Ubern't

- De-uberify the light shader
- Remove lightSources index
- Include LightShader in PipelineProgramKey and parameterize the
  pipeline fragment shader by it
- Profiling suggests that specializing the shaders uses significantly
  less GPU time, and we may want to do this for actual user-authored
  material shaders (and cutout?) as well
- Sort LightShader highest in the material comparator
- Implement a materialEquals method so IndirectCullingGroup can bucket
  draws on more that just material reference equality
- Do not store any particular draw program in IndirectCullingGroup

common/src/backend/java/dev/engine_room/flywheel/backend/MaterialShaderIndices.java

@@ -8,7 +8,6 @@ import org.jetbrains.annotations.Unmodifiable;
 
 import dev.engine_room.flywheel.api.material.CutoutShader;
 import dev.engine_room.flywheel.api.material.FogShader;
-import dev.engine_room.flywheel.api.material.LightShader;
 import dev.engine_room.flywheel.api.material.MaterialShaders;
 import dev.engine_room.flywheel.api.registry.Registry;
 import it.unimi.dsi.fastutil.objects.Object2IntMap;
@@ -26,8 +25,6 @@ public final class MaterialShaderIndices {
 	private static Index fogSources;
 	@Nullable
 	private static Index cutoutSources;
-	@Nullable
-	private static Index lightSources;
 
 	private MaterialShaderIndices() {
 	}
@@ -60,13 +57,6 @@ public final class MaterialShaderIndices {
 		return cutoutSources;
 	}
 
-	public static Index lightSources() {
-		if (lightSources == null) {
-			lightSources = indexFromRegistry(LightShader.REGISTRY, LightShader::source);
-		}
-		return lightSources;
-	}
-
 	public static int vertexIndex(MaterialShaders shaders) {
 		return vertexSources().index(shaders.vertexSource());
 	}
@@ -83,10 +73,6 @@ public final class MaterialShaderIndices {
 		return cutoutSources().index(cutoutShader.source());
 	}
 
-	public static int lightIndex(LightShader lightShader) {
-		return lightSources().index(lightShader.source());
-	}
-
 	private static <T> Index indexFromRegistry(Registry<T> registry, Function<T, ResourceLocation> sourceFunc) {
 		if (!registry.isFrozen()) {
 			throw new IllegalStateException("Cannot create index from registry that is not frozen!");

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

@@ -10,6 +10,7 @@ import com.google.common.collect.ImmutableList;
 
 import dev.engine_room.flywheel.api.Flywheel;
 import dev.engine_room.flywheel.api.instance.InstanceType;
+import dev.engine_room.flywheel.api.material.LightShader;
 import dev.engine_room.flywheel.backend.MaterialShaderIndices;
 import dev.engine_room.flywheel.backend.compile.component.UberShaderComponent;
 import dev.engine_room.flywheel.backend.compile.core.CompilerStats;
@@ -46,16 +47,15 @@ public final class FlwPrograms {
 		var fragmentMaterialComponent = createFragmentMaterialComponent(loader);
 		var fogComponent = createFogComponent(loader);
 		var cutoutComponent = createCutoutComponent(loader);
-		var lightComponent = createLightComponent(loader);
 
-		if (stats.errored() || vertexComponentsHeader == null || fragmentComponentsHeader == null || vertexMaterialComponent == null || fragmentMaterialComponent == null || fogComponent == null || cutoutComponent == null || lightComponent == null) {
+		if (stats.errored() || vertexComponentsHeader == null || fragmentComponentsHeader == null || vertexMaterialComponent == null || fragmentMaterialComponent == null || fogComponent == null || cutoutComponent == null) {
 			// Probably means the shader sources are missing.
 			stats.emitErrorLog();
 			return;
 		}
 
 		List<SourceComponent> vertexComponents = List.of(vertexComponentsHeader, vertexMaterialComponent);
-		List<SourceComponent> fragmentComponents = List.of(fragmentComponentsHeader, fragmentMaterialComponent, fogComponent, cutoutComponent, lightComponent);
+		List<SourceComponent> fragmentComponents = List.of(fragmentComponentsHeader, fragmentMaterialComponent, fogComponent, cutoutComponent);
 
 		var pipelineKeys = createPipelineKeys();
 		InstancingPrograms.reload(sources, pipelineKeys, vertexComponents, fragmentComponents);
@@ -66,7 +66,9 @@ public final class FlwPrograms {
 		ImmutableList.Builder<PipelineProgramKey> builder = ImmutableList.builder();
 		for (ContextShader contextShader : ContextShader.values()) {
 			for (InstanceType<?> instanceType : InstanceType.REGISTRY) {
-				builder.add(new PipelineProgramKey(instanceType, contextShader));
+				for (LightShader light : LightShader.REGISTRY.getAll()) {
+					builder.add(new PipelineProgramKey(instanceType, contextShader, light));
+				}
 			}
 		}
 		return builder.build();
@@ -119,18 +121,4 @@ public final class FlwPrograms {
 				.switchOn(GlslExpr.variable("_flw_uberCutoutIndex"))
 				.build(loader);
 	}
-
-	// TODO: Do not uber this component. Shader compile times are very high now
-	@Nullable
-	private static UberShaderComponent createLightComponent(SourceLoader loader) {
-		return UberShaderComponent.builder(Flywheel.rl("light"))
-				.materialSources(MaterialShaderIndices.lightSources()
-						.all())
-				.adapt(FnSignature.create()
-						.returnType("void")
-						.name("flw_shaderLight")
-						.build())
-				.switchOn(GlslExpr.variable("_flw_uberLightIndex"))
-				.build(loader);
-	}
 }

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

@@ -9,6 +9,7 @@ import com.google.common.collect.ImmutableList;
 
 import dev.engine_room.flywheel.api.Flywheel;
 import dev.engine_room.flywheel.api.instance.InstanceType;
+import dev.engine_room.flywheel.api.material.LightShader;
 import dev.engine_room.flywheel.backend.compile.component.InstanceStructComponent;
 import dev.engine_room.flywheel.backend.compile.component.SsboInstanceComponent;
 import dev.engine_room.flywheel.backend.compile.core.CompilationHarness;
@@ -167,8 +168,8 @@ public class IndirectPrograms extends AtomicReferenceCounted {
 		return instance != null;
 	}
 
-	public GlProgram getIndirectProgram(InstanceType<?> instanceType, ContextShader contextShader) {
-		return pipeline.get(new PipelineProgramKey(instanceType, contextShader));
+	public GlProgram getIndirectProgram(InstanceType<?> instanceType, ContextShader contextShader, LightShader light) {
+		return pipeline.get(new PipelineProgramKey(instanceType, contextShader, light));
 	}
 
 	public GlProgram getCullingProgram(InstanceType<?> instanceType) {

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

@@ -8,6 +8,7 @@ import org.jetbrains.annotations.Nullable;
 import com.google.common.collect.ImmutableList;
 
 import dev.engine_room.flywheel.api.instance.InstanceType;
+import dev.engine_room.flywheel.api.material.LightShader;
 import dev.engine_room.flywheel.backend.gl.GlCompat;
 import dev.engine_room.flywheel.backend.gl.shader.GlProgram;
 import dev.engine_room.flywheel.backend.glsl.GlslVersion;
@@ -78,8 +79,8 @@ public class InstancingPrograms extends AtomicReferenceCounted {
 		return instance != null;
 	}
 
-	public GlProgram get(InstanceType<?> instanceType, ContextShader contextShader) {
-		return pipeline.get(new PipelineProgramKey(instanceType, contextShader));
+	public GlProgram get(InstanceType<?> instanceType, ContextShader contextShader, LightShader light) {
+		return pipeline.get(new PipelineProgramKey(instanceType, contextShader, light));
 	}
 
 	@Override

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

@@ -56,7 +56,8 @@ public final class PipelineCompiler {
 						.nameMapper(key -> {
 							var context = key.contextShader()
 									.nameLowerCase();
-							return "pipeline/" + pipeline.compilerMarker() + "/" + context;
+							return "pipeline/" + pipeline.compilerMarker() + "/" + ResourceUtil.toDebugFileNameNoExtension(key.light()
+									.source()) + "_" + context;
 						})
 						.requireExtensions(extensions)
 						.enableExtension("GL_ARB_conservative_depth")
@@ -65,6 +66,8 @@ public final class PipelineCompiler {
 						.onCompile((key, comp) -> lightSmoothness.onCompile(comp))
 						.withResource(API_IMPL_FRAG)
 						.withComponents(fragmentComponents)
+						.withResource(key -> key.light()
+								.source())
 						.withResource(pipeline.fragmentMain()))
 				.preLink((key, program) -> {
 					program.bindAttribLocation("_flw_aPos", 0);

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

@@ -1,12 +1,14 @@
 package dev.engine_room.flywheel.backend.compile;
 
 import dev.engine_room.flywheel.api.instance.InstanceType;
+import dev.engine_room.flywheel.api.material.LightShader;
 
 /**
  * Represents the entire context of a program's usage.
  *
  * @param instanceType  The instance shader to use.
  * @param contextShader The context shader to use.
+ * @param light The light shader to use.
  */
-public record PipelineProgramKey(InstanceType<?> instanceType, ContextShader contextShader) {
+public record PipelineProgramKey(InstanceType<?> instanceType, ContextShader contextShader, LightShader light) {
 }

common/src/backend/java/dev/engine_room/flywheel/backend/engine/MaterialEncoder.java

@@ -55,8 +55,7 @@ public final class MaterialEncoder {
 	public static int packUberShader(Material material) {
 		var fog = MaterialShaderIndices.fogIndex(material.fog());
 		var cutout = MaterialShaderIndices.cutoutIndex(material.cutout());
-		var light = MaterialShaderIndices.lightIndex(material.light());
-		return (light & 0x3FF) | (cutout & 0x3FF) << 10 | (fog & 0x3FF) << 20;
+		return (cutout & 0xFFFF) | (fog & 0xFFFF) << 16;
 	}
 
 	// Packed format:

common/src/backend/java/dev/engine_room/flywheel/backend/engine/MaterialRenderState.java

@@ -16,7 +16,9 @@ import net.minecraft.client.Minecraft;
 import net.minecraft.client.renderer.texture.AbstractTexture;
 
 public final class MaterialRenderState {
-	public static final Comparator<Material> COMPARATOR = Comparator.comparing(Material::texture)
+	public static final Comparator<Material> COMPARATOR = Comparator.comparing((Material m) -> m.light()
+					.source())
+			.thenComparing(Material::texture)
 			.thenComparing(Material::blur)
 			.thenComparing(Material::mipmap)
 			.thenComparing(Material::backfaceCulling)
@@ -177,4 +179,18 @@ public final class MaterialRenderState {
 		RenderSystem.depthMask(true);
 		RenderSystem.colorMask(true, true, true, true);
 	}
+
+	public static boolean materialEquals(Material lhs, Material rhs) {
+		if (lhs == rhs) {
+			return true;
+		}
+
+		// Not here because ubershader: useLight, useOverlay, diffuse, shaders, fog shader, and cutout shader
+		// Everything in the comparator should be here.
+		return lhs.blur() == rhs.blur() && lhs.mipmap() == rhs.mipmap() && lhs.backfaceCulling() == rhs.backfaceCulling() && lhs.polygonOffset() == rhs.polygonOffset() && lhs.light()
+				.source()
+				.equals(rhs.light()
+						.source()) && lhs.texture()
+				.equals(rhs.texture()) && lhs.depthTest() == rhs.depthTest() && lhs.transparency() == rhs.transparency() && lhs.writeMask() == rhs.writeMask();
+	}
 }

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

@@ -28,6 +28,7 @@ import dev.engine_room.flywheel.backend.engine.embed.Environment;
 import dev.engine_room.flywheel.backend.engine.uniform.Uniforms;
 import dev.engine_room.flywheel.backend.gl.GlCompat;
 import dev.engine_room.flywheel.backend.gl.shader.GlProgram;
+import dev.engine_room.flywheel.lib.material.LightShaders;
 import dev.engine_room.flywheel.lib.math.MoreMath;
 
 public class IndirectCullingGroup<I extends Instance> {
@@ -49,7 +50,6 @@ public class IndirectCullingGroup<I extends Instance> {
 	private final IndirectPrograms programs;
 	private final GlProgram cullProgram;
 	private final GlProgram applyProgram;
-	private final GlProgram drawProgram;
 
 	private boolean needsDrawBarrier;
 	private boolean needsDrawSort;
@@ -65,7 +65,6 @@ public class IndirectCullingGroup<I extends Instance> {
 		this.programs = programs;
 		cullProgram = programs.getCullingProgram(instanceType);
 		applyProgram = programs.getApplyProgram();
-		drawProgram = programs.getIndirectProgram(instanceType, environment.contextShader());
 	}
 
 	public void flushInstancers() {
@@ -158,20 +157,23 @@ public class IndirectCullingGroup<I extends Instance> {
 
 		for (int start = 0, i = 0; i < indirectDraws.size(); i++) {
 			var draw1 = indirectDraws.get(i);
-			var material1 = draw1.material();
-			var visualType1 = draw1.visualType();
 
 			// if the next draw call has a different VisualType or Material, start a new MultiDraw
-			if (i == indirectDraws.size() - 1 || visualType1 != indirectDraws.get(i + 1)
-					.visualType() || !material1.equals(indirectDraws.get(i + 1)
-					.material())) {
-				multiDraws.computeIfAbsent(visualType1, s -> new ArrayList<>())
-						.add(new MultiDraw(material1, start, i + 1));
+			if (i == indirectDraws.size() - 1 || incompatibleDraws(draw1, indirectDraws.get(i + 1))) {
+				multiDraws.computeIfAbsent(draw1.visualType(), s -> new ArrayList<>())
+						.add(new MultiDraw(draw1.material(), start, i + 1));
 				start = i + 1;
 			}
 		}
 	}
 
+	private boolean incompatibleDraws(IndirectDraw draw1, IndirectDraw draw2) {
+		if (draw1.visualType() != draw2.visualType()) {
+			return true;
+		}
+		return !MaterialRenderState.materialEquals(draw1.material(), draw2.material());
+	}
+
 	public boolean hasVisualType(VisualType visualType) {
 		return multiDraws.containsKey(visualType);
 	}
@@ -199,17 +201,25 @@ public class IndirectCullingGroup<I extends Instance> {
 			return;
 		}
 
-		drawProgram.bind();
 		buffers.bindForDraw();
 
-		environment.setupDraw(drawProgram);
-
 		drawBarrier();
 
-		var flwBaseDraw = drawProgram.getUniformLocation("_flw_baseDraw");
+		GlProgram lastProgram = null;
+		int baseDrawUniformLoc = -1;
 
 		for (var multiDraw : multiDraws.get(visualType)) {
-			glUniform1ui(flwBaseDraw, multiDraw.start);
+			var drawProgram = programs.getIndirectProgram(instanceType, environment.contextShader(), multiDraw.material.light());
+			if (drawProgram != lastProgram) {
+				lastProgram = drawProgram;
+
+				// Don't need to do this unless the program changes.
+				drawProgram.bind();
+				environment.setupDraw(drawProgram);
+				baseDrawUniformLoc = drawProgram.getUniformLocation("_flw_baseDraw");
+			}
+
+			glUniform1ui(baseDrawUniformLoc, multiDraw.start);
 
 			MaterialRenderState.setup(multiDraw.material);
 
@@ -218,7 +228,7 @@ public class IndirectCullingGroup<I extends Instance> {
 	}
 
 	public void bindWithContextShader(ContextShader override) {
-		var program = programs.getIndirectProgram(instanceType, override);
+		var program = programs.getIndirectProgram(instanceType, override, LightShaders.SMOOTH_WHEN_EMBEDDED);
 
 		program.bind();
 
@@ -226,7 +236,7 @@ public class IndirectCullingGroup<I extends Instance> {
 
 		drawBarrier();
 
-		var flwBaseDraw = drawProgram.getUniformLocation("_flw_baseDraw");
+		var flwBaseDraw = program.getUniformLocation("_flw_baseDraw");
 		glUniform1ui(flwBaseDraw, 0);
 	}
 

common/src/backend/java/dev/engine_room/flywheel/backend/engine/instancing/InstancedDrawManager.java

@@ -28,6 +28,7 @@ import dev.engine_room.flywheel.backend.gl.GlStateTracker;
 import dev.engine_room.flywheel.backend.gl.TextureBuffer;
 import dev.engine_room.flywheel.backend.gl.array.GlVertexArray;
 import dev.engine_room.flywheel.backend.gl.shader.GlProgram;
+import dev.engine_room.flywheel.lib.material.LightShaders;
 import dev.engine_room.flywheel.lib.material.SimpleMaterial;
 import net.minecraft.client.resources.model.ModelBakery;
 
@@ -170,7 +171,7 @@ public class InstancedDrawManager extends DrawManager<InstancedInstancer<?>> {
 
 				GroupKey<?> shader = groupEntry.getKey();
 
-				var program = programs.get(shader.instanceType(), ContextShader.CRUMBLING);
+				var program = programs.get(shader.instanceType(), ContextShader.CRUMBLING, LightShaders.SMOOTH_WHEN_EMBEDDED);
 				program.bind();
 
 				for (var progressEntry : byProgress.int2ObjectEntrySet()) {

common/src/backend/java/dev/engine_room/flywheel/backend/engine/instancing/InstancedRenderStage.java

@@ -54,12 +54,13 @@ public class InstancedRenderStage {
 
 			var environment = shader.environment();
 
-			var program = programs.get(shader.instanceType(), environment.contextShader());
+			for (var drawCall : drawCalls.draws) {
+				var program = programs.get(shader.instanceType(), environment.contextShader(), drawCall.material()
+						.light());
 				program.bind();
 
 				environment.setupDraw(program);
 
-			for (var drawCall : drawCalls.draws) {
 				var material = drawCall.material();
 
 				uploadMaterialUniform(program, material);

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

@@ -1,4 +1,3 @@
 uint _flw_uberMaterialFragmentIndex;
 uint _flw_uberFogIndex;
 uint _flw_uberCutoutIndex;
-uint _flw_uberLightIndex;

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

@@ -6,7 +6,7 @@ flat in uvec3 _flw_packedMaterial;
 
 void main() {
     _flw_uberMaterialFragmentIndex = _flw_packedMaterial.x;
-    _flw_unpackUint3x10(_flw_packedMaterial.y, _flw_uberFogIndex, _flw_uberCutoutIndex, _flw_uberLightIndex);
+    _flw_unpackUint2x16(_flw_packedMaterial.y, _flw_uberFogIndex, _flw_uberCutoutIndex);
     _flw_unpackMaterialProperties(_flw_packedMaterial.z, flw_material);
 
     _flw_main();

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

@@ -5,7 +5,7 @@ uniform uvec4 _flw_packedMaterial;
 
 void main() {
     _flw_uberMaterialFragmentIndex = _flw_packedMaterial.y;
-    _flw_unpackUint3x10(_flw_packedMaterial.z, _flw_uberFogIndex, _flw_uberCutoutIndex, _flw_uberLightIndex);
+    _flw_unpackUint2x16(_flw_packedMaterial.z, _flw_uberFogIndex, _flw_uberCutoutIndex);
     _flw_unpackMaterialProperties(_flw_packedMaterial.w, flw_material);
 
     _flw_main();

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

@@ -53,9 +53,3 @@ void _flw_unpackUint2x16(uint s, out uint hi, out uint lo) {
     hi = (s >> 16) & 0xFFFFu;
     lo = s & 0xFFFFu;
 }
-
-void _flw_unpackUint3x10(uint s, out uint hi, out uint mi, out uint lo) {
-    hi = (s >> 20) & 0x3FFu;
-    mi = (s >> 10) & 0x3FFu;
-    lo = s & 0x3FFu;
-}