 12.92 : pow((srgb + 0.055) / 1.055, 2.4) The EOCF is simply the reciprocal of the OECF. While it is highly recommended to use the exact sRGB conversion functions everywhere possible, it is sometimes useful or beneficial to rely on approximations: - pow(x,2.2) and pow(x,1/2.2) - x^2 and sqrt(x) The latter is particularly useful in fragment shaders (for instance to apply dithering in sRGB space), especially if the sqrt() can be replaced with an inversesqrt(). Here is a fairly exhaustive list of modifications implemented in this CL: - Set TARGET_ENABLE_LINEAR_BLENDING := false in BoardConfig.mk to disable linear blending. This is only for GLES 2.0 GPUs with no hardware sRGB support. This flag is currently assumed to be false (see note above) - sRGB writes are disabled when entering a functor (WebView). This will need to be fixed at some point - Skia bitmaps are created with the sRGB color space - Bitmaps using a 565 config are expanded to 888 - Linear blending is disabled when entering a functor - External textures are not properly sampled (see below) - Gradients are interpolated in linear space - Texture-based dithering was replaced with analytical dithering - Dithering is done in the quantization color space, which is why we must do EOCF(OECF(color)+dither) - Text is now gamma corrected differently depending on the luminance of the source pixel. The asumption is that a bright pixel will be blended on a dark background and the other way around. The source alpha is gamma corrected to thicken dark on bright and thin bright on dark to match the intended design of fonts. This also matches the behavior of popular design/drawing applications - Removed the asset atlas. It did not contain anything useful and could not be sampled in sRGB without a yet-to-be-defined GL extension - The last column of color matrices is converted to linear space because its value are added to linear colors Missing features: - Resource qualifier? - Regeneration of goldeng images for automated tests - Handle alpha8/grey8 properly - Disable sRGB write for layers with external textures Test: Manual testing while work in progress Bug: 29940137 Change-Id: I6a07b15ab49b554377cd33a36b6d9971a15e9a0b
2016-09-28 17:34:42 -07:00
// Dithering must be done in the quantization space
// When we are writing to an sRGB framebuffer, we must do the following:
// EOTF(OETF(color) + dither)
// The dithering pattern is generated with a triangle noise generator in the range [-1.0,1.0]
Linear blending, step 1 NOTE: Linear blending is currently disabled in this CL as the feature is still a work in progress Android currently performs all blending (any kind of linear math on colors really) on gamma-encoded colors. Since Android assumes that the default color space is sRGB, all bitmaps and colors are encoded with the sRGB Opto-Electronic Conversion Function (OECF, which can be approximated with a power function). Since the power curve is not linear, our linear math is incorrect. The result is that we generate colors that tend to be too dark; this affects blending but also anti-aliasing, gradients, blurs, etc. The solution is to convert gamma-encoded colors back to linear space before doing any math on them, using the sRGB Electo-Optical Conversion Function (EOCF). This is achieved in different ways in different parts of the pipeline: - Using hardware conversions when sampling from OpenGL textures or writing into OpenGL frame buffers - Using software conversion functions, to translate app-supplied colors to and from sRGB - Using Skia's color spaces Any type of processing on colors must roughly ollow these steps: [sRGB input]->EOCF->[linear data]->[processing]->OECF->[sRGB output] For the sRGB color space, the conversion functions are defined as follows: OECF(linear) := linear <= 0.0031308 ? linear * 12.92 : (pow(linear, 1/2.4) * 1.055) - 0.055 EOCF(srgb) := srgb <= 0.04045 ? srgb / 12.92 : pow((srgb + 0.055) / 1.055, 2.4) The EOCF is simply the reciprocal of the OECF. While it is highly recommended to use the exact sRGB conversion functions everywhere possible, it is sometimes useful or beneficial to rely on approximations: - pow(x,2.2) and pow(x,1/2.2) - x^2 and sqrt(x) The latter is particularly useful in fragment shaders (for instance to apply dithering in sRGB space), especially if the sqrt() can be replaced with an inversesqrt(). Here is a fairly exhaustive list of modifications implemented in this CL: - Set TARGET_ENABLE_LINEAR_BLENDING := false in BoardConfig.mk to disable linear blending. This is only for GLES 2.0 GPUs with no hardware sRGB support. This flag is currently assumed to be false (see note above) - sRGB writes are disabled when entering a functor (WebView). This will need to be fixed at some point - Skia bitmaps are created with the sRGB color space - Bitmaps using a 565 config are expanded to 888 - Linear blending is disabled when entering a functor - External textures are not properly sampled (see below) - Gradients are interpolated in linear space - Texture-based dithering was replaced with analytical dithering - Dithering is done in the quantization color space, which is why we must do EOCF(OECF(color)+dither) - Text is now gamma corrected differently depending on the luminance of the source pixel. The asumption is that a bright pixel will be blended on a dark background and the other way around. The source alpha is gamma corrected to thicken dark on bright and thin bright on dark to match the intended design of fonts. This also matches the behavior of popular design/drawing applications - Removed the asset atlas. It did not contain anything useful and could not be sampled in sRGB without a yet-to-be-defined GL extension - The last column of color matrices is converted to linear space because its value are added to linear colors Missing features: - Resource qualifier? - Regeneration of goldeng images for automated tests - Handle alpha8/grey8 properly - Disable sRGB write for layers with external textures Test: Manual testing while work in progress Bug: 29940137 Change-Id: I6a07b15ab49b554377cd33a36b6d9971a15e9a0b
2016-09-28 17:34:42 -07:00
// TODO: Handle linear fp16 render targets
const char* gFS_GradientFunctions = R"__SHADER__(
float triangleNoise(const highp vec2 n) {
highp vec2 p = fract(n * vec2(5.3987, 5.4421));
p += dot(p.yx, p.xy + vec2(21.5351, 14.3137));
highp float xy = p.x * p.y;
return fract(xy * 95.4307) + fract(xy * 75.04961) - 1.0;
}
)__SHADER__";
const char* gFS_GradientPreamble[2] = {
Linear blending, step 1 NOTE: Linear blending is currently disabled in this CL as the feature is still a work in progress Android currently performs all blending (any kind of linear math on colors really) on gamma-encoded colors. Since Android assumes that the default color space is sRGB, all bitmaps and colors are encoded with the sRGB Opto-Electronic Conversion Function (OECF, which can be approximated with a power function). Since the power curve is not linear, our linear math is incorrect. The result is that we generate colors that tend to be too dark; this affects blending but also anti-aliasing, gradients, blurs, etc. The solution is to convert gamma-encoded colors back to linear space before doing any math on them, using the sRGB Electo-Optical Conversion Function (EOCF). This is achieved in different ways in different parts of the pipeline: - Using hardware conversions when sampling from OpenGL textures or writing into OpenGL frame buffers - Using software conversion functions, to translate app-supplied colors to and from sRGB - Using Skia's color spaces Any type of processing on colors must roughly ollow these steps: [sRGB input]->EOCF->[linear data]->[processing]->OECF->[sRGB output] For the sRGB color space, the conversion functions are defined as follows: OECF(linear) := linear <= 0.0031308 ? linear * 12.92 : (pow(linear, 1/2.4) * 1.055) - 0.055 EOCF(srgb) := srgb <= 0.04045 ? srgb / 12.92 : pow((srgb + 0.055) / 1.055, 2.4) The EOCF is simply the reciprocal of the OECF. While it is highly recommended to use the exact sRGB conversion functions everywhere possible, it is sometimes useful or beneficial to rely on approximations: - pow(x,2.2) and pow(x,1/2.2) - x^2 and sqrt(x) The latter is particularly useful in fragment shaders (for instance to apply dithering in sRGB space), especially if the sqrt() can be replaced with an inversesqrt(). Here is a fairly exhaustive list of modifications implemented in this CL: - Set TARGET_ENABLE_LINEAR_BLENDING := false in BoardConfig.mk to disable linear blending. This is only for GLES 2.0 GPUs with no hardware sRGB support. This flag is currently assumed to be false (see note above) - sRGB writes are disabled when entering a functor (WebView). This will need to be fixed at some point - Skia bitmaps are created with the sRGB color space - Bitmaps using a 565 config are expanded to 888 - Linear blending is disabled when entering a functor - External textures are not properly sampled (see below) - Gradients are interpolated in linear space - Texture-based dithering was replaced with analytical dithering - Dithering is done in the quantization color space, which is why we must do EOCF(OECF(color)+dither) - Text is now gamma corrected differently depending on the luminance of the source pixel. The asumption is that a bright pixel will be blended on a dark background and the other way around. The source alpha is gamma corrected to thicken dark on bright and thin bright on dark to match the intended design of fonts. This also matches the behavior of popular design/drawing applications - Removed the asset atlas. It did not contain anything useful and could not be sampled in sRGB without a yet-to-be-defined GL extension - The last column of color matrices is converted to linear space because its value are added to linear colors Missing features: - Resource qualifier? - Regeneration of goldeng images for automated tests - Handle alpha8/grey8 properly - Disable sRGB write for layers with external textures Test: Manual testing while work in progress Bug: 29940137 Change-Id: I6a07b15ab49b554377cd33a36b6d9971a15e9a0b
2016-09-28 17:34:42 -07:00
// Linear framebuffer
R"__SHADER__(
vec4 dither(const vec4 color) {
return color + (triangleNoise(gl_FragCoord.xy * screenSize.xy) / 255.0);
}
)__SHADER__",
Linear blending, step 1 NOTE: Linear blending is currently disabled in this CL as the feature is still a work in progress Android currently performs all blending (any kind of linear math on colors really) on gamma-encoded colors. Since Android assumes that the default color space is sRGB, all bitmaps and colors are encoded with the sRGB Opto-Electronic Conversion Function (OECF, which can be approximated with a power function). Since the power curve is not linear, our linear math is incorrect. The result is that we generate colors that tend to be too dark; this affects blending but also anti-aliasing, gradients, blurs, etc. The solution is to convert gamma-encoded colors back to linear space before doing any math on them, using the sRGB Electo-Optical Conversion Function (EOCF). This is achieved in different ways in different parts of the pipeline: - Using hardware conversions when sampling from OpenGL textures or writing into OpenGL frame buffers - Using software conversion functions, to translate app-supplied colors to and from sRGB - Using Skia's color spaces Any type of processing on colors must roughly ollow these steps: [sRGB input]->EOCF->[linear data]->[processing]->OECF->[sRGB output] For the sRGB color space, the conversion functions are defined as follows: OECF(linear) := linear <= 0.0031308 ? linear * 12.92 : (pow(linear, 1/2.4) * 1.055) - 0.055 EOCF(srgb) := srgb <= 0.04045 ? srgb / 12.92 : pow((srgb + 0.055) / 1.055, 2.4) The EOCF is simply the reciprocal of the OECF. While it is highly recommended to use the exact sRGB conversion functions everywhere possible, it is sometimes useful or beneficial to rely on approximations: - pow(x,2.2) and pow(x,1/2.2) - x^2 and sqrt(x) The latter is particularly useful in fragment shaders (for instance to apply dithering in sRGB space), especially if the sqrt() can be replaced with an inversesqrt(). Here is a fairly exhaustive list of modifications implemented in this CL: - Set TARGET_ENABLE_LINEAR_BLENDING := false in BoardConfig.mk to disable linear blending. This is only for GLES 2.0 GPUs with no hardware sRGB support. This flag is currently assumed to be false (see note above) - sRGB writes are disabled when entering a functor (WebView). This will need to be fixed at some point - Skia bitmaps are created with the sRGB color space - Bitmaps using a 565 config are expanded to 888 - Linear blending is disabled when entering a functor - External textures are not properly sampled (see below) - Gradients are interpolated in linear space - Texture-based dithering was replaced with analytical dithering - Dithering is done in the quantization color space, which is why we must do EOCF(OECF(color)+dither) - Text is now gamma corrected differently depending on the luminance of the source pixel. The asumption is that a bright pixel will be blended on a dark background and the other way around. The source alpha is gamma corrected to thicken dark on bright and thin bright on dark to match the intended design of fonts. This also matches the behavior of popular design/drawing applications - Removed the asset atlas. It did not contain anything useful and could not be sampled in sRGB without a yet-to-be-defined GL extension - The last column of color matrices is converted to linear space because its value are added to linear colors Missing features: - Resource qualifier? - Regeneration of goldeng images for automated tests - Handle alpha8/grey8 properly - Disable sRGB write for layers with external textures Test: Manual testing while work in progress Bug: 29940137 Change-Id: I6a07b15ab49b554377cd33a36b6d9971a15e9a0b
2016-09-28 17:34:42 -07:00
// sRGB framebuffer
R"__SHADER__(
vec4 dither(const vec4 color) {
vec3 dithered = sqrt(color.rgb) + (triangleNoise(gl_FragCoord.xy * screenSize.xy) / 255.0);
return vec4(dithered * dithered, color.a);
}
)__SHADER__",
Linear blending, step 1 NOTE: Linear blending is currently disabled in this CL as the feature is still a work in progress Android currently performs all blending (any kind of linear math on colors really) on gamma-encoded colors. Since Android assumes that the default color space is sRGB, all bitmaps and colors are encoded with the sRGB Opto-Electronic Conversion Function (OECF, which can be approximated with a power function). Since the power curve is not linear, our linear math is incorrect. The result is that we generate colors that tend to be too dark; this affects blending but also anti-aliasing, gradients, blurs, etc. The solution is to convert gamma-encoded colors back to linear space before doing any math on them, using the sRGB Electo-Optical Conversion Function (EOCF). This is achieved in different ways in different parts of the pipeline: - Using hardware conversions when sampling from OpenGL textures or writing into OpenGL frame buffers - Using software conversion functions, to translate app-supplied colors to and from sRGB - Using Skia's color spaces Any type of processing on colors must roughly ollow these steps: [sRGB input]->EOCF->[linear data]->[processing]->OECF->[sRGB output] For the sRGB color space, the conversion functions are defined as follows: OECF(linear) := linear <= 0.0031308 ? linear * 12.92 : (pow(linear, 1/2.4) * 1.055) - 0.055 EOCF(srgb) := srgb <= 0.04045 ? srgb / 12.92 : pow((srgb + 0.055) / 1.055, 2.4) The EOCF is simply the reciprocal of the OECF. While it is highly recommended to use the exact sRGB conversion functions everywhere possible, it is sometimes useful or beneficial to rely on approximations: - pow(x,2.2) and pow(x,1/2.2) - x^2 and sqrt(x) The latter is particularly useful in fragment shaders (for instance to apply dithering in sRGB space), especially if the sqrt() can be replaced with an inversesqrt(). Here is a fairly exhaustive list of modifications implemented in this CL: - Set TARGET_ENABLE_LINEAR_BLENDING := false in BoardConfig.mk to disable linear blending. This is only for GLES 2.0 GPUs with no hardware sRGB support. This flag is currently assumed to be false (see note above) - sRGB writes are disabled when entering a functor (WebView). This will need to be fixed at some point - Skia bitmaps are created with the sRGB color space - Bitmaps using a 565 config are expanded to 888 - Linear blending is disabled when entering a functor - External textures are not properly sampled (see below) - Gradients are interpolated in linear space - Texture-based dithering was replaced with analytical dithering - Dithering is done in the quantization color space, which is why we must do EOCF(OECF(color)+dither) - Text is now gamma corrected differently depending on the luminance of the source pixel. The asumption is that a bright pixel will be blended on a dark background and the other way around. The source alpha is gamma corrected to thicken dark on bright and thin bright on dark to match the intended design of fonts. This also matches the behavior of popular design/drawing applications - Removed the asset atlas. It did not contain anything useful and could not be sampled in sRGB without a yet-to-be-defined GL extension - The last column of color matrices is converted to linear space because its value are added to linear colors Missing features: - Resource qualifier? - Regeneration of goldeng images for automated tests - Handle alpha8/grey8 properly - Disable sRGB write for layers with external textures Test: Manual testing while work in progress Bug: 29940137 Change-Id: I6a07b15ab49b554377cd33a36b6d9971a15e9a0b
2016-09-28 17:34:42 -07:00
};
// Uses luminance coefficients from Rec.709 to choose the appropriate gamma
// The gamma() function assumes that bright text will be displayed on a dark
// background and that dark text will be displayed on bright background
// The gamma coefficient is chosen to thicken or thin the text accordingly
// The dot product used to compute the luminance could be approximated with
// a simple max(color.r, color.g, color.b)
const char* gFS_Gamma_Preamble = R"__SHADER__(
#define GAMMA (%.2f)
#define GAMMA_INV (%.2f)
float gamma(float a, const vec3 color) {
float luminance = dot(color, vec3(0.2126, 0.7152, 0.0722));
return pow(a, luminance < 0.5 ? GAMMA_INV : GAMMA);
}
)__SHADER__";
Linear blending, step 1 NOTE: Linear blending is currently disabled in this CL as the feature is still a work in progress Android currently performs all blending (any kind of linear math on colors really) on gamma-encoded colors. Since Android assumes that the default color space is sRGB, all bitmaps and colors are encoded with the sRGB Opto-Electronic Conversion Function (OECF, which can be approximated with a power function). Since the power curve is not linear, our linear math is incorrect. The result is that we generate colors that tend to be too dark; this affects blending but also anti-aliasing, gradients, blurs, etc. The solution is to convert gamma-encoded colors back to linear space before doing any math on them, using the sRGB Electo-Optical Conversion Function (EOCF). This is achieved in different ways in different parts of the pipeline: - Using hardware conversions when sampling from OpenGL textures or writing into OpenGL frame buffers - Using software conversion functions, to translate app-supplied colors to and from sRGB - Using Skia's color spaces Any type of processing on colors must roughly ollow these steps: [sRGB input]->EOCF->[linear data]->[processing]->OECF->[sRGB output] For the sRGB color space, the conversion functions are defined as follows: OECF(linear) := linear <= 0.0031308 ? linear * 12.92 : (pow(linear, 1/2.4) * 1.055) - 0.055 EOCF(srgb) := srgb <= 0.04045 ? srgb / 12.92 : pow((srgb + 0.055) / 1.055, 2.4) The EOCF is simply the reciprocal of the OECF. While it is highly recommended to use the exact sRGB conversion functions everywhere possible, it is sometimes useful or beneficial to rely on approximations: - pow(x,2.2) and pow(x,1/2.2) - x^2 and sqrt(x) The latter is particularly useful in fragment shaders (for instance to apply dithering in sRGB space), especially if the sqrt() can be replaced with an inversesqrt(). Here is a fairly exhaustive list of modifications implemented in this CL: - Set TARGET_ENABLE_LINEAR_BLENDING := false in BoardConfig.mk to disable linear blending. This is only for GLES 2.0 GPUs with no hardware sRGB support. This flag is currently assumed to be false (see note above) - sRGB writes are disabled when entering a functor (WebView). This will need to be fixed at some point - Skia bitmaps are created with the sRGB color space - Bitmaps using a 565 config are expanded to 888 - Linear blending is disabled when entering a functor - External textures are not properly sampled (see below) - Gradients are interpolated in linear space - Texture-based dithering was replaced with analytical dithering - Dithering is done in the quantization color space, which is why we must do EOCF(OECF(color)+dither) - Text is now gamma corrected differently depending on the luminance of the source pixel. The asumption is that a bright pixel will be blended on a dark background and the other way around. The source alpha is gamma corrected to thicken dark on bright and thin bright on dark to match the intended design of fonts. This also matches the behavior of popular design/drawing applications - Removed the asset atlas. It did not contain anything useful and could not be sampled in sRGB without a yet-to-be-defined GL extension - The last column of color matrices is converted to linear space because its value are added to linear colors Missing features: - Resource qualifier? - Regeneration of goldeng images for automated tests - Handle alpha8/grey8 properly - Disable sRGB write for layers with external textures Test: Manual testing while work in progress Bug: 29940137 Change-Id: I6a07b15ab49b554377cd33a36b6d9971a15e9a0b
2016-09-28 17:34:42 -07:00
const char* gFS_Main =
"\nvoid main(void) {\n"
Linear blending, step 1 NOTE: Linear blending is currently disabled in this CL as the feature is still a work in progress Android currently performs all blending (any kind of linear math on colors really) on gamma-encoded colors. Since Android assumes that the default color space is sRGB, all bitmaps and colors are encoded with the sRGB Opto-Electronic Conversion Function (OECF, which can be approximated with a power function). Since the power curve is not linear, our linear math is incorrect. The result is that we generate colors that tend to be too dark; this affects blending but also anti-aliasing, gradients, blurs, etc. The solution is to convert gamma-encoded colors back to linear space before doing any math on them, using the sRGB Electo-Optical Conversion Function (EOCF). This is achieved in different ways in different parts of the pipeline: - Using hardware conversions when sampling from OpenGL textures or writing into OpenGL frame buffers - Using software conversion functions, to translate app-supplied colors to and from sRGB - Using Skia's color spaces Any type of processing on colors must roughly ollow these steps: [sRGB input]->EOCF->[linear data]->[processing]->OECF->[sRGB output] For the sRGB color space, the conversion functions are defined as follows: OECF(linear) := linear <= 0.0031308 ? linear * 12.92 : (pow(linear, 1/2.4) * 1.055) - 0.055 EOCF(srgb) := srgb <= 0.04045 ? srgb / 12.92 : pow((srgb + 0.055) / 1.055, 2.4) The EOCF is simply the reciprocal of the OECF. While it is highly recommended to use the exact sRGB conversion functions everywhere possible, it is sometimes useful or beneficial to rely on approximations: - pow(x,2.2) and pow(x,1/2.2) - x^2 and sqrt(x) The latter is particularly useful in fragment shaders (for instance to apply dithering in sRGB space), especially if the sqrt() can be replaced with an inversesqrt(). Here is a fairly exhaustive list of modifications implemented in this CL: - Set TARGET_ENABLE_LINEAR_BLENDING := false in BoardConfig.mk to disable linear blending. This is only for GLES 2.0 GPUs with no hardware sRGB support. This flag is currently assumed to be false (see note above) - sRGB writes are disabled when entering a functor (WebView). This will need to be fixed at some point - Skia bitmaps are created with the sRGB color space - Bitmaps using a 565 config are expanded to 888 - Linear blending is disabled when entering a functor - External textures are not properly sampled (see below) - Gradients are interpolated in linear space - Texture-based dithering was replaced with analytical dithering - Dithering is done in the quantization color space, which is why we must do EOCF(OECF(color)+dither) - Text is now gamma corrected differently depending on the luminance of the source pixel. The asumption is that a bright pixel will be blended on a dark background and the other way around. The source alpha is gamma corrected to thicken dark on bright and thin bright on dark to match the intended design of fonts. This also matches the behavior of popular design/drawing applications - Removed the asset atlas. It did not contain anything useful and could not be sampled in sRGB without a yet-to-be-defined GL extension - The last column of color matrices is converted to linear space because its value are added to linear colors Missing features: - Resource qualifier? - Regeneration of goldeng images for automated tests - Handle alpha8/grey8 properly - Disable sRGB write for layers with external textures Test: Manual testing while work in progress Bug: 29940137 Change-Id: I6a07b15ab49b554377cd33a36b6d9971a15e9a0b
2016-09-28 17:34:42 -07:00
" vec4 fragColor;\n";
const char* gFS_Main_AddDither = " fragColor = dither(fragColor);\n";
// General case
const char* gFS_Main_FetchColor = " fragColor = color;\n";
const char* gFS_Main_ModulateColor = " fragColor *= color.a;\n";
const char* gFS_Main_ApplyVertexAlphaLinearInterp = " fragColor *= alpha;\n";
const char* gFS_Main_ApplyVertexAlphaShadowInterp =
// map alpha through shadow alpha sampler
" fragColor *= texture2D(baseSampler, vec2(alpha, 0.5)).a;\n";
const char* gFS_Main_FetchTexture[2] = {
// Don't modulate
" fragColor = colorConvert(texture2D(baseSampler, outTexCoords));\n",
// Modulate
" fragColor = color * colorConvert(texture2D(baseSampler, outTexCoords));\n"};
Linear blending, step 1 NOTE: Linear blending is currently disabled in this CL as the feature is still a work in progress Android currently performs all blending (any kind of linear math on colors really) on gamma-encoded colors. Since Android assumes that the default color space is sRGB, all bitmaps and colors are encoded with the sRGB Opto-Electronic Conversion Function (OECF, which can be approximated with a power function). Since the power curve is not linear, our linear math is incorrect. The result is that we generate colors that tend to be too dark; this affects blending but also anti-aliasing, gradients, blurs, etc. The solution is to convert gamma-encoded colors back to linear space before doing any math on them, using the sRGB Electo-Optical Conversion Function (EOCF). This is achieved in different ways in different parts of the pipeline: - Using hardware conversions when sampling from OpenGL textures or writing into OpenGL frame buffers - Using software conversion functions, to translate app-supplied colors to and from sRGB - Using Skia's color spaces Any type of processing on colors must roughly ollow these steps: [sRGB input]->EOCF->[linear data]->[processing]->OECF->[sRGB output] For the sRGB color space, the conversion functions are defined as follows: OECF(linear) := linear <= 0.0031308 ? linear * 12.92 : (pow(linear, 1/2.4) * 1.055) - 0.055 EOCF(srgb) := srgb <= 0.04045 ? srgb / 12.92 : pow((srgb + 0.055) / 1.055, 2.4) The EOCF is simply the reciprocal of the OECF. While it is highly recommended to use the exact sRGB conversion functions everywhere possible, it is sometimes useful or beneficial to rely on approximations: - pow(x,2.2) and pow(x,1/2.2) - x^2 and sqrt(x) The latter is particularly useful in fragment shaders (for instance to apply dithering in sRGB space), especially if the sqrt() can be replaced with an inversesqrt(). Here is a fairly exhaustive list of modifications implemented in this CL: - Set TARGET_ENABLE_LINEAR_BLENDING := false in BoardConfig.mk to disable linear blending. This is only for GLES 2.0 GPUs with no hardware sRGB support. This flag is currently assumed to be false (see note above) - sRGB writes are disabled when entering a functor (WebView). This will need to be fixed at some point - Skia bitmaps are created with the sRGB color space - Bitmaps using a 565 config are expanded to 888 - Linear blending is disabled when entering a functor - External textures are not properly sampled (see below) - Gradients are interpolated in linear space - Texture-based dithering was replaced with analytical dithering - Dithering is done in the quantization color space, which is why we must do EOCF(OECF(color)+dither) - Text is now gamma corrected differently depending on the luminance of the source pixel. The asumption is that a bright pixel will be blended on a dark background and the other way around. The source alpha is gamma corrected to thicken dark on bright and thin bright on dark to match the intended design of fonts. This also matches the behavior of popular design/drawing applications - Removed the asset atlas. It did not contain anything useful and could not be sampled in sRGB without a yet-to-be-defined GL extension - The last column of color matrices is converted to linear space because its value are added to linear colors Missing features: - Resource qualifier? - Regeneration of goldeng images for automated tests - Handle alpha8/grey8 properly - Disable sRGB write for layers with external textures Test: Manual testing while work in progress Bug: 29940137 Change-Id: I6a07b15ab49b554377cd33a36b6d9971a15e9a0b
2016-09-28 17:34:42 -07:00
const char* gFS_Main_FetchA8Texture[4] = {
// Don't modulate
" fragColor = texture2D(baseSampler, outTexCoords);\n",
Linear blending, step 1 NOTE: Linear blending is currently disabled in this CL as the feature is still a work in progress Android currently performs all blending (any kind of linear math on colors really) on gamma-encoded colors. Since Android assumes that the default color space is sRGB, all bitmaps and colors are encoded with the sRGB Opto-Electronic Conversion Function (OECF, which can be approximated with a power function). Since the power curve is not linear, our linear math is incorrect. The result is that we generate colors that tend to be too dark; this affects blending but also anti-aliasing, gradients, blurs, etc. The solution is to convert gamma-encoded colors back to linear space before doing any math on them, using the sRGB Electo-Optical Conversion Function (EOCF). This is achieved in different ways in different parts of the pipeline: - Using hardware conversions when sampling from OpenGL textures or writing into OpenGL frame buffers - Using software conversion functions, to translate app-supplied colors to and from sRGB - Using Skia's color spaces Any type of processing on colors must roughly ollow these steps: [sRGB input]->EOCF->[linear data]->[processing]->OECF->[sRGB output] For the sRGB color space, the conversion functions are defined as follows: OECF(linear) := linear <= 0.0031308 ? linear * 12.92 : (pow(linear, 1/2.4) * 1.055) - 0.055 EOCF(srgb) := srgb <= 0.04045 ? srgb / 12.92 : pow((srgb + 0.055) / 1.055, 2.4) The EOCF is simply the reciprocal of the OECF. While it is highly recommended to use the exact sRGB conversion functions everywhere possible, it is sometimes useful or beneficial to rely on approximations: - pow(x,2.2) and pow(x,1/2.2) - x^2 and sqrt(x) The latter is particularly useful in fragment shaders (for instance to apply dithering in sRGB space), especially if the sqrt() can be replaced with an inversesqrt(). Here is a fairly exhaustive list of modifications implemented in this CL: - Set TARGET_ENABLE_LINEAR_BLENDING := false in BoardConfig.mk to disable linear blending. This is only for GLES 2.0 GPUs with no hardware sRGB support. This flag is currently assumed to be false (see note above) - sRGB writes are disabled when entering a functor (WebView). This will need to be fixed at some point - Skia bitmaps are created with the sRGB color space - Bitmaps using a 565 config are expanded to 888 - Linear blending is disabled when entering a functor - External textures are not properly sampled (see below) - Gradients are interpolated in linear space - Texture-based dithering was replaced with analytical dithering - Dithering is done in the quantization color space, which is why we must do EOCF(OECF(color)+dither) - Text is now gamma corrected differently depending on the luminance of the source pixel. The asumption is that a bright pixel will be blended on a dark background and the other way around. The source alpha is gamma corrected to thicken dark on bright and thin bright on dark to match the intended design of fonts. This also matches the behavior of popular design/drawing applications - Removed the asset atlas. It did not contain anything useful and could not be sampled in sRGB without a yet-to-be-defined GL extension - The last column of color matrices is converted to linear space because its value are added to linear colors Missing features: - Resource qualifier? - Regeneration of goldeng images for automated tests - Handle alpha8/grey8 properly - Disable sRGB write for layers with external textures Test: Manual testing while work in progress Bug: 29940137 Change-Id: I6a07b15ab49b554377cd33a36b6d9971a15e9a0b
2016-09-28 17:34:42 -07:00
" fragColor = texture2D(baseSampler, outTexCoords);\n",
// Modulate
" fragColor = color * texture2D(baseSampler, outTexCoords).a;\n",
Linear blending, step 1 NOTE: Linear blending is currently disabled in this CL as the feature is still a work in progress Android currently performs all blending (any kind of linear math on colors really) on gamma-encoded colors. Since Android assumes that the default color space is sRGB, all bitmaps and colors are encoded with the sRGB Opto-Electronic Conversion Function (OECF, which can be approximated with a power function). Since the power curve is not linear, our linear math is incorrect. The result is that we generate colors that tend to be too dark; this affects blending but also anti-aliasing, gradients, blurs, etc. The solution is to convert gamma-encoded colors back to linear space before doing any math on them, using the sRGB Electo-Optical Conversion Function (EOCF). This is achieved in different ways in different parts of the pipeline: - Using hardware conversions when sampling from OpenGL textures or writing into OpenGL frame buffers - Using software conversion functions, to translate app-supplied colors to and from sRGB - Using Skia's color spaces Any type of processing on colors must roughly ollow these steps: [sRGB input]->EOCF->[linear data]->[processing]->OECF->[sRGB output] For the sRGB color space, the conversion functions are defined as follows: OECF(linear) := linear <= 0.0031308 ? linear * 12.92 : (pow(linear, 1/2.4) * 1.055) - 0.055 EOCF(srgb) := srgb <= 0.04045 ? srgb / 12.92 : pow((srgb + 0.055) / 1.055, 2.4) The EOCF is simply the reciprocal of the OECF. While it is highly recommended to use the exact sRGB conversion functions everywhere possible, it is sometimes useful or beneficial to rely on approximations: - pow(x,2.2) and pow(x,1/2.2) - x^2 and sqrt(x) The latter is particularly useful in fragment shaders (for instance to apply dithering in sRGB space), especially if the sqrt() can be replaced with an inversesqrt(). Here is a fairly exhaustive list of modifications implemented in this CL: - Set TARGET_ENABLE_LINEAR_BLENDING := false in BoardConfig.mk to disable linear blending. This is only for GLES 2.0 GPUs with no hardware sRGB support. This flag is currently assumed to be false (see note above) - sRGB writes are disabled when entering a functor (WebView). This will need to be fixed at some point - Skia bitmaps are created with the sRGB color space - Bitmaps using a 565 config are expanded to 888 - Linear blending is disabled when entering a functor - External textures are not properly sampled (see below) - Gradients are interpolated in linear space - Texture-based dithering was replaced with analytical dithering - Dithering is done in the quantization color space, which is why we must do EOCF(OECF(color)+dither) - Text is now gamma corrected differently depending on the luminance of the source pixel. The asumption is that a bright pixel will be blended on a dark background and the other way around. The source alpha is gamma corrected to thicken dark on bright and thin bright on dark to match the intended design of fonts. This also matches the behavior of popular design/drawing applications - Removed the asset atlas. It did not contain anything useful and could not be sampled in sRGB without a yet-to-be-defined GL extension - The last column of color matrices is converted to linear space because its value are added to linear colors Missing features: - Resource qualifier? - Regeneration of goldeng images for automated tests - Handle alpha8/grey8 properly - Disable sRGB write for layers with external textures Test: Manual testing while work in progress Bug: 29940137 Change-Id: I6a07b15ab49b554377cd33a36b6d9971a15e9a0b
2016-09-28 17:34:42 -07:00
" fragColor = color * gamma(texture2D(baseSampler, outTexCoords).a, color.rgb);\n",
};
const char* gFS_Main_FetchGradient[6] = {
// Linear
" vec4 gradientColor = texture2D(gradientSampler, linear);\n",
" vec4 gradientColor = mix(startColor, endColor, clamp(linear, 0.0, 1.0));\n",
// Circular
" vec4 gradientColor = texture2D(gradientSampler, vec2(length(circular), 0.5));\n",
" vec4 gradientColor = mix(startColor, endColor, clamp(length(circular), 0.0, 1.0));\n",
// Sweep
" highp float index = atan(sweep.y, sweep.x) * 0.15915494309; // inv(2 * PI)\n"
" vec4 gradientColor = texture2D(gradientSampler, vec2(index - floor(index), 0.5));\n",
" highp float index = atan(sweep.y, sweep.x) * 0.15915494309; // inv(2 * PI)\n"
" vec4 gradientColor = mix(startColor, endColor, clamp(index - floor(index), 0.0, "
"1.0));\n"};
const char* gFS_Main_FetchBitmap =
" vec4 bitmapColor = colorConvert(texture2D(bitmapSampler, outBitmapTexCoords));\n";
const char* gFS_Main_FetchBitmapNpot =
" vec4 bitmapColor = colorConvert(texture2D(bitmapSampler, "
"wrap(outBitmapTexCoords)));\n";
const char* gFS_Main_BlendShadersBG = " fragColor = blendShaders(gradientColor, bitmapColor)";
const char* gFS_Main_BlendShadersGB = " fragColor = blendShaders(bitmapColor, gradientColor)";
Linear blending, step 1 NOTE: Linear blending is currently disabled in this CL as the feature is still a work in progress Android currently performs all blending (any kind of linear math on colors really) on gamma-encoded colors. Since Android assumes that the default color space is sRGB, all bitmaps and colors are encoded with the sRGB Opto-Electronic Conversion Function (OECF, which can be approximated with a power function). Since the power curve is not linear, our linear math is incorrect. The result is that we generate colors that tend to be too dark; this affects blending but also anti-aliasing, gradients, blurs, etc. The solution is to convert gamma-encoded colors back to linear space before doing any math on them, using the sRGB Electo-Optical Conversion Function (EOCF). This is achieved in different ways in different parts of the pipeline: - Using hardware conversions when sampling from OpenGL textures or writing into OpenGL frame buffers - Using software conversion functions, to translate app-supplied colors to and from sRGB - Using Skia's color spaces Any type of processing on colors must roughly ollow these steps: [sRGB input]->EOCF->[linear data]->[processing]->OECF->[sRGB output] For the sRGB color space, the conversion functions are defined as follows: OECF(linear) := linear <= 0.0031308 ? linear * 12.92 : (pow(linear, 1/2.4) * 1.055) - 0.055 EOCF(srgb) := srgb <= 0.04045 ? srgb / 12.92 : pow((srgb + 0.055) / 1.055, 2.4) The EOCF is simply the reciprocal of the OECF. While it is highly recommended to use the exact sRGB conversion functions everywhere possible, it is sometimes useful or beneficial to rely on approximations: - pow(x,2.2) and pow(x,1/2.2) - x^2 and sqrt(x) The latter is particularly useful in fragment shaders (for instance to apply dithering in sRGB space), especially if the sqrt() can be replaced with an inversesqrt(). Here is a fairly exhaustive list of modifications implemented in this CL: - Set TARGET_ENABLE_LINEAR_BLENDING := false in BoardConfig.mk to disable linear blending. This is only for GLES 2.0 GPUs with no hardware sRGB support. This flag is currently assumed to be false (see note above) - sRGB writes are disabled when entering a functor (WebView). This will need to be fixed at some point - Skia bitmaps are created with the sRGB color space - Bitmaps using a 565 config are expanded to 888 - Linear blending is disabled when entering a functor - External textures are not properly sampled (see below) - Gradients are interpolated in linear space - Texture-based dithering was replaced with analytical dithering - Dithering is done in the quantization color space, which is why we must do EOCF(OECF(color)+dither) - Text is now gamma corrected differently depending on the luminance of the source pixel. The asumption is that a bright pixel will be blended on a dark background and the other way around. The source alpha is gamma corrected to thicken dark on bright and thin bright on dark to match the intended design of fonts. This also matches the behavior of popular design/drawing applications - Removed the asset atlas. It did not contain anything useful and could not be sampled in sRGB without a yet-to-be-defined GL extension - The last column of color matrices is converted to linear space because its value are added to linear colors Missing features: - Resource qualifier? - Regeneration of goldeng images for automated tests - Handle alpha8/grey8 properly - Disable sRGB write for layers with external textures Test: Manual testing while work in progress Bug: 29940137 Change-Id: I6a07b15ab49b554377cd33a36b6d9971a15e9a0b
2016-09-28 17:34:42 -07:00
const char* gFS_Main_BlendShaders_Modulate[6] = {
// Don't modulate
";\n", ";\n",
// Modulate
" * color.a;\n", " * color.a;\n",
// Modulate with alpha 8 texture
" * texture2D(baseSampler, outTexCoords).a;\n",
Linear blending, step 1 NOTE: Linear blending is currently disabled in this CL as the feature is still a work in progress Android currently performs all blending (any kind of linear math on colors really) on gamma-encoded colors. Since Android assumes that the default color space is sRGB, all bitmaps and colors are encoded with the sRGB Opto-Electronic Conversion Function (OECF, which can be approximated with a power function). Since the power curve is not linear, our linear math is incorrect. The result is that we generate colors that tend to be too dark; this affects blending but also anti-aliasing, gradients, blurs, etc. The solution is to convert gamma-encoded colors back to linear space before doing any math on them, using the sRGB Electo-Optical Conversion Function (EOCF). This is achieved in different ways in different parts of the pipeline: - Using hardware conversions when sampling from OpenGL textures or writing into OpenGL frame buffers - Using software conversion functions, to translate app-supplied colors to and from sRGB - Using Skia's color spaces Any type of processing on colors must roughly ollow these steps: [sRGB input]->EOCF->[linear data]->[processing]->OECF->[sRGB output] For the sRGB color space, the conversion functions are defined as follows: OECF(linear) := linear <= 0.0031308 ? linear * 12.92 : (pow(linear, 1/2.4) * 1.055) - 0.055 EOCF(srgb) := srgb <= 0.04045 ? srgb / 12.92 : pow((srgb + 0.055) / 1.055, 2.4) The EOCF is simply the reciprocal of the OECF. While it is highly recommended to use the exact sRGB conversion functions everywhere possible, it is sometimes useful or beneficial to rely on approximations: - pow(x,2.2) and pow(x,1/2.2) - x^2 and sqrt(x) The latter is particularly useful in fragment shaders (for instance to apply dithering in sRGB space), especially if the sqrt() can be replaced with an inversesqrt(). Here is a fairly exhaustive list of modifications implemented in this CL: - Set TARGET_ENABLE_LINEAR_BLENDING := false in BoardConfig.mk to disable linear blending. This is only for GLES 2.0 GPUs with no hardware sRGB support. This flag is currently assumed to be false (see note above) - sRGB writes are disabled when entering a functor (WebView). This will need to be fixed at some point - Skia bitmaps are created with the sRGB color space - Bitmaps using a 565 config are expanded to 888 - Linear blending is disabled when entering a functor - External textures are not properly sampled (see below) - Gradients are interpolated in linear space - Texture-based dithering was replaced with analytical dithering - Dithering is done in the quantization color space, which is why we must do EOCF(OECF(color)+dither) - Text is now gamma corrected differently depending on the luminance of the source pixel. The asumption is that a bright pixel will be blended on a dark background and the other way around. The source alpha is gamma corrected to thicken dark on bright and thin bright on dark to match the intended design of fonts. This also matches the behavior of popular design/drawing applications - Removed the asset atlas. It did not contain anything useful and could not be sampled in sRGB without a yet-to-be-defined GL extension - The last column of color matrices is converted to linear space because its value are added to linear colors Missing features: - Resource qualifier? - Regeneration of goldeng images for automated tests - Handle alpha8/grey8 properly - Disable sRGB write for layers with external textures Test: Manual testing while work in progress Bug: 29940137 Change-Id: I6a07b15ab49b554377cd33a36b6d9971a15e9a0b
2016-09-28 17:34:42 -07:00
" * gamma(texture2D(baseSampler, outTexCoords).a, color.rgb);\n",
};
Linear blending, step 1 NOTE: Linear blending is currently disabled in this CL as the feature is still a work in progress Android currently performs all blending (any kind of linear math on colors really) on gamma-encoded colors. Since Android assumes that the default color space is sRGB, all bitmaps and colors are encoded with the sRGB Opto-Electronic Conversion Function (OECF, which can be approximated with a power function). Since the power curve is not linear, our linear math is incorrect. The result is that we generate colors that tend to be too dark; this affects blending but also anti-aliasing, gradients, blurs, etc. The solution is to convert gamma-encoded colors back to linear space before doing any math on them, using the sRGB Electo-Optical Conversion Function (EOCF). This is achieved in different ways in different parts of the pipeline: - Using hardware conversions when sampling from OpenGL textures or writing into OpenGL frame buffers - Using software conversion functions, to translate app-supplied colors to and from sRGB - Using Skia's color spaces Any type of processing on colors must roughly ollow these steps: [sRGB input]->EOCF->[linear data]->[processing]->OECF->[sRGB output] For the sRGB color space, the conversion functions are defined as follows: OECF(linear) := linear <= 0.0031308 ? linear * 12.92 : (pow(linear, 1/2.4) * 1.055) - 0.055 EOCF(srgb) := srgb <= 0.04045 ? srgb / 12.92 : pow((srgb + 0.055) / 1.055, 2.4) The EOCF is simply the reciprocal of the OECF. While it is highly recommended to use the exact sRGB conversion functions everywhere possible, it is sometimes useful or beneficial to rely on approximations: - pow(x,2.2) and pow(x,1/2.2) - x^2 and sqrt(x) The latter is particularly useful in fragment shaders (for instance to apply dithering in sRGB space), especially if the sqrt() can be replaced with an inversesqrt(). Here is a fairly exhaustive list of modifications implemented in this CL: - Set TARGET_ENABLE_LINEAR_BLENDING := false in BoardConfig.mk to disable linear blending. This is only for GLES 2.0 GPUs with no hardware sRGB support. This flag is currently assumed to be false (see note above) - sRGB writes are disabled when entering a functor (WebView). This will need to be fixed at some point - Skia bitmaps are created with the sRGB color space - Bitmaps using a 565 config are expanded to 888 - Linear blending is disabled when entering a functor - External textures are not properly sampled (see below) - Gradients are interpolated in linear space - Texture-based dithering was replaced with analytical dithering - Dithering is done in the quantization color space, which is why we must do EOCF(OECF(color)+dither) - Text is now gamma corrected differently depending on the luminance of the source pixel. The asumption is that a bright pixel will be blended on a dark background and the other way around. The source alpha is gamma corrected to thicken dark on bright and thin bright on dark to match the intended design of fonts. This also matches the behavior of popular design/drawing applications - Removed the asset atlas. It did not contain anything useful and could not be sampled in sRGB without a yet-to-be-defined GL extension - The last column of color matrices is converted to linear space because its value are added to linear colors Missing features: - Resource qualifier? - Regeneration of goldeng images for automated tests - Handle alpha8/grey8 properly - Disable sRGB write for layers with external textures Test: Manual testing while work in progress Bug: 29940137 Change-Id: I6a07b15ab49b554377cd33a36b6d9971a15e9a0b
2016-09-28 17:34:42 -07:00
const char* gFS_Main_GradientShader_Modulate[6] = {
// Don't modulate
" fragColor = gradientColor;\n", " fragColor = gradientColor;\n",
// Modulate
" fragColor = gradientColor * color.a;\n", " fragColor = gradientColor * color.a;\n",
// Modulate with alpha 8 texture
" fragColor = gradientColor * texture2D(baseSampler, outTexCoords).a;\n",
" fragColor = gradientColor * gamma(texture2D(baseSampler, outTexCoords).a, "
"gradientColor.rgb);\n",
};
Linear blending, step 1 NOTE: Linear blending is currently disabled in this CL as the feature is still a work in progress Android currently performs all blending (any kind of linear math on colors really) on gamma-encoded colors. Since Android assumes that the default color space is sRGB, all bitmaps and colors are encoded with the sRGB Opto-Electronic Conversion Function (OECF, which can be approximated with a power function). Since the power curve is not linear, our linear math is incorrect. The result is that we generate colors that tend to be too dark; this affects blending but also anti-aliasing, gradients, blurs, etc. The solution is to convert gamma-encoded colors back to linear space before doing any math on them, using the sRGB Electo-Optical Conversion Function (EOCF). This is achieved in different ways in different parts of the pipeline: - Using hardware conversions when sampling from OpenGL textures or writing into OpenGL frame buffers - Using software conversion functions, to translate app-supplied colors to and from sRGB - Using Skia's color spaces Any type of processing on colors must roughly ollow these steps: [sRGB input]->EOCF->[linear data]->[processing]->OECF->[sRGB output] For the sRGB color space, the conversion functions are defined as follows: OECF(linear) := linear <= 0.0031308 ? linear * 12.92 : (pow(linear, 1/2.4) * 1.055) - 0.055 EOCF(srgb) := srgb <= 0.04045 ? srgb / 12.92 : pow((srgb + 0.055) / 1.055, 2.4) The EOCF is simply the reciprocal of the OECF. While it is highly recommended to use the exact sRGB conversion functions everywhere possible, it is sometimes useful or beneficial to rely on approximations: - pow(x,2.2) and pow(x,1/2.2) - x^2 and sqrt(x) The latter is particularly useful in fragment shaders (for instance to apply dithering in sRGB space), especially if the sqrt() can be replaced with an inversesqrt(). Here is a fairly exhaustive list of modifications implemented in this CL: - Set TARGET_ENABLE_LINEAR_BLENDING := false in BoardConfig.mk to disable linear blending. This is only for GLES 2.0 GPUs with no hardware sRGB support. This flag is currently assumed to be false (see note above) - sRGB writes are disabled when entering a functor (WebView). This will need to be fixed at some point - Skia bitmaps are created with the sRGB color space - Bitmaps using a 565 config are expanded to 888 - Linear blending is disabled when entering a functor - External textures are not properly sampled (see below) - Gradients are interpolated in linear space - Texture-based dithering was replaced with analytical dithering - Dithering is done in the quantization color space, which is why we must do EOCF(OECF(color)+dither) - Text is now gamma corrected differently depending on the luminance of the source pixel. The asumption is that a bright pixel will be blended on a dark background and the other way around. The source alpha is gamma corrected to thicken dark on bright and thin bright on dark to match the intended design of fonts. This also matches the behavior of popular design/drawing applications - Removed the asset atlas. It did not contain anything useful and could not be sampled in sRGB without a yet-to-be-defined GL extension - The last column of color matrices is converted to linear space because its value are added to linear colors Missing features: - Resource qualifier? - Regeneration of goldeng images for automated tests - Handle alpha8/grey8 properly - Disable sRGB write for layers with external textures Test: Manual testing while work in progress Bug: 29940137 Change-Id: I6a07b15ab49b554377cd33a36b6d9971a15e9a0b
2016-09-28 17:34:42 -07:00
const char* gFS_Main_BitmapShader_Modulate[6] = {
// Don't modulate
" fragColor = bitmapColor;\n", " fragColor = bitmapColor;\n",
// Modulate
" fragColor = bitmapColor * color.a;\n", " fragColor = bitmapColor * color.a;\n",
// Modulate with alpha 8 texture
" fragColor = bitmapColor * texture2D(baseSampler, outTexCoords).a;\n",
" fragColor = bitmapColor * gamma(texture2D(baseSampler, outTexCoords).a, "
"bitmapColor.rgb);\n",
};
const char* gFS_Main_FragColor = " gl_FragColor = fragColor;\n";
const char* gFS_Main_FragColor_HasColors = " gl_FragColor *= outColors;\n";
const char* gFS_Main_FragColor_Blend =
" gl_FragColor = blendFramebuffer(fragColor, gl_LastFragColor);\n";
const char* gFS_Main_FragColor_Blend_Swap =
" gl_FragColor = blendFramebuffer(gl_LastFragColor, fragColor);\n";
const char* gFS_Main_ApplyColorOp[3] = {
// None
"",
// Matrix
" fragColor.rgb /= (fragColor.a + 0.0019);\n" // un-premultiply
" fragColor *= colorMatrix;\n"
" fragColor += colorMatrixVector;\n"
" fragColor.rgb *= (fragColor.a + 0.0019);\n", // re-premultiply
// PorterDuff
" fragColor = blendColors(colorBlend, fragColor);\n"};
// Note: LTWH (left top width height) -> xyzw
// roundRectPos is now divided by roundRectRadius in vertex shader
// after we also subtract roundRectInnerRectLTWH.xy from roundRectPos
const char* gFS_Main_FragColor_HasRoundRectClip =
" mediump vec2 fragToLT = -roundRectPos;\n"
" mediump vec2 fragFromRB = roundRectPos - roundRectInnerRectLTWH.zw;\n"
// since distance is divided by radius, it's in [0;1] so precision is not an issue
// this also lets us clamp(0.0, 1.0) instead of max() which is cheaper on GPUs
" mediump vec2 dist = clamp(max(fragToLT, fragFromRB), 0.0, 1.0);\n"
" mediump float linearDist = clamp(roundRectRadius - (length(dist) * roundRectRadius), "
"0.0, 1.0);\n"
" gl_FragColor *= linearDist;\n";
const char* gFS_Main_DebugHighlight = " gl_FragColor.rgb = vec3(0.0, gl_FragColor.a, 0.0);\n";
const char* gFS_Footer = "}\n\n";
///////////////////////////////////////////////////////////////////////////////
// PorterDuff snippets
///////////////////////////////////////////////////////////////////////////////
const char* gBlendOps[18] = {
// Clear
"return vec4(0.0, 0.0, 0.0, 0.0);\n",
// Src
"return src;\n",
// Dst
"return dst;\n",
// SrcOver
"return src + dst * (1.0 - src.a);\n",
// DstOver
"return dst + src * (1.0 - dst.a);\n",
// SrcIn
"return src * dst.a;\n",
// DstIn
"return dst * src.a;\n",
// SrcOut
"return src * (1.0 - dst.a);\n",
// DstOut
"return dst * (1.0 - src.a);\n",
// SrcAtop
"return vec4(src.rgb * dst.a + (1.0 - src.a) * dst.rgb, dst.a);\n",
// DstAtop
"return vec4(dst.rgb * src.a + (1.0 - dst.a) * src.rgb, src.a);\n",
// Xor
"return vec4(src.rgb * (1.0 - dst.a) + (1.0 - src.a) * dst.rgb, "
"src.a + dst.a - 2.0 * src.a * dst.a);\n",
// Plus
"return min(src + dst, 1.0);\n",
// Modulate
"return src * dst;\n",
// Screen
"return src + dst - src * dst;\n",
// Overlay
"return clamp(vec4(mix("
"2.0 * src.rgb * dst.rgb + src.rgb * (1.0 - dst.a) + dst.rgb * (1.0 - src.a), "
"src.a * dst.a - 2.0 * (dst.a - dst.rgb) * (src.a - src.rgb) + src.rgb * (1.0 - dst.a) + "
"dst.rgb * (1.0 - src.a), "
"step(dst.a, 2.0 * dst.rgb)), "
"src.a + dst.a - src.a * dst.a), 0.0, 1.0);\n",
// Darken
"return vec4(src.rgb * (1.0 - dst.a) + (1.0 - src.a) * dst.rgb + "
"min(src.rgb * dst.a, dst.rgb * src.a), src.a + dst.a - src.a * dst.a);\n",
// Lighten
"return vec4(src.rgb * (1.0 - dst.a) + (1.0 - src.a) * dst.rgb + "
"max(src.rgb * dst.a, dst.rgb * src.a), src.a + dst.a - src.a * dst.a);\n",
};
///////////////////////////////////////////////////////////////////////////////
// Constructors/destructors
///////////////////////////////////////////////////////////////////////////////
ProgramCache::ProgramCache(const Extensions& extensions)
Linear blending, step 1 NOTE: Linear blending is currently disabled in this CL as the feature is still a work in progress Android currently performs all blending (any kind of linear math on colors really) on gamma-encoded colors. Since Android assumes that the default color space is sRGB, all bitmaps and colors are encoded with the sRGB Opto-Electronic Conversion Function (OECF, which can be approximated with a power function). Since the power curve is not linear, our linear math is incorrect. The result is that we generate colors that tend to be too dark; this affects blending but also anti-aliasing, gradients, blurs, etc. The solution is to convert gamma-encoded colors back to linear space before doing any math on them, using the sRGB Electo-Optical Conversion Function (EOCF). This is achieved in different ways in different parts of the pipeline: - Using hardware conversions when sampling from OpenGL textures or writing into OpenGL frame buffers - Using software conversion functions, to translate app-supplied colors to and from sRGB - Using Skia's color spaces Any type of processing on colors must roughly ollow these steps: [sRGB input]->EOCF->[linear data]->[processing]->OECF->[sRGB output] For the sRGB color space, the conversion functions are defined as follows: OECF(linear) := linear <= 0.0031308 ? linear * 12.92 : (pow(linear, 1/2.4) * 1.055) - 0.055 EOCF(srgb) := srgb <= 0.04045 ? srgb / 12.92 : pow((srgb + 0.055) / 1.055, 2.4) The EOCF is simply the reciprocal of the OECF. While it is highly recommended to use the exact sRGB conversion functions everywhere possible, it is sometimes useful or beneficial to rely on approximations: - pow(x,2.2) and pow(x,1/2.2) - x^2 and sqrt(x) The latter is particularly useful in fragment shaders (for instance to apply dithering in sRGB space), especially if the sqrt() can be replaced with an inversesqrt(). Here is a fairly exhaustive list of modifications implemented in this CL: - Set TARGET_ENABLE_LINEAR_BLENDING := false in BoardConfig.mk to disable linear blending. This is only for GLES 2.0 GPUs with no hardware sRGB support. This flag is currently assumed to be false (see note above) - sRGB writes are disabled when entering a functor (WebView). This will need to be fixed at some point - Skia bitmaps are created with the sRGB color space - Bitmaps using a 565 config are expanded to 888 - Linear blending is disabled when entering a functor - External textures are not properly sampled (see below) - Gradients are interpolated in linear space - Texture-based dithering was replaced with analytical dithering - Dithering is done in the quantization color space, which is why we must do EOCF(OECF(color)+dither) - Text is now gamma corrected differently depending on the luminance of the source pixel. The asumption is that a bright pixel will be blended on a dark background and the other way around. The source alpha is gamma corrected to thicken dark on bright and thin bright on dark to match the intended design of fonts. This also matches the behavior of popular design/drawing applications - Removed the asset atlas. It did not contain anything useful and could not be sampled in sRGB without a yet-to-be-defined GL extension - The last column of color matrices is converted to linear space because its value are added to linear colors Missing features: - Resource qualifier? - Regeneration of goldeng images for automated tests - Handle alpha8/grey8 properly - Disable sRGB write for layers with external textures Test: Manual testing while work in progress Bug: 29940137 Change-Id: I6a07b15ab49b554377cd33a36b6d9971a15e9a0b
2016-09-28 17:34:42 -07:00
: mHasES3(extensions.getMajorGlVersion() >= 3)
, mHasLinearBlending(extensions.hasLinearBlending()) {}
ProgramCache::~ProgramCache() {
clear();
}
///////////////////////////////////////////////////////////////////////////////
// Cache management
///////////////////////////////////////////////////////////////////////////////
void ProgramCache::clear() {
PROGRAM_LOGD("Clearing program cache");
mCache.clear();
}
Program* ProgramCache::get(const ProgramDescription& description) {
programid key = description.key();
if (key == (PROGRAM_KEY_TEXTURE | PROGRAM_KEY_A8_TEXTURE)) {
// program for A8, unmodulated, texture w/o shader (black text/path textures) is equivalent
// to standard texture program (bitmaps, patches). Consider them equivalent.
key = PROGRAM_KEY_TEXTURE;
}
auto iter = mCache.find(key);
Program* program = nullptr;
if (iter == mCache.end()) {
description.log("Could not find program");
program = generateProgram(description, key);
mCache[key] = std::unique_ptr<Program>(program);
} else {
program = iter->second.get();
}
return program;
}
///////////////////////////////////////////////////////////////////////////////
// Program generation
///////////////////////////////////////////////////////////////////////////////
Program* ProgramCache::generateProgram(const ProgramDescription& description, programid key) {
String8 vertexShader = generateVertexShader(description);
String8 fragmentShader = generateFragmentShader(description);
return new Program(description, vertexShader.string(), fragmentShader.string());
}
static inline size_t gradientIndex(const ProgramDescription& description) {
return description.gradientType * 2 + description.isSimpleGradient;
}
String8 ProgramCache::generateVertexShader(const ProgramDescription& description) {
// Add attributes
String8 shader(gVS_Header_Start);
if (description.hasTexture || description.hasExternalTexture) {
shader.append(gVS_Header_Attributes_TexCoords);
}
if (description.hasVertexAlpha) {
shader.append(gVS_Header_Attributes_VertexAlphaParameters);
}
if (description.hasColors) {
shader.append(gVS_Header_Attributes_Colors);
}
// Uniforms
shader.append(gVS_Header_Uniforms);
if (description.hasTextureTransform) {
shader.append(gVS_Header_Uniforms_TextureTransform);
}
if (description.hasGradient) {
shader.append(gVS_Header_Uniforms_HasGradient);
}
if (description.hasBitmap) {
shader.append(gVS_Header_Uniforms_HasBitmap);
}
if (description.hasRoundRectClip) {
shader.append(gVS_Header_Uniforms_HasRoundRectClip);
}
// Varyings
if (description.hasTexture || description.hasExternalTexture) {
shader.append(gVS_Header_Varyings_HasTexture);
}
if (description.hasVertexAlpha) {
shader.append(gVS_Header_Varyings_HasVertexAlpha);
}
if (description.hasColors) {
shader.append(gVS_Header_Varyings_HasColors);
}
if (description.hasGradient) {
shader.append(gVS_Header_Varyings_HasGradient[gradientIndex(description)]);
}
if (description.hasBitmap) {
shader.append(gVS_Header_Varyings_HasBitmap);
}
if (description.hasRoundRectClip) {
shader.append(gVS_Header_Varyings_HasRoundRectClip);
}
// Begin the shader
shader.append(gVS_Main);
{
if (description.hasTextureTransform) {
shader.append(gVS_Main_OutTransformedTexCoords);
} else if (description.hasTexture || description.hasExternalTexture) {
shader.append(gVS_Main_OutTexCoords);
}
if (description.hasVertexAlpha) {
shader.append(gVS_Main_VertexAlpha);
}
if (description.hasColors) {
shader.append(gVS_Main_OutColors);
}
if (description.hasBitmap) {
shader.append(gVS_Main_OutBitmapTexCoords);
}
// Output transformed position
shader.append(gVS_Main_Position);
if (description.hasGradient) {
shader.append(gVS_Main_OutGradient[gradientIndex(description)]);
}
if (description.hasRoundRectClip) {
shader.append(gVS_Main_HasRoundRectClip);
}
}
// End the shader
shader.append(gVS_Footer);
PROGRAM_LOGD("*** Generated vertex shader:\n\n%s", shader.string());
return shader;
}
static bool shaderOp(const ProgramDescription& description, String8& shader, const int modulateOp,
const char** snippets) {
int op = description.hasAlpha8Texture ? MODULATE_OP_MODULATE_A8 : modulateOp;
Linear blending, step 1 NOTE: Linear blending is currently disabled in this CL as the feature is still a work in progress Android currently performs all blending (any kind of linear math on colors really) on gamma-encoded colors. Since Android assumes that the default color space is sRGB, all bitmaps and colors are encoded with the sRGB Opto-Electronic Conversion Function (OECF, which can be approximated with a power function). Since the power curve is not linear, our linear math is incorrect. The result is that we generate colors that tend to be too dark; this affects blending but also anti-aliasing, gradients, blurs, etc. The solution is to convert gamma-encoded colors back to linear space before doing any math on them, using the sRGB Electo-Optical Conversion Function (EOCF). This is achieved in different ways in different parts of the pipeline: - Using hardware conversions when sampling from OpenGL textures or writing into OpenGL frame buffers - Using software conversion functions, to translate app-supplied colors to and from sRGB - Using Skia's color spaces Any type of processing on colors must roughly ollow these steps: [sRGB input]->EOCF->[linear data]->[processing]->OECF->[sRGB output] For the sRGB color space, the conversion functions are defined as follows: OECF(linear) := linear <= 0.0031308 ? linear * 12.92 : (pow(linear, 1/2.4) * 1.055) - 0.055 EOCF(srgb) := srgb <= 0.04045 ? srgb / 12.92 : pow((srgb + 0.055) / 1.055, 2.4) The EOCF is simply the reciprocal of the OECF. While it is highly recommended to use the exact sRGB conversion functions everywhere possible, it is sometimes useful or beneficial to rely on approximations: - pow(x,2.2) and pow(x,1/2.2) - x^2 and sqrt(x) The latter is particularly useful in fragment shaders (for instance to apply dithering in sRGB space), especially if the sqrt() can be replaced with an inversesqrt(). Here is a fairly exhaustive list of modifications implemented in this CL: - Set TARGET_ENABLE_LINEAR_BLENDING := false in BoardConfig.mk to disable linear blending. This is only for GLES 2.0 GPUs with no hardware sRGB support. This flag is currently assumed to be false (see note above) - sRGB writes are disabled when entering a functor (WebView). This will need to be fixed at some point - Skia bitmaps are created with the sRGB color space - Bitmaps using a 565 config are expanded to 888 - Linear blending is disabled when entering a functor - External textures are not properly sampled (see below) - Gradients are interpolated in linear space - Texture-based dithering was replaced with analytical dithering - Dithering is done in the quantization color space, which is why we must do EOCF(OECF(color)+dither) - Text is now gamma corrected differently depending on the luminance of the source pixel. The asumption is that a bright pixel will be blended on a dark background and the other way around. The source alpha is gamma corrected to thicken dark on bright and thin bright on dark to match the intended design of fonts. This also matches the behavior of popular design/drawing applications - Removed the asset atlas. It did not contain anything useful and could not be sampled in sRGB without a yet-to-be-defined GL extension - The last column of color matrices is converted to linear space because its value are added to linear colors Missing features: - Resource qualifier? - Regeneration of goldeng images for automated tests - Handle alpha8/grey8 properly - Disable sRGB write for layers with external textures Test: Manual testing while work in progress Bug: 29940137 Change-Id: I6a07b15ab49b554377cd33a36b6d9971a15e9a0b
2016-09-28 17:34:42 -07:00
op = op * 2 + description.hasGammaCorrection;
shader.append(snippets[op]);
return description.hasAlpha8Texture;
}
String8 ProgramCache::generateFragmentShader(const ProgramDescription& description) {
String8 shader(gFS_Header_Start);
const bool blendFramebuffer = description.framebufferMode >= SkBlendMode::kPlus;
if (blendFramebuffer) {
shader.append(gFS_Header_Extension_FramebufferFetch);
}
if (description.hasExternalTexture ||
(description.hasBitmap && description.isShaderBitmapExternal)) {
shader.append(gFS_Header_Extension_ExternalTexture);
}
shader.append(gFS_Header);
// Varyings
if (description.hasTexture || description.hasExternalTexture) {
shader.append(gVS_Header_Varyings_HasTexture);
}
if (description.hasVertexAlpha) {
shader.append(gVS_Header_Varyings_HasVertexAlpha);
}
if (description.hasColors) {
shader.append(gVS_Header_Varyings_HasColors);
}
if (description.hasGradient) {
shader.append(gVS_Header_Varyings_HasGradient[gradientIndex(description)]);
}
if (description.hasBitmap) {
shader.append(gVS_Header_Varyings_HasBitmap);
}
if (description.hasRoundRectClip) {
shader.append(gVS_Header_Varyings_HasRoundRectClip);
}
// Uniforms
int modulateOp = MODULATE_OP_NO_MODULATE;
const bool singleColor = !description.hasTexture && !description.hasExternalTexture &&
!description.hasGradient && !description.hasBitmap;
if (description.modulate || singleColor) {
shader.append(gFS_Uniforms_Color);
if (!singleColor) modulateOp = MODULATE_OP_MODULATE;
}
if (description.hasTexture || description.useShadowAlphaInterp) {
shader.append(gFS_Uniforms_TextureSampler);
} else if (description.hasExternalTexture) {
shader.append(gFS_Uniforms_ExternalTextureSampler);
}
if (description.hasGradient) {
Linear blending, step 1 NOTE: Linear blending is currently disabled in this CL as the feature is still a work in progress Android currently performs all blending (any kind of linear math on colors really) on gamma-encoded colors. Since Android assumes that the default color space is sRGB, all bitmaps and colors are encoded with the sRGB Opto-Electronic Conversion Function (OECF, which can be approximated with a power function). Since the power curve is not linear, our linear math is incorrect. The result is that we generate colors that tend to be too dark; this affects blending but also anti-aliasing, gradients, blurs, etc. The solution is to convert gamma-encoded colors back to linear space before doing any math on them, using the sRGB Electo-Optical Conversion Function (EOCF). This is achieved in different ways in different parts of the pipeline: - Using hardware conversions when sampling from OpenGL textures or writing into OpenGL frame buffers - Using software conversion functions, to translate app-supplied colors to and from sRGB - Using Skia's color spaces Any type of processing on colors must roughly ollow these steps: [sRGB input]->EOCF->[linear data]->[processing]->OECF->[sRGB output] For the sRGB color space, the conversion functions are defined as follows: OECF(linear) := linear <= 0.0031308 ? linear * 12.92 : (pow(linear, 1/2.4) * 1.055) - 0.055 EOCF(srgb) := srgb <= 0.04045 ? srgb / 12.92 : pow((srgb + 0.055) / 1.055, 2.4) The EOCF is simply the reciprocal of the OECF. While it is highly recommended to use the exact sRGB conversion functions everywhere possible, it is sometimes useful or beneficial to rely on approximations: - pow(x,2.2) and pow(x,1/2.2) - x^2 and sqrt(x) The latter is particularly useful in fragment shaders (for instance to apply dithering in sRGB space), especially if the sqrt() can be replaced with an inversesqrt(). Here is a fairly exhaustive list of modifications implemented in this CL: - Set TARGET_ENABLE_LINEAR_BLENDING := false in BoardConfig.mk to disable linear blending. This is only for GLES 2.0 GPUs with no hardware sRGB support. This flag is currently assumed to be false (see note above) - sRGB writes are disabled when entering a functor (WebView). This will need to be fixed at some point - Skia bitmaps are created with the sRGB color space - Bitmaps using a 565 config are expanded to 888 - Linear blending is disabled when entering a functor - External textures are not properly sampled (see below) - Gradients are interpolated in linear space - Texture-based dithering was replaced with analytical dithering - Dithering is done in the quantization color space, which is why we must do EOCF(OECF(color)+dither) - Text is now gamma corrected differently depending on the luminance of the source pixel. The asumption is that a bright pixel will be blended on a dark background and the other way around. The source alpha is gamma corrected to thicken dark on bright and thin bright on dark to match the intended design of fonts. This also matches the behavior of popular design/drawing applications - Removed the asset atlas. It did not contain anything useful and could not be sampled in sRGB without a yet-to-be-defined GL extension - The last column of color matrices is converted to linear space because its value are added to linear colors Missing features: - Resource qualifier? - Regeneration of goldeng images for automated tests - Handle alpha8/grey8 properly - Disable sRGB write for layers with external textures Test: Manual testing while work in progress Bug: 29940137 Change-Id: I6a07b15ab49b554377cd33a36b6d9971a15e9a0b
2016-09-28 17:34:42 -07:00
shader.append(gFS_Uniforms_GradientSampler[description.isSimpleGradient]);
}
if (description.hasRoundRectClip) {
shader.append(gFS_Uniforms_HasRoundRectClip);
}
Linear blending, step 1 NOTE: Linear blending is currently disabled in this CL as the feature is still a work in progress Android currently performs all blending (any kind of linear math on colors really) on gamma-encoded colors. Since Android assumes that the default color space is sRGB, all bitmaps and colors are encoded with the sRGB Opto-Electronic Conversion Function (OECF, which can be approximated with a power function). Since the power curve is not linear, our linear math is incorrect. The result is that we generate colors that tend to be too dark; this affects blending but also anti-aliasing, gradients, blurs, etc. The solution is to convert gamma-encoded colors back to linear space before doing any math on them, using the sRGB Electo-Optical Conversion Function (EOCF). This is achieved in different ways in different parts of the pipeline: - Using hardware conversions when sampling from OpenGL textures or writing into OpenGL frame buffers - Using software conversion functions, to translate app-supplied colors to and from sRGB - Using Skia's color spaces Any type of processing on colors must roughly ollow these steps: [sRGB input]->EOCF->[linear data]->[processing]->OECF->[sRGB output] For the sRGB color space, the conversion functions are defined as follows: OECF(linear) := linear <= 0.0031308 ? linear * 12.92 : (pow(linear, 1/2.4) * 1.055) - 0.055 EOCF(srgb) := srgb <= 0.04045 ? srgb / 12.92 : pow((srgb + 0.055) / 1.055, 2.4) The EOCF is simply the reciprocal of the OECF. While it is highly recommended to use the exact sRGB conversion functions everywhere possible, it is sometimes useful or beneficial to rely on approximations: - pow(x,2.2) and pow(x,1/2.2) - x^2 and sqrt(x) The latter is particularly useful in fragment shaders (for instance to apply dithering in sRGB space), especially if the sqrt() can be replaced with an inversesqrt(). Here is a fairly exhaustive list of modifications implemented in this CL: - Set TARGET_ENABLE_LINEAR_BLENDING := false in BoardConfig.mk to disable linear blending. This is only for GLES 2.0 GPUs with no hardware sRGB support. This flag is currently assumed to be false (see note above) - sRGB writes are disabled when entering a functor (WebView). This will need to be fixed at some point - Skia bitmaps are created with the sRGB color space - Bitmaps using a 565 config are expanded to 888 - Linear blending is disabled when entering a functor - External textures are not properly sampled (see below) - Gradients are interpolated in linear space - Texture-based dithering was replaced with analytical dithering - Dithering is done in the quantization color space, which is why we must do EOCF(OECF(color)+dither) - Text is now gamma corrected differently depending on the luminance of the source pixel. The asumption is that a bright pixel will be blended on a dark background and the other way around. The source alpha is gamma corrected to thicken dark on bright and thin bright on dark to match the intended design of fonts. This also matches the behavior of popular design/drawing applications - Removed the asset atlas. It did not contain anything useful and could not be sampled in sRGB without a yet-to-be-defined GL extension - The last column of color matrices is converted to linear space because its value are added to linear colors Missing features: - Resource qualifier? - Regeneration of goldeng images for automated tests - Handle alpha8/grey8 properly - Disable sRGB write for layers with external textures Test: Manual testing while work in progress Bug: 29940137 Change-Id: I6a07b15ab49b554377cd33a36b6d9971a15e9a0b
2016-09-28 17:34:42 -07:00
if (description.hasGammaCorrection) {
shader.appendFormat(gFS_Gamma_Preamble, Properties::textGamma,
1.0f / Properties::textGamma);
Linear blending, step 1 NOTE: Linear blending is currently disabled in this CL as the feature is still a work in progress Android currently performs all blending (any kind of linear math on colors really) on gamma-encoded colors. Since Android assumes that the default color space is sRGB, all bitmaps and colors are encoded with the sRGB Opto-Electronic Conversion Function (OECF, which can be approximated with a power function). Since the power curve is not linear, our linear math is incorrect. The result is that we generate colors that tend to be too dark; this affects blending but also anti-aliasing, gradients, blurs, etc. The solution is to convert gamma-encoded colors back to linear space before doing any math on them, using the sRGB Electo-Optical Conversion Function (EOCF). This is achieved in different ways in different parts of the pipeline: - Using hardware conversions when sampling from OpenGL textures or writing into OpenGL frame buffers - Using software conversion functions, to translate app-supplied colors to and from sRGB - Using Skia's color spaces Any type of processing on colors must roughly ollow these steps: [sRGB input]->EOCF->[linear data]->[processing]->OECF->[sRGB output] For the sRGB color space, the conversion functions are defined as follows: OECF(linear) := linear <= 0.0031308 ? linear * 12.92 : (pow(linear, 1/2.4) * 1.055) - 0.055 EOCF(srgb) := srgb <= 0.04045 ? srgb / 12.92 : pow((srgb + 0.055) / 1.055, 2.4) The EOCF is simply the reciprocal of the OECF. While it is highly recommended to use the exact sRGB conversion functions everywhere possible, it is sometimes useful or beneficial to rely on approximations: - pow(x,2.2) and pow(x,1/2.2) - x^2 and sqrt(x) The latter is particularly useful in fragment shaders (for instance to apply dithering in sRGB space), especially if the sqrt() can be replaced with an inversesqrt(). Here is a fairly exhaustive list of modifications implemented in this CL: - Set TARGET_ENABLE_LINEAR_BLENDING := false in BoardConfig.mk to disable linear blending. This is only for GLES 2.0 GPUs with no hardware sRGB support. This flag is currently assumed to be false (see note above) - sRGB writes are disabled when entering a functor (WebView). This will need to be fixed at some point - Skia bitmaps are created with the sRGB color space - Bitmaps using a 565 config are expanded to 888 - Linear blending is disabled when entering a functor - External textures are not properly sampled (see below) - Gradients are interpolated in linear space - Texture-based dithering was replaced with analytical dithering - Dithering is done in the quantization color space, which is why we must do EOCF(OECF(color)+dither) - Text is now gamma corrected differently depending on the luminance of the source pixel. The asumption is that a bright pixel will be blended on a dark background and the other way around. The source alpha is gamma corrected to thicken dark on bright and thin bright on dark to match the intended design of fonts. This also matches the behavior of popular design/drawing applications - Removed the asset atlas. It did not contain anything useful and could not be sampled in sRGB without a yet-to-be-defined GL extension - The last column of color matrices is converted to linear space because its value are added to linear colors Missing features: - Resource qualifier? - Regeneration of goldeng images for automated tests - Handle alpha8/grey8 properly - Disable sRGB write for layers with external textures Test: Manual testing while work in progress Bug: 29940137 Change-Id: I6a07b15ab49b554377cd33a36b6d9971a15e9a0b
2016-09-28 17:34:42 -07:00
}
if (description.hasBitmap) {
if (description.isShaderBitmapExternal) {
shader.append(gFS_Uniforms_BitmapExternalSampler);
} else {
shader.append(gFS_Uniforms_BitmapSampler);
}
}
shader.append(gFS_Uniforms_ColorOp[static_cast<int>(description.colorOp)]);
if (description.hasColorSpaceConversion) {
shader.append(gFS_Uniforms_ColorSpaceConversion);
}
shader.append(gFS_Uniforms_TransferFunction[static_cast<int>(description.transferFunction)]);
// Generate required functions
if (description.hasGradient && description.hasBitmap) {
generateBlend(shader, "blendShaders", description.shadersMode);
}
if (description.colorOp == ProgramDescription::ColorFilterMode::Blend) {
generateBlend(shader, "blendColors", description.colorMode);
}
if (blendFramebuffer) {
generateBlend(shader, "blendFramebuffer", description.framebufferMode);
}
if (description.useShaderBasedWrap) {
generateTextureWrap(shader, description.bitmapWrapS, description.bitmapWrapT);
}
if (description.hasGradient || description.hasLinearTexture ||
description.hasColorSpaceConversion) {
shader.append(gFS_sRGB_TransferFunctions);
}
if (description.hasBitmap || ((description.hasTexture || description.hasExternalTexture) &&
!description.hasAlpha8Texture)) {
shader.append(gFS_TransferFunction[static_cast<int>(description.transferFunction)]);
shader.append(
gFS_OETF[(description.hasLinearTexture || description.hasColorSpaceConversion) &&
!mHasLinearBlending]);
shader.append(gFS_ColorConvert[description.hasColorSpaceConversion
? 1 + description.hasTranslucentConversion
: 0]);
}
Linear blending, step 1 NOTE: Linear blending is currently disabled in this CL as the feature is still a work in progress Android currently performs all blending (any kind of linear math on colors really) on gamma-encoded colors. Since Android assumes that the default color space is sRGB, all bitmaps and colors are encoded with the sRGB Opto-Electronic Conversion Function (OECF, which can be approximated with a power function). Since the power curve is not linear, our linear math is incorrect. The result is that we generate colors that tend to be too dark; this affects blending but also anti-aliasing, gradients, blurs, etc. The solution is to convert gamma-encoded colors back to linear space before doing any math on them, using the sRGB Electo-Optical Conversion Function (EOCF). This is achieved in different ways in different parts of the pipeline: - Using hardware conversions when sampling from OpenGL textures or writing into OpenGL frame buffers - Using software conversion functions, to translate app-supplied colors to and from sRGB - Using Skia's color spaces Any type of processing on colors must roughly ollow these steps: [sRGB input]->EOCF->[linear data]->[processing]->OECF->[sRGB output] For the sRGB color space, the conversion functions are defined as follows: OECF(linear) := linear <= 0.0031308 ? linear * 12.92 : (pow(linear, 1/2.4) * 1.055) - 0.055 EOCF(srgb) := srgb <= 0.04045 ? srgb / 12.92 : pow((srgb + 0.055) / 1.055, 2.4) The EOCF is simply the reciprocal of the OECF. While it is highly recommended to use the exact sRGB conversion functions everywhere possible, it is sometimes useful or beneficial to rely on approximations: - pow(x,2.2) and pow(x,1/2.2) - x^2 and sqrt(x) The latter is particularly useful in fragment shaders (for instance to apply dithering in sRGB space), especially if the sqrt() can be replaced with an inversesqrt(). Here is a fairly exhaustive list of modifications implemented in this CL: - Set TARGET_ENABLE_LINEAR_BLENDING := false in BoardConfig.mk to disable linear blending. This is only for GLES 2.0 GPUs with no hardware sRGB support. This flag is currently assumed to be false (see note above) - sRGB writes are disabled when entering a functor (WebView). This will need to be fixed at some point - Skia bitmaps are created with the sRGB color space - Bitmaps using a 565 config are expanded to 888 - Linear blending is disabled when entering a functor - External textures are not properly sampled (see below) - Gradients are interpolated in linear space - Texture-based dithering was replaced with analytical dithering - Dithering is done in the quantization color space, which is why we must do EOCF(OECF(color)+dither) - Text is now gamma corrected differently depending on the luminance of the source pixel. The asumption is that a bright pixel will be blended on a dark background and the other way around. The source alpha is gamma corrected to thicken dark on bright and thin bright on dark to match the intended design of fonts. This also matches the behavior of popular design/drawing applications - Removed the asset atlas. It did not contain anything useful and could not be sampled in sRGB without a yet-to-be-defined GL extension - The last column of color matrices is converted to linear space because its value are added to linear colors Missing features: - Resource qualifier? - Regeneration of goldeng images for automated tests - Handle alpha8/grey8 properly - Disable sRGB write for layers with external textures Test: Manual testing while work in progress Bug: 29940137 Change-Id: I6a07b15ab49b554377cd33a36b6d9971a15e9a0b
2016-09-28 17:34:42 -07:00
if (description.hasGradient) {
shader.append(gFS_GradientFunctions);
shader.append(gFS_GradientPreamble[mHasLinearBlending]);
Linear blending, step 1 NOTE: Linear blending is currently disabled in this CL as the feature is still a work in progress Android currently performs all blending (any kind of linear math on colors really) on gamma-encoded colors. Since Android assumes that the default color space is sRGB, all bitmaps and colors are encoded with the sRGB Opto-Electronic Conversion Function (OECF, which can be approximated with a power function). Since the power curve is not linear, our linear math is incorrect. The result is that we generate colors that tend to be too dark; this affects blending but also anti-aliasing, gradients, blurs, etc. The solution is to convert gamma-encoded colors back to linear space before doing any math on them, using the sRGB Electo-Optical Conversion Function (EOCF). This is achieved in different ways in different parts of the pipeline: - Using hardware conversions when sampling from OpenGL textures or writing into OpenGL frame buffers - Using software conversion functions, to translate app-supplied colors to and from sRGB - Using Skia's color spaces Any type of processing on colors must roughly ollow these steps: [sRGB input]->EOCF->[linear data]->[processing]->OECF->[sRGB output] For the sRGB color space, the conversion functions are defined as follows: OECF(linear) := linear <= 0.0031308 ? linear * 12.92 : (pow(linear, 1/2.4) * 1.055) - 0.055 EOCF(srgb) := srgb <= 0.04045 ? srgb / 12.92 : pow((srgb + 0.055) / 1.055, 2.4) The EOCF is simply the reciprocal of the OECF. While it is highly recommended to use the exact sRGB conversion functions everywhere possible, it is sometimes useful or beneficial to rely on approximations: - pow(x,2.2) and pow(x,1/2.2) - x^2 and sqrt(x) The latter is particularly useful in fragment shaders (for instance to apply dithering in sRGB space), especially if the sqrt() can be replaced with an inversesqrt(). Here is a fairly exhaustive list of modifications implemented in this CL: - Set TARGET_ENABLE_LINEAR_BLENDING := false in BoardConfig.mk to disable linear blending. This is only for GLES 2.0 GPUs with no hardware sRGB support. This flag is currently assumed to be false (see note above) - sRGB writes are disabled when entering a functor (WebView). This will need to be fixed at some point - Skia bitmaps are created with the sRGB color space - Bitmaps using a 565 config are expanded to 888 - Linear blending is disabled when entering a functor - External textures are not properly sampled (see below) - Gradients are interpolated in linear space - Texture-based dithering was replaced with analytical dithering - Dithering is done in the quantization color space, which is why we must do EOCF(OECF(color)+dither) - Text is now gamma corrected differently depending on the luminance of the source pixel. The asumption is that a bright pixel will be blended on a dark background and the other way around. The source alpha is gamma corrected to thicken dark on bright and thin bright on dark to match the intended design of fonts. This also matches the behavior of popular design/drawing applications - Removed the asset atlas. It did not contain anything useful and could not be sampled in sRGB without a yet-to-be-defined GL extension - The last column of color matrices is converted to linear space because its value are added to linear colors Missing features: - Resource qualifier? - Regeneration of goldeng images for automated tests - Handle alpha8/grey8 properly - Disable sRGB write for layers with external textures Test: Manual testing while work in progress Bug: 29940137 Change-Id: I6a07b15ab49b554377cd33a36b6d9971a15e9a0b
2016-09-28 17:34:42 -07:00
}
// Begin the shader
shader.append(gFS_Main);
{
// Stores the result in fragColor directly
if (description.hasTexture || description.hasExternalTexture) {
if (description.hasAlpha8Texture) {
if (!description.hasGradient && !description.hasBitmap) {
shader.append(gFS_Main_FetchA8Texture[modulateOp * 2 +
description.hasGammaCorrection]);
}
} else {
shader.append(gFS_Main_FetchTexture[modulateOp]);
}
} else {
if (!description.hasGradient && !description.hasBitmap) {
shader.append(gFS_Main_FetchColor);
}
}
if (description.hasGradient) {
shader.append(gFS_Main_FetchGradient[gradientIndex(description)]);
}
if (description.hasBitmap) {
if (!description.useShaderBasedWrap) {
shader.append(gFS_Main_FetchBitmap);
} else {
shader.append(gFS_Main_FetchBitmapNpot);
}
}
bool applyModulate = false;
// Case when we have two shaders set
if (description.hasGradient && description.hasBitmap) {
if (description.isBitmapFirst) {
shader.append(gFS_Main_BlendShadersBG);
} else {
shader.append(gFS_Main_BlendShadersGB);
}
applyModulate =
shaderOp(description, shader, modulateOp, gFS_Main_BlendShaders_Modulate);
} else {
if (description.hasGradient) {
applyModulate =
shaderOp(description, shader, modulateOp, gFS_Main_GradientShader_Modulate);
} else if (description.hasBitmap) {
applyModulate =
shaderOp(description, shader, modulateOp, gFS_Main_BitmapShader_Modulate);
}
}
if (description.modulate && applyModulate) {
shader.append(gFS_Main_ModulateColor);
}
// Apply the color op if needed
shader.append(gFS_Main_ApplyColorOp[static_cast<int>(description.colorOp)]);
if (description.hasVertexAlpha) {
if (description.useShadowAlphaInterp) {
shader.append(gFS_Main_ApplyVertexAlphaShadowInterp);
} else {
shader.append(gFS_Main_ApplyVertexAlphaLinearInterp);
}
}
Linear blending, step 1 NOTE: Linear blending is currently disabled in this CL as the feature is still a work in progress Android currently performs all blending (any kind of linear math on colors really) on gamma-encoded colors. Since Android assumes that the default color space is sRGB, all bitmaps and colors are encoded with the sRGB Opto-Electronic Conversion Function (OECF, which can be approximated with a power function). Since the power curve is not linear, our linear math is incorrect. The result is that we generate colors that tend to be too dark; this affects blending but also anti-aliasing, gradients, blurs, etc. The solution is to convert gamma-encoded colors back to linear space before doing any math on them, using the sRGB Electo-Optical Conversion Function (EOCF). This is achieved in different ways in different parts of the pipeline: - Using hardware conversions when sampling from OpenGL textures or writing into OpenGL frame buffers - Using software conversion functions, to translate app-supplied colors to and from sRGB - Using Skia's color spaces Any type of processing on colors must roughly ollow these steps: [sRGB input]->EOCF->[linear data]->[processing]->OECF->[sRGB output] For the sRGB color space, the conversion functions are defined as follows: OECF(linear) := linear <= 0.0031308 ? linear * 12.92 : (pow(linear, 1/2.4) * 1.055) - 0.055 EOCF(srgb) := srgb <= 0.04045 ? srgb / 12.92 : pow((srgb + 0.055) / 1.055, 2.4) The EOCF is simply the reciprocal of the OECF. While it is highly recommended to use the exact sRGB conversion functions everywhere possible, it is sometimes useful or beneficial to rely on approximations: - pow(x,2.2) and pow(x,1/2.2) - x^2 and sqrt(x) The latter is particularly useful in fragment shaders (for instance to apply dithering in sRGB space), especially if the sqrt() can be replaced with an inversesqrt(). Here is a fairly exhaustive list of modifications implemented in this CL: - Set TARGET_ENABLE_LINEAR_BLENDING := false in BoardConfig.mk to disable linear blending. This is only for GLES 2.0 GPUs with no hardware sRGB support. This flag is currently assumed to be false (see note above) - sRGB writes are disabled when entering a functor (WebView). This will need to be fixed at some point - Skia bitmaps are created with the sRGB color space - Bitmaps using a 565 config are expanded to 888 - Linear blending is disabled when entering a functor - External textures are not properly sampled (see below) - Gradients are interpolated in linear space - Texture-based dithering was replaced with analytical dithering - Dithering is done in the quantization color space, which is why we must do EOCF(OECF(color)+dither) - Text is now gamma corrected differently depending on the luminance of the source pixel. The asumption is that a bright pixel will be blended on a dark background and the other way around. The source alpha is gamma corrected to thicken dark on bright and thin bright on dark to match the intended design of fonts. This also matches the behavior of popular design/drawing applications - Removed the asset atlas. It did not contain anything useful and could not be sampled in sRGB without a yet-to-be-defined GL extension - The last column of color matrices is converted to linear space because its value are added to linear colors Missing features: - Resource qualifier? - Regeneration of goldeng images for automated tests - Handle alpha8/grey8 properly - Disable sRGB write for layers with external textures Test: Manual testing while work in progress Bug: 29940137 Change-Id: I6a07b15ab49b554377cd33a36b6d9971a15e9a0b
2016-09-28 17:34:42 -07:00
if (description.hasGradient) {
shader.append(gFS_Main_AddDither);
}
// Output the fragment
if (!blendFramebuffer) {
shader.append(gFS_Main_FragColor);
} else {
shader.append(!description.swapSrcDst ? gFS_Main_FragColor_Blend
: gFS_Main_FragColor_Blend_Swap);
}
if (description.hasColors) {
shader.append(gFS_Main_FragColor_HasColors);
}
if (description.hasRoundRectClip) {
shader.append(gFS_Main_FragColor_HasRoundRectClip);
}
if (description.hasDebugHighlight) {
shader.append(gFS_Main_DebugHighlight);
}
}
// End the shader
shader.append(gFS_Footer);
#if DEBUG_PROGRAMS
PROGRAM_LOGD("*** Generated fragment shader:\n\n");
printLongString(shader);
#endif
return shader;
}
void ProgramCache::generateBlend(String8& shader, const char* name, SkBlendMode mode) {
shader.append("\nvec4 ");
shader.append(name);
shader.append("(vec4 src, vec4 dst) {\n");
shader.append(" ");
shader.append(gBlendOps[(int)mode]);
shader.append("}\n");
}
void ProgramCache::generateTextureWrap(String8& shader, GLenum wrapS, GLenum wrapT) {
shader.append("\nhighp vec2 wrap(highp vec2 texCoords) {\n");
if (wrapS == GL_MIRRORED_REPEAT) {
shader.append(" highp float xMod2 = mod(texCoords.x, 2.0);\n");
shader.append(" if (xMod2 > 1.0) xMod2 = 2.0 - xMod2;\n");
}
if (wrapT == GL_MIRRORED_REPEAT) {
shader.append(" highp float yMod2 = mod(texCoords.y, 2.0);\n");
shader.append(" if (yMod2 > 1.0) yMod2 = 2.0 - yMod2;\n");
}
shader.append(" return vec2(");
switch (wrapS) {
case GL_CLAMP_TO_EDGE:
shader.append("texCoords.x");
break;
case GL_REPEAT:
shader.append("mod(texCoords.x, 1.0)");
break;
case GL_MIRRORED_REPEAT:
shader.append("xMod2");
break;
}
shader.append(", ");
switch (wrapT) {
case GL_CLAMP_TO_EDGE:
shader.append("texCoords.y");
break;
case GL_REPEAT:
shader.append("mod(texCoords.y, 1.0)");
break;
case GL_MIRRORED_REPEAT:
shader.append("yMod2");
break;
}
shader.append(");\n");
shader.append("}\n");
}
void ProgramCache::printLongString(const String8& shader) const {
ssize_t index = 0;
ssize_t lastIndex = 0;
const char* str = shader.string();
while ((index = shader.find("\n", index)) > -1) {
String8 line(str, index - lastIndex);
if (line.length() == 0) line.append("\n");
ALOGD("%s", line.string());
index++;
str += (index - lastIndex);
lastIndex = index;
}
}
}; // namespace uirenderer
}; // namespace android