CAD DTM Rendering

include/nbl/core/containers/LRUCache.h

@@ -389,7 +389,7 @@ class ResizableLRUCache : protected impl::LRUCacheBase<Key, Value, MapHash, MapE
 			return nullptr;
 	}
 
-	//remove element at key if present
+	// remove element at key if present
 	inline void erase(const Key& key)
 	{
 		bool success;
@@ -400,6 +400,20 @@ class ResizableLRUCache : protected impl::LRUCacheBase<Key, Value, MapHash, MapE
 			m_shortcut_map.erase(iterator);
 		}
 	}
+
+	// returns key for least recently used
+	// use in evictin callback to know which Key is being evicted
+	inline const Key* get_least_recently_used() const
+	{
+		if (m_shortcut_map.size() > 0)
+			return &base_t::m_list.getBack()->data.first;
+		else
+			return nullptr;
+	}
+
+	inline size_t size() const { return m_shortcut_map.size(); }
+
+	inline bool empty() const { return size() <= 0ull;  }
 
 	/**
 	* @brief Resizes the cache by extending its capacity so it can hold more elements. Returns a bool indicating if capacity was indeed increased.

include/nbl/video/alloc/SubAllocatedDescriptorSet.h

@@ -28,11 +28,18 @@ class SubAllocatedDescriptorSet : public core::IReferenceCounted
 		class DeferredFreeFunctor
 		{
 			public:
-				inline DeferredFreeFunctor(SubAllocatedDescriptorSet* composed, uint32_t binding, size_type count, const value_type* addresses)
-					: m_addresses(std::move(core::make_refctd_dynamic_array<core::smart_refctd_dynamic_array<value_type>>(count))), 
-					  m_binding(binding), m_composed(composed)
+				 using ref_t = core::smart_refctd_ptr<const core::IReferenceCounted>;
+
+				template<typename T> requires std::is_base_of_v<core::IReferenceCounted,T>
+				inline DeferredFreeFunctor(SubAllocatedDescriptorSet* composed, uint32_t binding, size_type count, const value_type* addresses, const T*const *const objectsToHold)
+					: m_addresses(core::make_refctd_dynamic_array<core::smart_refctd_dynamic_array<value_type>>(count)) 
+					, m_objectsToHold(core::make_refctd_dynamic_array<core::smart_refctd_dynamic_array<ref_t>>(count))
+					, m_binding(binding)
+					, m_composed(composed)
 				{
 					memcpy(m_addresses->data(), addresses, count * sizeof(value_type));
+					for (size_t i=0u; i<count; i++)
+						(*m_objectsToHold)[i] = objectsToHold ? ref_t(objectsToHold[i]) : nullptr;
 				}
 				inline DeferredFreeFunctor(DeferredFreeFunctor&& other) 
 				{
@@ -58,11 +65,13 @@ class SubAllocatedDescriptorSet : public core::IReferenceCounted
 				{
 					m_composed = other.m_composed;
 					m_addresses = other.m_addresses;
+					m_objectsToHold = other.m_objectsToHold;
 					m_binding = other.m_binding;
 
 					// Nullifying other
 					other.m_composed = nullptr;
 					other.m_addresses = nullptr;
+					other.m_objectsToHold = nullptr;
 					other.m_binding = 0;
 					return *this;
 				}
@@ -99,6 +108,7 @@ class SubAllocatedDescriptorSet : public core::IReferenceCounted
 				}
 			protected:
 				core::smart_refctd_dynamic_array<value_type> m_addresses;
+				core::smart_refctd_dynamic_array<ref_t> m_objectsToHold;
 				SubAllocatedDescriptorSet* m_composed; // TODO: shouldn't be called `composed`, maybe `parent` or something
 				uint32_t m_binding;
 		};
@@ -209,9 +219,9 @@ class SubAllocatedDescriptorSet : public core::IReferenceCounted
 				remainingFrees = cull_frees();
 			} while (remainingFrees > 0);
 
-			for (uint32_t i = 0; i < m_allocatableRanges.size(); i++)
+			for (auto& it : m_allocatableRanges)
 			{
-				auto& range = m_allocatableRanges[i];
+				auto& range = it.second;
 				if (range.reservedSize == 0)
 					continue;
 				assert(range.eventHandler->getTimelines().size() == 0);
@@ -355,10 +365,11 @@ class SubAllocatedDescriptorSet : public core::IReferenceCounted
 		}
 
 		// defers based on the conservative estimation if `futureWait` needs to be waited on, if doesn't will call nullify descriiptors internally immediately
-		inline void multi_deallocate(uint32_t binding, size_type count, const value_type* addr, const ISemaphore::SWaitInfo& futureWait) noexcept
+		template<typename T=core::IReferenceCounted>
+		inline void multi_deallocate(uint32_t binding, size_type count, const value_type* addr, const ISemaphore::SWaitInfo& futureWait, const T*const *const objectsToDrop=nullptr) noexcept
 		{
 			if (futureWait.semaphore)
-				multi_deallocate(binding, futureWait, DeferredFreeFunctor(this, binding, count, addr));
+				multi_deallocate(binding, futureWait, DeferredFreeFunctor(this, binding, count, addr, objectsToDrop));
 			else
 			{
 				core::vector<IGPUDescriptorSet::SDropDescriptorSet> nulls(count);
@@ -376,10 +387,9 @@ class SubAllocatedDescriptorSet : public core::IReferenceCounted
 			uint32_t frees = 0;
 			core::vector<IGPUDescriptorSet::SDropDescriptorSet> nulls(m_totalDeferredFrees);
 			auto outNulls = nulls.data();
-			for (uint32_t i = 0; i < m_allocatableRanges.size(); i++)
+			for (auto& it : m_allocatableRanges)
 			{
-				auto& it = m_allocatableRanges[i];
-				frees += it.eventHandler->poll(outNulls).eventsLeft;
+				frees += it.second.eventHandler->poll(outNulls).eventsLeft;
 			}
 			getDevice()->nullifyDescriptors({nulls.data(),outNulls});
 			return frees;

src/nbl/video/CVulkanLogicalDevice.cpp

@@ -764,7 +764,7 @@ void CVulkanLogicalDevice::updateDescriptorSets_impl(const SUpdateDescriptorSets
 void CVulkanLogicalDevice::nullifyDescriptors_impl(const SDropDescriptorSetsParams& params)
 {
     const auto& drops = params.drops;
-    if (getEnabledFeatures().nullDescriptor)
+    if (!getEnabledFeatures().nullDescriptor)
     {
         return;
     }

src/nbl/video/CVulkanPhysicalDevice.cpp

@@ -1564,7 +1564,8 @@ core::smart_refctd_ptr<ILogicalDevice> CVulkanPhysicalDevice::createLogicalDevic
 
         enableExtensionIfAvailable(VK_KHR_SHADER_NON_SEMANTIC_INFO_EXTENSION_NAME);
 
-        enableExtensionIfAvailable(VK_KHR_FRAGMENT_SHADER_BARYCENTRIC_EXTENSION_NAME);
+        VkPhysicalDeviceFragmentShaderBarycentricFeaturesKHR fragmentShaderBarycentricFeatures = { VK_STRUCTURE_TYPE_PHYSICAL_DEVICE_FRAGMENT_SHADER_BARYCENTRIC_FEATURES_KHR, nullptr};
+        enableExtensionIfAvailable(VK_KHR_FRAGMENT_SHADER_BARYCENTRIC_EXTENSION_NAME,&fragmentShaderBarycentricFeatures);
 
         VkPhysicalDeviceShaderSubgroupUniformControlFlowFeaturesKHR subgroupUniformControlFlowFeatures = { VK_STRUCTURE_TYPE_PHYSICAL_DEVICE_SHADER_SUBGROUP_UNIFORM_CONTROL_FLOW_FEATURES_KHR,nullptr };
         enableExtensionIfAvailable(VK_KHR_SHADER_SUBGROUP_UNIFORM_CONTROL_FLOW_EXTENSION_NAME,&subgroupUniformControlFlowFeatures);

src/nbl/video/IGPUDescriptorSet.cpp

@@ -174,10 +174,10 @@ void IGPUDescriptorSet::processWrite(const IGPUDescriptorSet::SWriteDescriptorSe
 
     for (auto j = 0; j < write.count; ++j)
     {
-        descriptors[j] = write.info[j].desc;
+        descriptors[j + write.arrayElement] = write.info[j].desc;
 
         if (mutableSamplers)
-            mutableSamplers[j] = write.info[j].info.combinedImageSampler.sampler;
+            mutableSamplers[j + write.arrayElement] = write.info[j].info.combinedImageSampler.sampler;
     }
     auto& bindingRedirect = m_layout->getDescriptorRedirect(validationResult.type);
     auto bindingCreateFlags = bindingRedirect.getCreateFlags(validationResult.descriptorRedirectBindingIndex);