general improvements (#1049)

- Entity is now 64-bit uint
- terrain chunk hashing improvement
- structure memory alignment improvements
- light shafts improvements

Editor/GeneralWindow.cpp

@@ -411,7 +411,7 @@ void GeneralWindow::Create(EditorComponent* _editor)
 		}
 
 		theme.shadow_highlight = !focusModeCheckBox.GetCheck();
-		theme.shadow_highlight_spread = 0.6f;
+		theme.shadow_highlight_spread = 0.4f;
 		theme.shadow_highlight_color = theme_color_focus;
 		theme.shadow_highlight_color.x *= 1.4f;
 		theme.shadow_highlight_color.y *= 1.4f;

WickedEngine/shaders/ShaderInterop_Renderer.h

@@ -653,15 +653,18 @@ struct alignas(16) ShaderTransform
 
 struct alignas(16) ShaderMeshInstance
 {
-	uint uid;
+	uint64_t uid;
 	uint flags;	// high 8 bits: user stencilRef
 	uint layerMask;
-	uint meshletOffset; // offset in the global meshlet buffer for first subset (for LOD0)
 
-	uint geometryOffset;	// offset of all geometries for currently active LOD
-	uint geometryCount;		// number of all geometries in currently active LOD
+	uint meshletOffset;			// offset in the global meshlet buffer for first subset (for LOD0)
+	uint geometryOffset;		// offset of all geometries for currently active LOD
+	uint geometryCount;			// number of all geometries in currently active LOD
 	uint baseGeometryOffset;	// offset of all geometries of the instance (if no LODs, then it is equal to geometryOffset)
+
+	uint2 rimHighlight;			// packed half4
 	uint baseGeometryCount;		// number of all geometries of the instance (if no LODs, then it is equal to geometryCount)
+	float fadeDistance;
 
 	uint2 color; // packed half4
 	uint2 emissive; // packed half4
@@ -671,10 +674,6 @@ struct alignas(16) ShaderMeshInstance
 	int lightmap;
 	uint alphaTest_size; // packed half2
 
-	uint2 rimHighlight; // packed half4
-	float fadeDistance;
-	float padding;
-
 	float3 center;
 	float radius;
 

WickedEngine/shaders/ShaderInterop_SurfelGI.h

@@ -49,14 +49,13 @@ struct Surfel
 // This per-surfel structure will store all additional persistent data per surfel that isn't needed at GI lookup
 struct SurfelData
 {
+	uint64_t uid;
 	uint2 primitiveID;
-	uint bary;
-	uint uid;
 
+	uint bary;
 	uint raydata; // 24bit rayOffset, 8bit rayCount
 	uint properties; // 8bit life frames, 8bit recycle frames, 1bit backface normal
 	float max_inconsistency;
-	int padding1;
 
 	inline uint GetRayOffset() { return raydata & 0xFFFFFF; }
 	inline uint GetRayCount() { return (raydata >> 24u) & 0xFF; }

WickedEngine/shaders/lightShaftsCS.hlsl

@@ -3,40 +3,34 @@
 
 PUSHCONSTANT(postprocess, PostProcess);
 
-Texture2D<float4> input : register(t0);
+Texture2D<half4> input : register(t0);
 
-RWTexture2D<float4> output : register(u0);
+RWTexture2D<half4> output : register(u0);
 
-static const uint NUM_SAMPLES = 32;
-static const uint UNROLL_GRANULARITY = 8;
+static const uint NUM_SAMPLES = 64;
 
 [numthreads(POSTPROCESS_BLOCKSIZE, POSTPROCESS_BLOCKSIZE, 1)]
 void main(uint3 DTid : SV_DispatchThreadID)
 {
-	float2 uv = (DTid.xy + 0.5f) * postprocess.resolution_rcp;
+	float2 uv = (DTid.xy + 0.5) * postprocess.resolution_rcp;
 
-	float3 color = input.SampleLevel(sampler_linear_clamp, uv, 0).rgb;
+	half3 color = input.SampleLevel(sampler_linear_clamp, uv, 0).rgb;
 
 	float2 lightPos = postprocess.params1.xy;
 	float2 deltaTexCoord =  uv - lightPos;
 	deltaTexCoord *= postprocess.params0.x / NUM_SAMPLES;
-	float illuminationDecay = 1.0f;
+	half illuminationDecay = 1.0;
 	
-	[loop] // loop big part (balance register pressure)
-	for (uint i = 0; i < NUM_SAMPLES / UNROLL_GRANULARITY; i++)
+	for (uint i = 0; i < NUM_SAMPLES; i++)
 	{
-		[unroll] // unroll small parts (balance register pressure)
-		for (uint j = 0; j < UNROLL_GRANULARITY; ++j)
-		{
-			uv.xy -= deltaTexCoord;
-			float3 sam = input.SampleLevel(sampler_linear_clamp, uv.xy, 0).rgb;
-			sam *= illuminationDecay * postprocess.params0.y;
-			color.rgb += sam;
-			illuminationDecay *= postprocess.params0.z;
-		}
+		uv.xy -= deltaTexCoord;
+		half3 sam = input.SampleLevel(sampler_linear_clamp, uv.xy, 0).rgb;
+		sam *= illuminationDecay * postprocess.params0.y;
+		color.rgb += sam;
+		illuminationDecay *= postprocess.params0.z;
 	}
 
 	color *= postprocess.params0.w;
 
-	output[DTid.xy] = float4(color, 1);
+	output[DTid.xy] = half4(color, 1);
 }

WickedEngine/wiECS.h

@@ -19,9 +19,10 @@ namespace wi::ecs
 	// The Entity is a global unique persistent identifier within the entity-component system
 	//	It can be stored and used for the duration of the application
 	//	The entity can be a different value on a different run of the application, if it was serialized
-	//	It must be only serialized with the SerializeEntity() function. It will ensure that entities still match with their components correctly after serialization
-	using Entity = uint32_t;
-	inline constexpr Entity INVALID_ENTITY = 0;
+	//	It must be only serialized with the SerializeEntity() function if persistence is needed across different program runs,
+	//		this will ensure that entities still match with their components correctly after serialization
+	using Entity = uint64_t;
+	inline static constexpr Entity INVALID_ENTITY = 0;
 	// Runtime can create a new entity with this
 	inline Entity CreateEntity()
 	{
@@ -142,9 +143,14 @@ namespace wi::ecs
 		// reservedCount : how much components can be held initially before growing the container
 		ComponentManager(size_t reservedCount = 0)
 		{
-			components.reserve(reservedCount);
-			entities.reserve(reservedCount);
-			lookup.reserve(reservedCount);
+			Reserve(reservedCount);
+		}
+
+		inline void Reserve(size_t count)
+		{
+			components.reserve(count);
+			entities.reserve(count);
+			lookup.reserve(count);
 		}
 
 		// Clear the whole container

WickedEngine/wiHairParticle.h

@@ -60,7 +60,6 @@ namespace wi
 			wi::graphics::CommandList cmd
 		);
 
-		mutable bool gpu_initialized = false;
 		void InitializeGPUDataIfNeeded(wi::graphics::CommandList cmd);
 
 		void Draw(
@@ -69,6 +68,8 @@ namespace wi
 			wi::graphics::CommandList cmd
 		) const;
 
+		wi::ecs::Entity meshID = wi::ecs::INVALID_ENTITY;
+
 		enum FLAGS
 		{
 			EMPTY = 0,
@@ -78,8 +79,6 @@ namespace wi
 		};
 		uint32_t _flags = EMPTY;
 
-		wi::ecs::Entity meshID = wi::ecs::INVALID_ENTITY;
-
 		uint32_t strandCount = 0;
 		uint32_t segmentCount = 1;
 		uint32_t randomSeed = 1;
@@ -106,6 +105,7 @@ namespace wi
 		mutable bool regenerate_frame = true;
 		wi::graphics::Format position_format = wi::graphics::Format::R16G16B16A16_UNORM;
 		mutable bool must_rebuild_blas = true;
+		mutable bool gpu_initialized = false;
 
 		void Serialize(wi::Archive& archive, wi::ecs::EntitySerializer& seri);
 

WickedEngine/wiRenderPath3D.cpp

@@ -36,6 +36,7 @@ namespace wi
 		rtShadow = {};
 		rtSun[0] = {};
 		rtSun[1] = {};
+		rtSun[2] = {};
 		rtSun_resolved = {};
 		rtGUIBlurredBackground[0] = {};
 		rtGUIBlurredBackground[1] = {};
@@ -1914,6 +1915,8 @@ namespace wi
 
 			device->EventBegin("Light Shafts", cmd);
 
+			const Texture* texture_fullres = nullptr;
+
 			// Render sun stencil cutout:
 			{
 				if (getMSAASampleCount() > 1)
@@ -1931,6 +1934,7 @@ namespace wi
 						RenderPassImage::Resolve(&rtSun_resolved),
 					};
 					device->RenderPassBegin(rp, arraysize(rp), cmd);
+					texture_fullres = &rtSun_resolved;
 				}
 				else
 				{
@@ -1946,6 +1950,7 @@ namespace wi
 						RenderPassImage::RenderTarget(&rtSun[0], RenderPassImage::LoadOp::CLEAR),
 					};
 					device->RenderPassBegin(rp, arraysize(rp), cmd);
+					texture_fullres = &rtSun[0];
 				}
 
 				Viewport vp;
@@ -1977,10 +1982,13 @@ namespace wi
 					1.0f, 1.0f, 0.1f, 1.0f,
 					camera->GetProjection(), camera->GetView(), XMMatrixIdentity());
 				{
+					// Downsample to low res first:
+					wi::renderer::Postprocess_Downsample4x(*texture_fullres, rtSun[2], cmd);
+
 					XMFLOAT2 sun;
 					XMStoreFloat2(&sun, sunPos);
 					wi::renderer::Postprocess_LightShafts(
-						getMSAASampleCount() > 1 ? rtSun_resolved : rtSun[0],
+						rtSun[2],
 						rtSun[1],
 						cmd,
 						sun,
@@ -2919,10 +2927,12 @@ namespace wi
 
 			desc.bind_flags = BindFlag::SHADER_RESOURCE | BindFlag::UNORDERED_ACCESS;
 			desc.sample_count = 1;
-			desc.width = internalResolution.x / 2;
-			desc.height = internalResolution.y / 2;
+			desc.width = internalResolution.x / 4;
+			desc.height = internalResolution.y / 4;
 			device->CreateTexture(&desc, nullptr, &rtSun[1]);
 			device->SetName(&rtSun[1], "rtSun[1]");
+			device->CreateTexture(&desc, nullptr, &rtSun[2]);
+			device->SetName(&rtSun[2], "rtSun[2]");
 
 			if (getMSAASampleCount() > 1)
 			{
@@ -2937,6 +2947,7 @@ namespace wi
 		{
 			rtSun[0] = {};
 			rtSun[1] = {};
+			rtSun[2] = {};
 			rtSun_resolved = {};
 		}
 	}

WickedEngine/wiRenderPath3D.h

@@ -107,7 +107,7 @@ namespace wi
 		wi::graphics::Texture rtBloom_tmp; // temporary for bloom downsampling
 		wi::graphics::Texture rtAO; // full res AO
 		wi::graphics::Texture rtShadow; // raytraced shadows mask
-		wi::graphics::Texture rtSun[2]; // 0: sun render target used for lightshafts (can be MSAA), 1: radial blurred lightshafts
+		wi::graphics::Texture rtSun[3]; // 0: sun render target used for lightshafts (can be MSAA), 1: radial blurred lightshafts
 		wi::graphics::Texture rtSun_resolved; // sun render target, but the resolved version if MSAA is enabled
 		wi::graphics::Texture rtGUIBlurredBackground[3];	// downsampled, gaussian blurred scene for GUI
 		wi::graphics::Texture rtShadingRate; // UINT8 shading rate per tile

WickedEngine/wiScene.cpp

@@ -4417,7 +4417,7 @@ namespace wi::scene
 						const float dist = std::sqrt(distsq);
 						const float dist_to_sphere = dist - radius;
 						object.lod = uint32_t(dist_to_sphere * object.lod_distance_multiplier);
-						object.lod = std::min(object.lod, mesh.GetLODCount() - 1);
+						object.lod = std::min(object.lod, uint16_t(mesh.GetLODCount() - 1));
 					}
 				}
 

WickedEngine/wiScene_BindLua.cpp

@@ -4796,7 +4796,7 @@ int MeshComponent_BindLua::SetMeshSubsetMaterialID(lua_State* L)
 	if (argc >= 2)
 	{
 		size_t subsetindex = (uint32_t)wi::lua::SGetLongLong(L, 1);
-		Entity entity = (uint32_t)wi::lua::SGetLongLong(L, 2);
+		Entity entity = (Entity)wi::lua::SGetLongLong(L, 2);
 
 		const uint32_t lod_count = component->GetLODCount();
 		for (uint32_t lod = 0; lod < lod_count; ++lod)

WickedEngine/wiTerrain.cpp

@@ -332,7 +332,7 @@ namespace wi::terrain
 			lod_count++;
 		}
 
-		locker.lock();
+		std::scoped_lock lck(locker);
 		free(atlas);
 		if (tile_count > 1)
 		{
@@ -852,7 +852,7 @@ namespace wi::terrain
 		}
 
 		// Start the generation on a background thread and keep it running until the next frame
-		wi::jobsystem::Execute(generator->workload, [=](wi::jobsystem::JobArgs args) {
+		wi::jobsystem::Execute(generator->workload, [=](wi::jobsystem::JobArgs a) {
 
 			wi::Timer timer;
 			bool generated_something = false;
@@ -926,6 +926,7 @@ namespace wi::terrain
 
 					// Do a parallel for loop over all the chunk's vertices and compute their properties:
 					wi::jobsystem::context ctx;
+					ctx.priority = wi::jobsystem::Priority::Low;
 					wi::jobsystem::Dispatch(ctx, vertexCount, chunk_width, [&](wi::jobsystem::JobArgs args) {
 						uint32_t index = args.jobIndex;
 						const float x = (float(index % chunk_width) - chunk_half_width) * chunk_scale;

WickedEngine/wiTerrain.h

@@ -14,14 +14,21 @@ namespace wi::terrain
 {
 	struct Chunk
 	{
-		int x, z;
+		union
+		{
+			struct
+			{
+				int x, z;
+			};
+			uint64_t raw = 0;
+		};
 		constexpr bool operator==(const Chunk& other) const
 		{
-			return (x == other.x) && (z == other.z);
+			return raw == other.raw;
 		}
-		inline size_t compute_hash() const
+		constexpr uint64_t compute_hash() const
 		{
-			return ((std::hash<int>()(x) ^ (std::hash<int>()(z) << 1)) >> 1);
+			return raw;
 		}
 	};
 }
@@ -31,7 +38,7 @@ namespace std
 	template <>
 	struct hash<wi::terrain::Chunk>
 	{
-		inline size_t operator()(const wi::terrain::Chunk& chunk) const
+		constexpr uint64_t operator()(const wi::terrain::Chunk& chunk) const
 		{
 			return chunk.compute_hash();
 		}
@@ -95,7 +102,7 @@ namespace wi::terrain
 			wi::graphics::GPUBuffer pageBuffer;
 			wi::graphics::GPUBuffer pageBuffer_CPU_upload[wi::graphics::GraphicsDevice::GetBufferCount()];
 			bool data_available_CPU[wi::graphics::GraphicsDevice::GetBufferCount()] = {};
-			int cpu_resource_id = 0;
+			int16_t cpu_resource_id = 0;
 			uint32_t resolution = 0;
 
 			void init(uint32_t resolution);
@@ -213,10 +220,10 @@ namespace wi::terrain
 		wi::ecs::Entity entity = wi::ecs::INVALID_ENTITY;
 		wi::ecs::Entity grass_entity = wi::ecs::INVALID_ENTITY;
 		wi::ecs::Entity props_entity = wi::ecs::INVALID_ENTITY;
-		const XMFLOAT3* mesh_vertex_positions = nullptr;
 		float prop_density_current = 1;
-		wi::HairParticleSystem grass;
 		float grass_density_current = 1;
+		const XMFLOAT3* mesh_vertex_positions = nullptr;
+		wi::HairParticleSystem grass;
 		wi::vector<BlendmapLayer> blendmap_layers;
 		wi::graphics::Texture blendmap;
 		wi::primitive::Sphere sphere;
@@ -231,7 +238,6 @@ namespace wi::terrain
 			while (blendmap_layers.size() < materialIndex + 1)
 			{
 				blendmap_layers.emplace_back().pixels.resize(vertexCount);
-				std::fill(blendmap_layers.back().pixels.begin(), blendmap_layers.back().pixels.end(), 0);
 			}
 		}
 	};
@@ -284,15 +290,15 @@ namespace wi::terrain
 		int grass_chunk_dist = 1;
 
 		// For generating scene on a background thread:
+		float generation_time_budget_milliseconds = 8; // after this much time, the generation thread will start to exit. This can help avoid a very long running, resource consuming and slow cancellation generation
 		std::shared_ptr<Generator> generator;
-		float generation_time_budget_milliseconds = 12; // after this much time, the generation thread will exit. This can help avoid a very long running, resource consuming and slow cancellation generation
 
 		wi::vector<VirtualTexture*> virtual_textures_in_use;
 		wi::graphics::Sampler sampler;
 		VirtualTextureAtlas atlas;
 
-		int chunk_buffer_range = 3; // how many chunks to upload to GPU in X and Z directions
 		wi::graphics::GPUBuffer chunk_buffer;
+		int chunk_buffer_range = 3; // how many chunks to upload to GPU in X and Z directions
 
 		constexpr bool IsCenterToCamEnabled() const { return _flags & CENTER_TO_CAM; }
 		constexpr bool IsRemovalEnabled() const { return _flags & REMOVAL; }

WickedEngine/wiVersion.cpp

@@ -9,7 +9,7 @@ namespace wi::version
 	// minor features, major updates, breaking compatibility changes
 	const int minor = 71;
 	// minor bug fixes, alterations, refactors, updates
-	const int revision = 668;
+	const int revision = 669;
 
 	const std::string version_string = std::to_string(major) + "." + std::to_string(minor) + "." + std::to_string(revision);
 
@@ -30,7 +30,7 @@ namespace wi::version
 		return version_string.c_str();
 	}
 
-#define SUPPORTERS "Nemerle, James Webb, Quifeng Jin, TheGameCreators, Joseph Goldin, Yuri, Sergey K, Yukawa Kanta, Dragon Josh, John, LurkingNinja, Bernardo Del Castillo, Invictus, Scott Hunt, Yazan Altaki, Tuan NV, Robert MacGregor, cybernescence, Alexander Dahlin, blueapples, Delhills, NI NI, Sherief, ktopoet, Justin Macklin, Cédric Fabre, TogetherTeam, Bartosz Boczula, Arne Koenig, Ivan Trajchev, nathants, Fahd Ahmed, Gabriel Jadderson, SAS_Controller, Dominik Madarász, Segfault, Mike amanfo, Dennis Brakhane, rookie, Peter Moore, therealjtgill, Nicolas Embleton, Desuuc, radino1977, Anthony Curtis, manni heck, Matthias Hölzl, Phyffer, Lucas Pinheiro, Tapkaara, gpman, Anthony Python, Gnowos, Klaus, slaughternaut, Paul Brain, Connor Greaves, Alexandr, Lee Bamber, MCAlarm MC2, Titoutan, Willow, Aldo, lokimx, K. Osterman, Nomad, ykl, Alex Krokos, Timmy, Avaflow, mat, Hexegonel Samael Michael, Joe Spataro, soru, GeniokV, Mammoth, Ignacio, datae, Jason Rice, MarsBEKET, Tim, Twisty, Zelf ieats kiezen, Romildo Franco, zNachoh, Dmitriy, Alex Minerva, Stefan Kent, Natty, Sunny Krishna, Vilmos Malárik, Ferrata, Rossakis, Stefana Andrei, Taylor, Gunnar Kriik, 赟 杨, Rex, Lemon Brother, fixy, meta_leap"
+#define SUPPORTERS "Nemerle, James Webb, Quifeng Jin, TheGameCreators, Joseph Goldin, Yuri, Sergey K, Yukawa Kanta, Dragon Josh, John, LurkingNinja, Bernardo Del Castillo, Invictus, Scott Hunt, Yazan Altaki, Tuan NV, Robert MacGregor, cybernescence, Alexander Dahlin, blueapples, Delhills, NI NI, Sherief, ktopoet, Justin Macklin, Cédric Fabre, TogetherTeam, Bartosz Boczula, Arne Koenig, Ivan Trajchev, nathants, Fahd Ahmed, Gabriel Jadderson, SAS_Controller, Dominik Madarász, Segfault, Mike amanfo, Dennis Brakhane, rookie, Peter Moore, therealjtgill, Nicolas Embleton, Desuuc, radino1977, Anthony Curtis, manni heck, Matthias Hölzl, Phyffer, Lucas Pinheiro, Tapkaara, gpman, Anthony Python, Gnowos, Klaus, slaughternaut, Paul Brain, Connor Greaves, Alexandr, Lee Bamber, MCAlarm MC2, Titoutan, Willow, Aldo, lokimx, K. Osterman, Nomad, ykl, Alex Krokos, Timmy, Avaflow, mat, Hexegonel Samael Michael, Joe Spataro, soru, GeniokV, Mammoth, Ignacio, datae, Jason Rice, MarsBEKET, Tim, Twisty, Zelf ieats kiezen, Romildo Franco, zNachoh, Dmitriy, Alex Minerva, Stefan Kent, Natty, Sunny Krishna, Vilmos Malárik, Ferrata, Rossakis, Stefana Andrei, Taylor, Gunnar Kriik, 赟 杨, Rex, Lemon Brother, fixy, meta_leap, Edik, jusik5348"
 
 	const char* GetCreditsString()
 	{