gpu: draw full screen widgets on gpu

This adjusts the draw callbacks api to take in the color of the full
screen overlay the client would have rendered. This is primarily used in
godwars, underwater, darkmeyer, etc. Having them rasterized on the cpu
is slow, especially with gpu on since we additionally have to compute
the alpha per-pixel.

runelite-api/src/main/java/net/runelite/api/hooks/DrawCallbacks.java

@@ -43,7 +43,12 @@ public interface DrawCallbacks
 						SceneTileModel model, int tileZ, int tileX, int tileY,
 						int zoom, int centerX, int centerY);
 
-	void draw();
+	/**
+	 * Called when a frame should be drawn.
+	 *
+	 * @param overlayColor Color of full-viewport overlays, if any
+	 */
+	void draw(int overlayColor);
 
 	boolean drawFace(Model model, int face);
 

runelite-client/src/main/java/net/runelite/client/plugins/gpu/GpuPlugin.java

@@ -264,6 +264,7 @@ public class GpuPlugin extends Plugin implements DrawCallbacks
 	private int uniTexSamplingMode;
 	private int uniTexSourceDimensions;
 	private int uniTexTargetDimensions;
+	private int uniUiAlphaOverlay;
 	private int uniTextures;
 	private int uniTextureOffsets;
 	private int uniBlockSmall;
@@ -545,6 +546,7 @@ public class GpuPlugin extends Plugin implements DrawCallbacks
 		uniTexTargetDimensions = gl.glGetUniformLocation(glUiProgram, "targetDimensions");
 		uniTexSourceDimensions = gl.glGetUniformLocation(glUiProgram, "sourceDimensions");
 		uniUiColorBlindMode = gl.glGetUniformLocation(glUiProgram, "colorBlindMode");
+		uniUiAlphaOverlay = gl.glGetUniformLocation(glUiProgram, "alphaOverlay");
 		uniTextures = gl.glGetUniformLocation(glProgram, "textures");
 		uniTextureOffsets = gl.glGetUniformLocation(glProgram, "textureOffsets");
 
@@ -855,9 +857,9 @@ public class GpuPlugin extends Plugin implements DrawCallbacks
 	}
 
 	@Override
-	public void draw()
+	public void draw(int overlayColor)
 	{
-		invokeOnMainThread(this::drawFrame);
+		invokeOnMainThread(() -> drawFrame(overlayColor));
 	}
 
 	private void resize(int canvasWidth, int canvasHeight, int viewportWidth, int viewportHeight)
@@ -883,7 +885,7 @@ public class GpuPlugin extends Plugin implements DrawCallbacks
 		}
 	}
 
-	private void drawFrame()
+	private void drawFrame(int overlayColor)
 	{
 		if (jawtWindow.getAWTComponent() != client.getCanvas())
 		{
@@ -1208,7 +1210,7 @@ public class GpuPlugin extends Plugin implements DrawCallbacks
 		tempUvOffset = 0;
 
 		// Texture on UI
-		drawUi(canvasHeight, canvasWidth);
+		drawUi(overlayColor, canvasHeight, canvasWidth);
 
 		glDrawable.swapBuffers();
 
@@ -1226,7 +1228,7 @@ public class GpuPlugin extends Plugin implements DrawCallbacks
 		};
 	}
 
-	private void drawUi(final int canvasHeight, final int canvasWidth)
+	private void drawUi(final int overlayColor, final int canvasHeight, final int canvasWidth)
 	{
 		final BufferProvider bufferProvider = client.getBufferProvider();
 		final int[] pixels = bufferProvider.getPixels();
@@ -1254,6 +1256,12 @@ public class GpuPlugin extends Plugin implements DrawCallbacks
 		gl.glUniform1i(uniTexSamplingMode, uiScalingMode.getMode());
 		gl.glUniform2i(uniTexSourceDimensions, canvasWidth, canvasHeight);
 		gl.glUniform1i(uniUiColorBlindMode, config.colorBlindMode().ordinal());
+		gl.glUniform4f(uniUiAlphaOverlay,
+			(overlayColor >> 16 & 0xFF) / 255f,
+			(overlayColor >> 8 & 0xFF) / 255f,
+			(overlayColor & 0xFF) / 255f,
+			(overlayColor >>> 24) / 255f
+		);
 
 		if (client.isStretchedEnabled())
 		{

runelite-client/src/main/resources/net/runelite/client/plugins/gpu/fragui.glsl

@@ -34,6 +34,7 @@ uniform int samplingMode;
 uniform ivec2 sourceDimensions;
 uniform ivec2 targetDimensions;
 uniform int colorBlindMode;
+uniform vec4 alphaOverlay;
 
 #include scale/bicubic.glsl
 #include scale/xbr_lv2_frag.glsl
@@ -44,6 +45,13 @@ in XBRTable xbrTable;
 
 out vec4 FragColor;
 
+vec4 alphaBlend(vec4 src, vec4 dst) {
+    return vec4(
+        src.rgb + dst.rgb * (1.0f - src.a),
+        src.a + dst.a * (1.0f - src.a)
+    );
+}
+
 void main() {
     vec4 c;
 
@@ -51,14 +59,17 @@ void main() {
         case SAMPLING_CATROM:
         case SAMPLING_MITCHELL:
             c = textureCubic(tex, TexCoord, samplingMode);
+            c = alphaBlend(c, alphaOverlay);
             c.rgb = colorblind(colorBlindMode, c.rgb);
             break;
         case SAMPLING_XBR:
             c = textureXBR(tex, TexCoord, xbrTable, ceil(1.0 * targetDimensions.x / sourceDimensions.x));
+            c = alphaBlend(c, alphaOverlay);
             c.rgb = colorblind(colorBlindMode, c.rgb);
             break;
         default: // NEAREST or LINEAR, which uses GL_TEXTURE_MIN_FILTER/GL_TEXTURE_MAG_FILTER to affect sampling
             c = texture(tex, TexCoord);
+            c = alphaBlend(c, alphaOverlay);
             c.rgb = colorblind(colorBlindMode, c.rgb);
     }