New denoiser, more async

Author: Cheaterdev

sources/Core/Math/Constants.h

@@ -66,7 +66,7 @@ namespace Math
 		return AlignUpWithMask(value, alignment - 1);
 	}
 
-	template <typename T> __forceinline T roundUp(T a, size_t b) { return (1 + (a - 1) / b) * b; }
+	template <typename T> __forceinline T roundUp(T num, size_t factor) { return num + factor - 1 - (num + factor - 1) % factor; }
 
 	template <typename T> __forceinline T AlignDown(T value, size_t alignment)
 	{

sources/DirectXFramework/DX12/CommandList.cpp

@@ -267,7 +267,10 @@ namespace DX12
 		list->OMSetRenderTargets(c, &rt.cpu, true, h.is_valid() ? &h.cpu : nullptr);
 	}
 
-
+void GraphicsContext::set_rtv(std::initializer_list<Handle> rt, Handle h)
+	{
+		set_rtv(place_rtv(rt), h);	
+	}
 	void GraphicsContext::flush_binds(bool force)
 	{
 		base.flush_heaps(force);

sources/DirectXFramework/DX12/CommandList.h

@@ -1126,10 +1126,11 @@ namespace DX12
 		void set_viewports(std::vector<Viewport> viewports);
 
 
+		void set_rtv(std::initializer_list<Handle> rt, Handle h);
 
 		void set_rtv(const HandleTable&, Handle);
 		void set_rtv(int c, Handle rt, Handle h);
-
+	
 		void set_rtv(const HandleTableLight&, Handle);
 
 		void draw(D3D12_DRAW_INDEXED_ARGUMENTS args)
@@ -1174,13 +1175,29 @@ namespace DX12
 			set_rtvs_internal(t, h);
 			set_rtvs_internal(t + 1, list...);
 		}
-
+		template<class T, size_t N>
+		constexpr size_t size(T(&)[N]) { return N; }
+		
 		template<class ...Handles>
-		void set_rtvs(const Handle& h, Handles... list)
+		void set_rtvs(const Handle& h, Handles... rtvlist)
 		{
-			D3D12_CPU_DESCRIPTOR_HANDLE handles[sizeof...(Handles)];
-			set_rtvs_internal(handles, list...);
-			list->OMSetRenderTargets(sizeof...(Handles), handles, false, h.is_valid() ? &h.cpu : nullptr);
+
+			auto f = [&](Handle h) {
+				get_base().transition_rtv(h.resource_info);
+			};
+		
+			(f(std::forward<Handles>(rtvlist)), ...);
+
+			
+			if (h.is_valid())
+				get_base().transition_dsv(h.resource_info);
+
+			get_base().flush_transitions();
+
+			
+			
+			CD3DX12_CPU_DESCRIPTOR_HANDLE ar[] = { (rtvlist.cpu)... };
+			list->OMSetRenderTargets(size(ar), ar, false, h.is_valid() ? &h.cpu : nullptr);
 		}
 
 

sources/DirectXFramework/DX12/ResourceViews.h

@@ -126,7 +126,7 @@ namespace DX12
 	class TextureView :public ResourceView
 	{
 		HandleTableLight hlsl;
-		//HandleTableLight rtv;
+		HandleTableLight rtv;
 	public:
 
 
@@ -139,6 +139,7 @@ namespace DX12
 		HLSL::TextureCube<> textureÑube;
 		HLSL::Texture2DArray<> texture2DArray;
 
+		Handle renderTarget;
 	public:
 		template<class F>
 		void init(F& frame)
@@ -170,15 +171,15 @@ namespace DX12
 
 				place_uav(hlsl[1]);
 			}
-			/*
+			
 			if (desc.Flags & D3D12_RESOURCE_FLAGS::D3D12_RESOURCE_FLAG_ALLOW_RENDER_TARGET) {
 
-				rtv = frame.rtv_cpu.place(1);
+				rtv = frame.get_cpu_heap(DescriptorHeapType::RTV).place(1);
 
 				renderTarget = rtv[0];
-				place_uav(renderTarget);
+				place_rtv(renderTarget);
 			}
-			*/
+			
 		}
 		TextureView() = default;
 

sources/DirectXFramework/DX12/States.h

@@ -120,28 +120,32 @@ namespace DX12
 		};
 
 
-		//mutable  SpinLock states_lock;
-
+	
 		mutable std::array<SubResourcesCPU, 128> cpu_state;
 
+		mutable  SpinLock states_lock;
+		mutable std::map<int, SubResourcesCPU> cpu_state_map;
+
 		SubResourcesCPU& get_state(int id) const
 		{
+			if(id<128)
 			return cpu_state[id];
-			/*
+			
+			
 			std::lock_guard<SpinLock> guard(states_lock);
 
-			auto it = cpu_state[id];
+			auto it = cpu_state_map.find(id);
 
-			if (it == cpu_state.end())
+			if (it == cpu_state_map.end())
 			{
-				auto & state = cpu_state[id];
+				auto & state = cpu_state_map[id];
 				state.subres.resize(gpu_state.subres.size());
 
 				return state;
 			}
 
 			return it->second;
-			*/
+			
 		}
 
 

sources/FileSystem/FileSystem/FileSystem.h

@@ -35,7 +35,7 @@ struct file
 	std::shared_ptr<istream> get_new_stream()
 	{
 		std::filesystem::path full_path(std::filesystem::current_path());
-		Log::get() << "Current path is : " << full_path.generic_string() << Log::endl;
+	//	Log::get() << "Current path is : " << full_path.generic_string() << Log::endl;
 
 		return provider->create_stream(file_name);
 	}

sources/FlowGraph/FrameGraph.cpp

@@ -774,6 +774,11 @@ void TaskBuilder::create_resources()
 				flags |= D3D12_RESOURCE_FLAGS::D3D12_RESOURCE_FLAG_ALLOW_UNORDERED_ACCESS;
 			}
 
+			if (check(info->flags & ResourceFlags::Counted))
+			{
+				//info->heap_type = Render::HeapType::READBACK;
+			}
+			
 			info->d3ddesc = CD3DX12_RESOURCE_DESC::Buffer(desc.size, flags);
 		}
 /*

sources/FlowGraph/FrameGraph.h

@@ -10,24 +10,25 @@ enum class ResourceType :int {
 
 enum class ResourceFlags :int {
 	None = 0,
-	PixelRead = 1,
-	ComputeRead = 2,
-	DSRead = 1024,
+	PixelRead = (1<<1),
+	ComputeRead = (1 << 2),
+	DSRead = (1 << 3),
 
-	UnorderedAccess = 4,
-	RenderTarget = 8,
-	DepthStencil = 16,
+	UnorderedAccess = (1 << 4),
+	RenderTarget = (1 << 5),
+	DepthStencil = (1 << 6),
 
 
-	GenCPU = 32,
-	ReadCPU = 64,
+	GenCPU = (1 << 7),
+	ReadCPU = (1 << 8),
 
 	Temporal = 0,
-	Static = 128,
-	Required = 256,
+	Static = (1 << 9),
+	Required = (1 << 10),
 
-	Cube = 512,
-	Changed = 1024
+	Cube = (1 << 11),
+	Counted = (1 << 12),
+	Changed = (1 << 13)
 
 	, GENERATE_OPS
 };

sources/RenderSystem/Assets/MeshAsset.cpp

@@ -389,7 +389,7 @@ bool MeshAssetInstance::update_transforms()
 {
 	bool res = scene_object::update_transforms();
 
-	if (res)
+	if (res|| need_update_mats)
 	{
 
 //		node_buffer.data().clear();
@@ -407,21 +407,23 @@ bool MeshAssetInstance::update_transforms()
 		if (!info.primitive_global)
 			info.primitive_global = std::make_shared<AABB>();
 
-
-		mat4x4 mat = nodes[index].asset_node->mesh_matrix*global_transform;
-		info.primitive_global->apply_transform(p, mat);
+		mat4x4 prev_mat = info.global_mat;
+		info.global_mat = nodes[index].asset_node->mesh_matrix*global_transform;
+		info.primitive_global->apply_transform(p, info.global_mat);
 
 		auto &my_node = gpu_nodes[info.mesh_info.GetNode_offset() - nodes_handle.get_offset()];
-		my_node.node_global_matrix = mat;
-		my_node.node_inverse_matrix = mat;
+		my_node.node_global_matrix = info.global_mat;
+		my_node.node_global_matrix_prev = prev_mat;
+
+		my_node.node_inverse_matrix = info.global_mat;
 		my_node.node_inverse_matrix.inverse();
 
 
 		my_node.aabb.min = info.primitive->get_min();
 		my_node.aabb.max = info.primitive->get_max();
 
 
-	
+		need_update_mats = prev_mat != info.global_mat;
 
 	}
 
@@ -579,9 +581,11 @@ nodes.emplace_back(node);
 
 			info.primitive_global = std::make_shared<AABB>();
 
-			mat4x4 mat = nodes.back().asset_node->mesh_matrix * global_transform;
 
-			info.primitive_global->apply_transform(info.primitive, mat);
+		
+			info.global_mat = nodes.back().asset_node->mesh_matrix * global_transform;
+
+			info.primitive_global->apply_transform(info.primitive, info.global_mat);
 
 
 		//	node_buffer[node->index] = mat;
@@ -591,11 +595,12 @@ nodes.emplace_back(node);
 			info.mesh_info.GetVertex_offset() = UINT(mesh_asset->vertex_handle.get_offset() + mesh_asset->meshes[m].vertex_offset) ;
 
 			auto &my_node = gpu_nodes[UINT(info.mesh_info.GetNode_offset() - nodes_handle.get_offset())];
-			my_node.node_global_matrix = mat;
-			my_node.node_inverse_matrix = mat;
+			my_node.node_global_matrix = info.global_mat;
+			my_node.node_inverse_matrix = info.global_mat;
 			my_node.node_inverse_matrix.inverse();
 
-
+			my_node.node_global_matrix_prev = info.global_mat;
+			
 			my_node.aabb.min = info.primitive->get_min();
 			my_node.aabb.max = info.primitive->get_max();
 

sources/RenderSystem/Assets/MeshAsset.h

@@ -259,6 +259,8 @@ class MeshAssetInstance : public scene_object, public material_holder, public As
 
         size_t nodes_count;
         size_t rendering_count;
+
+        bool need_update_mats = false;
     public:
 		MESH_TYPE type= MESH_TYPE::STATIC;
 		MeshAsset::ref mesh_asset;
@@ -287,7 +289,8 @@ class MeshAssetInstance : public scene_object, public material_holder, public As
             MaterialAsset* material;
 
             UINT node_id;
-
+            mat4x4 global_mat;
+            
             //compiled
 			Slots::MeshInfo::Compiled compiled_mesh_info;