Cleanups and depth

- Centralized wavelet code and roughly document
- Add a full screen pass to render the depth at which transmittance
  falls to zero

Author: Jozufozu

common/src/backend/java/dev/engine_room/flywheel/backend/compile/IndirectPrograms.java

@@ -34,6 +34,7 @@ public class IndirectPrograms extends AtomicReferenceCounted {
 
 	private static final ResourceLocation FULLSCREEN = Flywheel.rl("internal/indirect/fullscreen.vert");
 	private static final ResourceLocation OIT_COMPOSITE = Flywheel.rl("internal/indirect/oit_composite.frag");
+	private static final ResourceLocation OIT_DEPTH = Flywheel.rl("internal/indirect/oit_depth.frag");
 
 	private static final Compile<InstanceType<?>> CULL = new Compile<>();
 	private static final Compile<ResourceLocation> UTIL = new Compile<>();
@@ -138,6 +139,7 @@ private static CompilationHarness<ResourceLocation> createFullscreenCompiler(Sha
 				.link(UTIL.shader(GlCompat.MAX_GLSL_VERSION, ShaderType.FRAGMENT)
 						.nameMapper(rl -> "fullscreen/" + ResourceUtil.toDebugFileNameNoExtension(rl))
 						.withResource(s -> s))
+				.postLink((key, program) -> Uniforms.setUniformBlockBindings(program))
 				.harness("fullscreen", sources);
 	}
 
@@ -192,6 +194,10 @@ public GlProgram getOitCompositeProgram() {
 		return fullscreen.get(OIT_COMPOSITE);
 	}
 
+	public GlProgram getOitDepthProgram() {
+		return fullscreen.get(OIT_DEPTH);
+	}
+
 	@Override
 	protected void _delete() {
 		pipeline.delete();

common/src/backend/java/dev/engine_room/flywheel/backend/engine/indirect/IndirectDrawManager.java

@@ -158,9 +158,9 @@ public void render(LightStorage lightStorage, EnvironmentStorage environmentStor
 			group.submitTransparent(PipelineCompiler.OitMode.GENERATE_COEFFICIENTS);
 		}
 
-		//		wboitFrameBuffer.adjustBackgroundForTotalTransmittance();
+		wboitFrameBuffer.renderDepth();
 
-		//		vertexArray.bindForDraw();
+		vertexArray.bindForDraw();
 
 		wboitFrameBuffer.shade();
 

common/src/backend/java/dev/engine_room/flywheel/backend/engine/indirect/OitFramebuffer.java

@@ -14,13 +14,13 @@
 
 public class OitFramebuffer {
 
-	public final int fbo;
 	private final IndirectPrograms programs;
 	private final int vao;
 
-	public int depthBounds;
-	public int coefficients;
-	public int accumulate;
+	public int fbo = -1;
+	public int depthBounds = -1;
+	public int coefficients = -1;
+	public int accumulate = -1;
 
 	private int lastWidth = -1;
 	private int lastHeight = -1;
@@ -39,6 +39,7 @@ public void depthRange() {
 
 		// No depth writes, but we'll still use the depth test
 		RenderSystem.depthMask(false);
+		RenderSystem.colorMask(true, true, true, true);
 		RenderSystem.enableBlend();
 		RenderSystem.blendFunc(GlStateManager.SourceFactor.ONE, GlStateManager.DestFactor.ONE);
 		RenderSystem.blendEquation(GL46.GL_MAX);
@@ -57,6 +58,7 @@ public void depthRange() {
 	public void renderTransmittance() {
 		// No depth writes, but we'll still use the depth test
 		RenderSystem.depthMask(false);
+		RenderSystem.colorMask(true, true, true, true);
 		RenderSystem.enableBlend();
 		RenderSystem.blendFunc(GlStateManager.SourceFactor.ONE, GlStateManager.DestFactor.ONE);
 		RenderSystem.blendEquation(GL46.GL_FUNC_ADD);
@@ -85,6 +87,7 @@ public void renderTransmittance() {
 	public void shade() {
 		// No depth writes, but we'll still use the depth test
 		RenderSystem.depthMask(false);
+		RenderSystem.colorMask(true, true, true, true);
 		RenderSystem.enableBlend();
 		RenderSystem.blendFunc(GlStateManager.SourceFactor.ONE, GlStateManager.DestFactor.ONE);
 		RenderSystem.blendEquation(GL46.GL_FUNC_ADD);
@@ -106,9 +109,34 @@ public void shade() {
 		GlStateManager._glBindFramebuffer(GL46.GL_FRAMEBUFFER, fbo);
 	}
 
+	public void renderDepth() {
+		// No depth writes, but we'll still use the depth test
+		RenderSystem.depthMask(true);
+		RenderSystem.colorMask(false, false, false, false);
+		RenderSystem.disableBlend();
+
+		Samplers.COEFFICIENTS.makeActive();
+		GlStateManager._bindTexture(0);
+		GL46.glBindTextureUnit(0, coefficients);
+
+		Samplers.DEPTH_RANGE.makeActive();
+		GlStateManager._bindTexture(depthBounds);
+
+		GL46.glNamedFramebufferDrawBuffers(fbo, new int[]{});
+
+		programs.getOitDepthProgram()
+				.bind();
+
+		// Empty VAO, the actual full screen triangle is generated in the vertex shader
+		GlStateManager._glBindVertexArray(vao);
+
+		GL46.glDrawArrays(GL46.GL_TRIANGLES, 0, 3);
+	}
+
 	public void composite() {
 		// No depth writes, but we'll still use the depth test
 		RenderSystem.depthMask(false);
+		RenderSystem.colorMask(true, true, true, true);
 		RenderSystem.enableBlend();
 		RenderSystem.blendFunc(GlStateManager.SourceFactor.ONE_MINUS_SRC_ALPHA, GlStateManager.DestFactor.SRC_ALPHA);
 		RenderSystem.blendEquation(GL46.GL_FUNC_ADD);
@@ -141,9 +169,15 @@ public void delete() {
 	}
 
 	private void deleteTextures() {
-		GL46.glDeleteTextures(depthBounds);
-		GL46.glDeleteTextures(coefficients);
-		GL46.glDeleteTextures(accumulate);
+		if (depthBounds != -1) {
+			GL46.glDeleteTextures(depthBounds);
+		}
+		if (coefficients != -1) {
+			GL46.glDeleteTextures(coefficients);
+		}
+		if (accumulate != -1) {
+			GL46.glDeleteTextures(accumulate);
+		}
 	}
 
 	private void createTextures(int width, int height) {

common/src/backend/resources/assets/flywheel/flywheel/internal/common.frag

@@ -1,6 +1,8 @@
 #include "flywheel:internal/packed_material.glsl"
 #include "flywheel:internal/diffuse.glsl"
 #include "flywheel:internal/colorizer.glsl"
+#include "flywheel:internal/wavelet.glsl"
+#include "flywheel:internal/depth.glsl"
 
 // optimize discard usage
 #if defined(GL_ARB_conservative_depth) && defined(_FLW_USE_DISCARD)
@@ -19,10 +21,6 @@ flat in uvec2 _flw_ids;
 
 #ifdef _FLW_OIT
 
-#define TRANSPARENCY_WAVELET_RANK 3
-#define TRANSPARENCY_WAVELET_COEFFICIENT_COUNT 16
-#define REMOVE_SIGNAL true
-
 layout (binding = 7) uniform sampler2D _flw_depthRange;
 
 layout (binding = 8) uniform sampler2DArray _flw_coefficients;
@@ -44,11 +42,6 @@ float tented_blue_noise(float normalizedDepth) {
     return b * tent;
 }
 
-float linearize_depth(float d, float zNear, float zFar) {
-    float z_n = 2.0 * d - 1.0;
-    return 2.0 * zNear * zFar / (zFar + zNear - z_n * (zFar - zNear));
-}
-
 float linear_depth() {
     return linearize_depth(gl_FragCoord.z, _flw_cullData.znear, _flw_cullData.zfar);
 }
@@ -69,126 +62,19 @@ layout (location = 0) out vec2 _flw_depthRange_out;
 
 #endif
 
-
 #ifdef _FLW_COLLECT_COEFFS
 
-
 layout (location = 0) out vec4 _flw_coeffs0;
 layout (location = 1) out vec4 _flw_coeffs1;
 layout (location = 2) out vec4 _flw_coeffs2;
 layout (location = 3) out vec4 _flw_coeffs3;
 
-void add_to_index(inout vec4[4] coefficients, uint index, float addend) {
-    coefficients[index >> 2][index & 3u] = addend;
-}
-
-void add_event_to_wavelets(inout vec4[4] coefficients, float signal, float depth)
-{
-    depth *= float(TRANSPARENCY_WAVELET_COEFFICIENT_COUNT-1) / TRANSPARENCY_WAVELET_COEFFICIENT_COUNT;
-
-    int index = clamp(int(floor(depth * TRANSPARENCY_WAVELET_COEFFICIENT_COUNT)), 0, TRANSPARENCY_WAVELET_COEFFICIENT_COUNT - 1);
-    index += TRANSPARENCY_WAVELET_COEFFICIENT_COUNT - 1;
-
-    for (int i = 0; i < (TRANSPARENCY_WAVELET_RANK+1); ++i)
-    {
-        int power = TRANSPARENCY_WAVELET_RANK - i;
-        int new_index = (index - 1) >> 1;
-        float k = float((new_index + 1) & ((1 << power) - 1));
-
-        int wavelet_sign = ((index & 1) << 1) - 1;
-        float wavelet_phase = ((index + 1) & 1) * exp2(-power);
-        float addend = fma(fma(-exp2(-power), k, depth), wavelet_sign, wavelet_phase) * exp2(power * 0.5) * signal;
-        add_to_index(coefficients, new_index, addend);
-
-        index = new_index;
-    }
-
-    float addend = fma(signal, -depth, signal);
-    add_to_index(coefficients, TRANSPARENCY_WAVELET_COEFFICIENT_COUNT - 1, addend);
-}
-
-void add_transmittance_event_to_wavelets(inout vec4[4] coefficients, float transmittance, float depth)
-{
-    float absorbance = -log(max(transmittance, 0.00001));// transforming the signal from multiplicative transmittance to additive absorbance
-    add_event_to_wavelets(coefficients, absorbance, depth);
-}
-
 #endif
 
 #ifdef _FLW_EVALUATE
 
 layout (location = 0) out vec4 _flw_accumulate;
 
-
-float get_coefficients(in sampler2DArray coefficients, uint index) {
-    return texelFetch(coefficients, ivec3(gl_FragCoord.xy, index >> 2), 0)[index & 3u];
-}
-
-float evaluate_wavelets(in sampler2DArray coefficients, float depth, float signal)
-{
-    float scale_coefficient = get_coefficients(coefficients, TRANSPARENCY_WAVELET_COEFFICIENT_COUNT - 1);
-    if (scale_coefficient == 0)
-    {
-        return 0;
-    }
-
-    depth *= float(TRANSPARENCY_WAVELET_COEFFICIENT_COUNT-1) / TRANSPARENCY_WAVELET_COEFFICIENT_COUNT;
-
-    if (REMOVE_SIGNAL)
-    {
-        float scale_coefficient_addend = fma(signal, -depth, signal);
-        scale_coefficient -= scale_coefficient_addend;
-    }
-
-    float coefficient_depth = depth * TRANSPARENCY_WAVELET_COEFFICIENT_COUNT;
-    int index_b = clamp(int(floor(coefficient_depth)), 0, TRANSPARENCY_WAVELET_COEFFICIENT_COUNT - 1);
-    bool sample_a = index_b >= 1;
-    int index_a = sample_a ? (index_b - 1) : index_b;
-
-    index_b += TRANSPARENCY_WAVELET_COEFFICIENT_COUNT - 1;
-    index_a += TRANSPARENCY_WAVELET_COEFFICIENT_COUNT - 1;
-
-    float b = scale_coefficient;
-    float a = sample_a ? scale_coefficient : 0;
-
-    for (int i = 0; i < (TRANSPARENCY_WAVELET_RANK+1); ++i)
-    {
-        int power = TRANSPARENCY_WAVELET_RANK - i;
-
-        int new_index_b = (index_b - 1) >> 1;
-        int wavelet_sign_b = ((index_b & 1) << 1) - 1;
-        float coeff_b = get_coefficients(coefficients, new_index_b);
-        if (REMOVE_SIGNAL)
-        {
-            float wavelet_phase_b = ((index_b + 1) & 1) * exp2(-power);
-            float k = float((new_index_b + 1) & ((1 << power) - 1));
-            float addend = fma(fma(-exp2(-power), k, depth), wavelet_sign_b, wavelet_phase_b) * exp2(power * 0.5) * signal;
-            coeff_b -= addend;
-        }
-        b -= exp2(float(power) * 0.5) * coeff_b * wavelet_sign_b;
-        index_b = new_index_b;
-
-        if (sample_a)
-        {
-            int new_index_a = (index_a - 1) >> 1;
-            int wavelet_sign_a = ((index_a & 1) << 1) - 1;
-            float coeff_a = (new_index_a == new_index_b) ? coeff_b : get_coefficients(coefficients, new_index_a);// No addend here on purpose, the original signal didn't contribute to this coefficient
-            a -= exp2(float(power) * 0.5) * coeff_a * wavelet_sign_a;
-            index_a = new_index_a;
-        }
-    }
-
-    float t = coefficient_depth >= TRANSPARENCY_WAVELET_COEFFICIENT_COUNT ? 1.0 : fract(coefficient_depth);
-
-    return mix(a, b, t);
-}
-
-float evaluate_transmittance_wavelets(in sampler2DArray coefficients, float depth, float signal)
-{
-    float absorbance = evaluate_wavelets(coefficients, depth, signal);
-    return clamp(exp(-absorbance), 0., 1.);// undoing the transformation from absorbance back to transmittance
-}
-
 #endif
 
 #else
@@ -298,7 +184,7 @@ void _flw_main() {
     result[2] = vec4(0.);
     result[3] = vec4(0.);
 
-    add_transmittance_event_to_wavelets(result, 1. - color.a, depth());
+    add_transmittance(result, 1. - color.a, depth());
 
     _flw_coeffs0 = result[0];
     _flw_coeffs1 = result[1];
@@ -309,7 +195,7 @@ void _flw_main() {
 
     #ifdef _FLW_EVALUATE
 
-    float transmittance = evaluate_transmittance_wavelets(_flw_coefficients, depth(), 1. - color.a);
+    float transmittance = signal_corrected_transmittance(_flw_coefficients, depth(), 1. - color.a);
 
     _flw_accumulate = vec4(color.rgb * color.a, color.a) * transmittance;
 

common/src/backend/resources/assets/flywheel/flywheel/internal/depth.glsl

@@ -0,0 +1,9 @@
+float linearize_depth(float d, float zNear, float zFar) {
+    float z_n = 2.0 * d - 1.0;
+    return 2.0 * zNear * zFar / (zFar + zNear - z_n * (zFar - zNear));
+}
+
+float delinearize_depth(float linearDepth, float zNear, float zFar) {
+    float z_n = (2.0 * zNear * zFar / linearDepth) - (zFar + zNear);
+    return 0.5 * (z_n / (zNear - zFar) + 1.0);
+}