[ProtonAOSP] blur: Interpolate tap UV coordinates in vertex shader

Instead of calculating the UV coordinate to sample for each blur tap in
the vertex shader, calculate them as varying attributes in the vertex
shader and let the GPU interpolate them for each fragment. This allows
the GPU to prefetch pixel data because we no longer rely on dependent
texture reads.

When tested with 2 layers of another 6-pass blur implementation, this
saves ~100 µs of rendering time on an Adreno 640 GPU at 1440x3040
resolution.

Change-Id: Id2a700decef060ea23d1d5e2cccaea864581bf16

2 files changed, 35 insertions, 11 deletions

diff --git a/libs/renderengine/gl/filters/BlurFilter.cpp b/libs/renderengine/gl/filters/BlurFilter.cpp
index c317c5f7dd..d53fc8b25e 100644
--- a/libs/renderengine/gl/filters/BlurFilter.cpp
+++ b/libs/renderengine/gl/filters/BlurFilter.cpp
@@ -40,7 +40,7 @@ BlurFilter::BlurFilter(GLESRenderEngine& engine)
         mMixProgram(engine),
         mDitherMixProgram(engine),
         mBlurProgram(engine) {
-    mMixProgram.compile(getVertexShader(), getMixFragShader());
+    mMixProgram.compile(getMixVertShader(), getMixFragShader());
     mMPosLoc = mMixProgram.getAttributeLocation("aPosition");
     mMUvLoc = mMixProgram.getAttributeLocation("aUV");
     mMBlurTextureLoc = mMixProgram.getUniformLocation("uBlurTexture");
@@ -59,7 +59,7 @@ BlurFilter::BlurFilter(GLESRenderEngine& engine)
                                GL_NEAREST, GL_REPEAT,
                                GL_RGB8, GL_RGB, GL_UNSIGNED_BYTE);
 
-    mBlurProgram.compile(getVertexShader(), getFragmentShader());
+    mBlurProgram.compile(getBlurVertShader(), getBlurFragShader());
     mBPosLoc = mBlurProgram.getAttributeLocation("aPosition");
     mBUvLoc = mBlurProgram.getAttributeLocation("aUV");
     mBTextureLoc = mBlurProgram.getUniformLocation("uTexture");
@@ -247,43 +247,66 @@ status_t BlurFilter::render(size_t layers, int currentLayer) {
     return NO_ERROR;
 }
 
-string BlurFilter::getVertexShader() const {
+string BlurFilter::getBlurVertShader() const {
     return R"SHADER(#version 310 es
         precision mediump float;
 
+        uniform vec2 uOffset;
+
         in vec2 aPosition;
         in vec2 aUV;
         out vec2 vUV;
+        out vec2 vBlurTaps[4];
 
         void main() {
             vUV = aUV;
             gl_Position = vec4(aPosition, 0.0, 1.0);
+
+            vBlurTaps[0] = vUV + vec2( uOffset.x,  uOffset.y);
+            vBlurTaps[1] = vUV + vec2( uOffset.x, -uOffset.y);
+            vBlurTaps[2] = vUV + vec2(-uOffset.x,  uOffset.y);
+            vBlurTaps[3] = vUV + vec2(-uOffset.x, -uOffset.y);
         }
     )SHADER";
 }
 
-string BlurFilter::getFragmentShader() const {
+string BlurFilter::getBlurFragShader() const {
     return R"SHADER(#version 310 es
         precision mediump float;
 
         uniform sampler2D uTexture;
-        uniform vec2 uOffset;
 
         in vec2 vUV;
+        in vec2 vBlurTaps[4];
         out vec4 fragColor;
 
         void main() {
             vec3 sum = texture(uTexture, vUV).rgb;
-            sum += texture(uTexture, vUV + vec2( uOffset.x,  uOffset.y)).rgb;
-            sum += texture(uTexture, vUV + vec2( uOffset.x, -uOffset.y)).rgb;
-            sum += texture(uTexture, vUV + vec2(-uOffset.x,  uOffset.y)).rgb;
-            sum += texture(uTexture, vUV + vec2(-uOffset.x, -uOffset.y)).rgb;
+            sum += texture(uTexture, vBlurTaps[0]).rgb;
+            sum += texture(uTexture, vBlurTaps[1]).rgb;
+            sum += texture(uTexture, vBlurTaps[2]).rgb;
+            sum += texture(uTexture, vBlurTaps[3]).rgb;
 
             fragColor = vec4(sum * 0.2, 1.0);
         }
     )SHADER";
 }
 
+string BlurFilter::getMixVertShader() const {
+    return R"SHADER(#version 310 es
+        precision mediump float;
+
+        in vec2 aPosition;
+        in vec2 aUV;
+        out vec2 vUV;
+
+        void main() {
+            vUV = aUV;
+            gl_Position = vec4(aPosition, 0.0, 1.0);
+        }
+    )SHADER";
+}
+
 string BlurFilter::getMixFragShader() const {
     string shader = R"SHADER(#version 310 es
         precision mediump float;
diff --git a/libs/renderengine/gl/filters/BlurFilter.h b/libs/renderengine/gl/filters/BlurFilter.h
index 686c74914f..e5dae7b9ea 100644
--- a/libs/renderengine/gl/filters/BlurFilter.h
+++ b/libs/renderengine/gl/filters/BlurFilter.h
@@ -54,8 +54,9 @@ public:
 private:
     uint32_t mRadius;
     void drawMesh(GLuint uv, GLuint position);
-    string getVertexShader() const;
-    string getFragmentShader() const;
+    string getBlurVertShader() const;
+    string getBlurFragShader() const;
+    string getMixVertShader() const;
     string getMixFragShader() const;
     string getDitherMixVertShader() const;
     string getDitherMixFragShader() const;