Re-add unifom buffer writing

src/application.cpp

@@ -169,9 +169,6 @@ namespace engine {
 				m_window->resetAvgFPS();
 			}
 
-			/* draw */
-			gfx()->finishRender(renderData.drawBuffer);
-
 			/* poll events */
 			m_window->getInputAndEvents();
 

src/gfx_device_vulkan.cpp

@@ -48,7 +48,7 @@ inline static void checkVulkanError(VkResult errorCode, int lineNo)
 
 namespace engine {
 
-	static constexpr uint32_t FRAMES_IN_FLIGHT = 1; // This improved FPS by 5x! (on Intel IGPU)
+	static constexpr uint32_t FRAMES_IN_FLIGHT = 2; // This improved FPS by 5x! (on Intel IGPU)
 
 	static constexpr size_t PUSH_CONSTANT_MAX_SIZE = 128; // bytes
 	static constexpr VkIndexType INDEX_TYPE = VK_INDEX_TYPE_UINT32;
@@ -517,7 +517,7 @@ namespace engine {
 		VKCHECK(res);
 		res = vkResetFences(pimpl->device.device, 1, &frameData.renderFence);
 		VKCHECK(res);
-#if 0
+
 		/* perform any pending uniform buffer writes */
 		VKCHECK(vkResetCommandPool(pimpl->device.device, frameData.transferPool, 0));
 
@@ -531,11 +531,26 @@ namespace engine {
 
 		// transfer cmds...
 
-		std::vector<VkBufferMemoryBarrier> barriers(pimpl->descriptorBufferWriteQueues[currentFrameIndex].size());
+		std::vector<VkBufferMemoryBarrier2> barriers{};
 		for (gfx::DescriptorBuffer* descriptorBuffer : pimpl->descriptorBufferWriteQueues[currentFrameIndex]) {
-			VkBufferMemoryBarrier& barrier = barriers.emplace_back();
+
-			barrier.sType = VK_STRUCTURE_TYPE_BUFFER_MEMORY_BARRIER;
+			VkBufferCopy copyRegion{};
-			barrier.srcAccessMask = VK_ACCESS_TRANSFER_WRITE_BIT;
+			copyRegion.srcOffset = 0;
+			copyRegion.dstOffset = 0;
+			copyRegion.size = descriptorBuffer->stagingBuffer.size;
+			vkCmdCopyBuffer(
+				frameData.transferBuf,
+				descriptorBuffer->stagingBuffer.buffer,
+				descriptorBuffer->gpuBuffers[currentFrameIndex].buffer,
+				1,
+				&copyRegion
+			);
+
+			VkBufferMemoryBarrier2& barrier = barriers.emplace_back();
+			barrier.sType = VK_STRUCTURE_TYPE_BUFFER_MEMORY_BARRIER_2;
+			barrier.srcStageMask = VK_PIPELINE_STAGE_2_TRANSFER_BIT;
+			barrier.srcAccessMask = VK_ACCESS_2_TRANSFER_WRITE_BIT;
+			barrier.dstStageMask = VK_PIPELINE_STAGE_2_TRANSFER_BIT;
 			barrier.dstAccessMask = 0;
 			barrier.srcQueueFamilyIndex = pimpl->device.queues.transferQueueFamily;
 			barrier.dstQueueFamilyIndex = pimpl->device.queues.presentAndDrawQueueFamily;
@@ -543,14 +558,13 @@ namespace engine {
 			barrier.offset = 0;
 			barrier.size = descriptorBuffer->gpuBuffers[currentFrameIndex].size;
 		}
+		pimpl->descriptorBufferWriteQueues[currentFrameIndex].clear();
 
-		vkCmdPipelineBarrier(frameData.transferBuf,
+		VkDependencyInfo dependencyInfo{};
-				VK_PIPELINE_STAGE_TRANSFER_BIT, VK_PIPELINE_STAGE_TRANSFER_BIT,
+		dependencyInfo.sType = VK_STRUCTURE_TYPE_DEPENDENCY_INFO;
-				0,
+		dependencyInfo.bufferMemoryBarrierCount = (uint32_t)barriers.size();
-				0, nullptr,
+		dependencyInfo.pBufferMemoryBarriers = barriers.data();
-				(uint32_t)barriers.size(), barriers.data(),
+		vkCmdPipelineBarrier2(frameData.transferBuf, &dependencyInfo);
-				0, nullptr
-			);
 
 		VKCHECK(vkEndCommandBuffer(frameData.transferBuf));
 
@@ -567,7 +581,7 @@ namespace engine {
 		}; 
 		res = vkQueueSubmit(pimpl->device.queues.transferQueues[0], 1, &transferSubmitInfo, VK_NULL_HANDLE);
 		assert(res == VK_SUCCESS);
-#endif
+
 		uint32_t swapchainImageIndex;
 
 		do {
@@ -600,6 +614,15 @@ namespace engine {
 
 		{ // RECORDING
 
+			/* change barriers to perform a queue ownership acquire operation */
+			for (VkBufferMemoryBarrier2& barrier : barriers) {
+				barrier.srcStageMask = VK_PIPELINE_STAGE_2_VERTEX_SHADER_BIT;
+				barrier.srcAccessMask = 0;
+				barrier.dstStageMask = VK_PIPELINE_STAGE_2_VERTEX_SHADER_BIT;
+				barrier.dstAccessMask = VK_ACCESS_2_UNIFORM_READ_BIT;
+			}
+			vkCmdPipelineBarrier2(frameData.drawBuf, &dependencyInfo);
+
 			std::array<VkClearValue, 2> clearValues{}; // Using same value for all components enables compression according to NVIDIA Best Practices
 			clearValues[0].color.float32[0] = 1.0f;
 			clearValues[0].color.float32[1] = 1.0f;
@@ -662,8 +685,8 @@ namespace engine {
 
 		waitSemaphores.push_back(drawBuffer->frameData.presentSemaphore);
 		waitDstStageMasks.push_back(VK_PIPELINE_STAGE_COLOR_ATTACHMENT_OUTPUT_BIT);
-		//waitSemaphores.push_back(drawBuffer->frameData.transferSemaphore);
+		waitSemaphores.push_back(drawBuffer->frameData.transferSemaphore);
-		//waitDstStageMasks.push_back(VK_PIPELINE_STAGE_VERTEX_SHADER_BIT);
+		waitDstStageMasks.push_back(VK_PIPELINE_STAGE_VERTEX_SHADER_BIT);
 
 		VkSubmitInfo submitInfo{
 			.sType = VK_STRUCTURE_TYPE_SUBMIT_INFO,

src/systems/render.cpp

@@ -110,6 +110,9 @@ namespace engine {
 			}
 		}
 
+		/* draw */
+		m_gfx->finishRender(renderData.drawBuffer);
+
 	}
 
 	void RenderSystem::setCameraEntity(uint32_t entity)

src/vulkan/device.cpp

@@ -51,10 +51,6 @@ namespace engine {
 			res = vkEnumerateDeviceExtensionProperties(physDev, nullptr, &extensionCount, availableExtensions.data());
 			assert(res == VK_SUCCESS);
 
-			for ([[maybe_unused]] const VkExtensionProperties& ext : availableExtensions) {
-				LOG_TRACE("extension: {}", ext.extensionName);
-			}
-
 			bool foundRequiredExtensions = true;
 			for (const char* extToFind : requirements.requiredExtensions) {
 				bool extFound = false;
@@ -82,9 +78,12 @@ namespace engine {
 			/* check features */
 			VkPhysicalDeviceMemoryPriorityFeaturesEXT memoryPriorityFeatures{};
 			memoryPriorityFeatures.sType = VK_STRUCTURE_TYPE_PHYSICAL_DEVICE_MEMORY_PRIORITY_FEATURES_EXT;
+			VkPhysicalDeviceSynchronization2Features synchronization2Features{};
+			synchronization2Features.sType = VK_STRUCTURE_TYPE_PHYSICAL_DEVICE_SYNCHRONIZATION_2_FEATURES;
+			synchronization2Features.pNext = &memoryPriorityFeatures;
 			VkPhysicalDeviceFeatures2 devFeatures{};
 			devFeatures.sType = VK_STRUCTURE_TYPE_PHYSICAL_DEVICE_FEATURES_2;
-			devFeatures.pNext = &memoryPriorityFeatures;
+			devFeatures.pNext = &synchronization2Features;
 			vkGetPhysicalDeviceFeatures2(physDev, &devFeatures);
 			{
 				if (requirements.requiredFeatures.robustBufferAccess)
@@ -198,6 +197,9 @@ namespace engine {
 				if (requirements.requiredFeatures.inheritedQueries == VK_TRUE)
 					if (devFeatures.features.inheritedQueries == VK_FALSE) continue;
 
+				/* ensure synchronization 2 is found */
+				if (synchronization2Features.synchronization2 == VK_FALSE) continue;
+				
 				/* check the memory priority extension was even requested */
 				bool memoryPriorityRequired = false;
 				for (const char* ext : requirements.requiredExtensions) {
@@ -343,12 +345,16 @@ namespace engine {
 		}
 
 		/* set enabled features */
-		VkPhysicalDeviceMemoryPriorityFeaturesEXT memoryPriorityFeaturesToEnable{};
+		VkPhysicalDeviceMemoryPriorityFeaturesEXT memoryPriorityFeatures{};
-		memoryPriorityFeaturesToEnable.sType = VK_STRUCTURE_TYPE_PHYSICAL_DEVICE_MEMORY_PRIORITY_FEATURES_EXT;
+		memoryPriorityFeatures.sType = VK_STRUCTURE_TYPE_PHYSICAL_DEVICE_MEMORY_PRIORITY_FEATURES_EXT;
-		memoryPriorityFeaturesToEnable.memoryPriority = d.memoryPriorityFeature ? VK_TRUE : VK_FALSE;
+		memoryPriorityFeatures.memoryPriority = d.memoryPriorityFeature ? VK_TRUE : VK_FALSE;
+		VkPhysicalDeviceSynchronization2Features synchronization2Features{};
+		synchronization2Features.sType = VK_STRUCTURE_TYPE_PHYSICAL_DEVICE_SYNCHRONIZATION_2_FEATURES;
+		synchronization2Features.pNext = &memoryPriorityFeatures;
+		synchronization2Features.synchronization2 = VK_TRUE;
 		VkPhysicalDeviceFeatures2 featuresToEnable{};
 		featuresToEnable.sType = VK_STRUCTURE_TYPE_PHYSICAL_DEVICE_FEATURES_2;
-		featuresToEnable.pNext = &memoryPriorityFeaturesToEnable;
+		featuresToEnable.pNext = &synchronization2Features;
 		featuresToEnable.features = requirements.requiredFeatures;
 
 		/* get list of extensions to enable */

src/vulkan/swapchain.cpp

@@ -183,9 +183,9 @@ namespace engine {
 				// Depth buffer
 				.srcSubpass = VK_SUBPASS_EXTERNAL,
 				.dstSubpass = 0,
-				.srcStageMask = VK_PIPELINE_STAGE_EARLY_FRAGMENT_TESTS_BIT | VK_PIPELINE_STAGE_LATE_FRAGMENT_TESTS_BIT,
+				.srcStageMask = VK_PIPELINE_STAGE_NONE,
 				.dstStageMask = VK_PIPELINE_STAGE_EARLY_FRAGMENT_TESTS_BIT | VK_PIPELINE_STAGE_LATE_FRAGMENT_TESTS_BIT,
-				.srcAccessMask = VK_ACCESS_DEPTH_STENCIL_ATTACHMENT_WRITE_BIT,
+				.srcAccessMask = 0,
 				.dstAccessMask = VK_ACCESS_DEPTH_STENCIL_ATTACHMENT_READ_BIT | VK_ACCESS_DEPTH_STENCIL_ATTACHMENT_WRITE_BIT,
 				.dependencyFlags = 0,
 			},
@@ -193,7 +193,7 @@ namespace engine {
 				// Image Layout Transition
 				.srcSubpass = VK_SUBPASS_EXTERNAL,
 				.dstSubpass = 0,
-				.srcStageMask = VK_PIPELINE_STAGE_COLOR_ATTACHMENT_OUTPUT_BIT,
+				.srcStageMask = VK_PIPELINE_STAGE_NONE,
 				.dstStageMask = VK_PIPELINE_STAGE_COLOR_ATTACHMENT_OUTPUT_BIT,
 				.srcAccessMask = 0,
 				.dstAccessMask = VK_ACCESS_COLOR_ATTACHMENT_WRITE_BIT | VK_ACCESS_COLOR_ATTACHMENT_READ_BIT,

test/src/camera_controller.cpp

@@ -3,6 +3,7 @@
 #include "application.hpp"
 #include "window.hpp"
 #include "input_manager.hpp"
+#include "scene_manager.hpp"
 #include "scene.hpp"
 
 #include "components/transform.hpp"
@@ -157,6 +158,11 @@ void CameraControllerSystem::onUpdate(float ts)
 		LOG_INFO("position: " + pos_string);
 	}
 
+	if (m_scene->app()->window()->getKeyPress(engine::inputs::Key::K_R)) {
+		t->position = { 0.0f, 5.0f, 0.0f };
+		c->dy = 0.0f;
+	}
+
 	if (m_scene->app()->inputManager()->getButtonPress("fullscreen")) {
 		m_scene->app()->window()->toggleFullscreen();
 	}

test/src/game.cpp

@@ -56,7 +56,7 @@ void playGame(GameSettings settings)
 	app.setFrameLimiter(settings.enableFrameLimiter);
 
 	// configure window
-	app.window()->setRelativeMouseMode(true);
+	//app.window()->setRelativeMouseMode(true);
 	
 	configureInputs(app.inputManager());
 
@@ -112,7 +112,7 @@ void playGame(GameSettings settings)
 	/* cube */
 	{
 		uint32_t cube = myScene->createEntity("cube");
-		myScene->getComponent<engine::TransformComponent>(cube)->position = glm::vec3{ -0.5f, -0.5f + 5.0f, -0.5f };
+		myScene->getComponent<engine::TransformComponent>(cube)->position = glm::vec3{ -0.5f + 5.0f, -0.5f + 5.0f, -0.5f + 5.0f };
 		auto cubeRenderable = myScene->addComponent<engine::RenderableComponent>(cube);
 		cubeRenderable->material = std::make_shared<engine::resources::Material>(app.getResource<engine::resources::Shader>("builtin.standard"));
 		cubeRenderable->mesh = genCuboidMesh(app.gfx(), 1.0f, 1.0f, 1.0f, 1);