Merge pull request SaschaWillems#1224 from SaschaWillems/sync_rework_second_attempt

Rework synchronization

Author: SaschaWillems

README.md

@@ -85,10 +85,6 @@ Note that some examples require specific device features, and if you are on a mu
 
 Vulkan consumes shaders in an intermediate representation called SPIR-V. This makes it possible to use different shader languages by compiling them to that bytecode format. The primary shader language used here is [GLSL](shaders/glsl), most samples also come with [slang](shaders/slang/) and [HLSL](shaders/hlsl) shader sources, making it easy to compare the differences between those shading languages. The [Rust GPU](https://rust-gpu.github.io/) project maintains [Rust](https://www.rust-lang.org/) shader sources in a [separate repo](https://github.com/Rust-GPU/VulkanShaderExamples/tree/master/shaders/rust).
 
-## A note on synchronization
-
-Synchronization in the master branch currently isn't optimal und uses ```vkDeviceQueueWaitIdle``` at the end of each frame. This is a heavy operation and is suboptimal in regards to having CPU and GPU operations run in parallel. I'm currently reworking this in the [this branch](https://github.com/SaschaWillems/Vulkan/tree/sync_rework_second_attempt). While still work-in-progress, if you're interested in a more proper way of synchronization in Vulkan, please take a look at the [PR for that branch](https://github.com/SaschaWillems/Vulkan/pull/1224) to see progress.
-
 ## Examples
 
 ### Basics

base/VulkanBuffer.cpp

@@ -3,7 +3,7 @@
 *
 * Encapsulates a Vulkan buffer
 *
-* Copyright (C) 2016 by Sascha Willems - www.saschawillems.de
+* Copyright (C) 2016-2025 by Sascha Willems - www.saschawillems.de
 *
 * This code is licensed under the MIT license (MIT) (http://opensource.org/licenses/MIT)
 */
@@ -126,10 +126,12 @@ namespace vks
 		if (buffer)
 		{
 			vkDestroyBuffer(device, buffer, nullptr);
+			buffer = VK_NULL_HANDLE;
 		}
 		if (memory)
 		{
 			vkFreeMemory(device, memory, nullptr);
+			memory = VK_NULL_HANDLE;
 		}
 	}
 };

base/VulkanBuffer.h

@@ -3,7 +3,7 @@
 *
 * Encapsulates a Vulkan buffer
 *
-* Copyright (C) 2016 by Sascha Willems - www.saschawillems.de
+* Copyright (C) 2016-2025 by Sascha Willems - www.saschawillems.de
 *
 * This code is licensed under the MIT license (MIT) (http://opensource.org/licenses/MIT)
 */
@@ -34,6 +34,7 @@ namespace vks
 		VkBufferUsageFlags usageFlags;
 		/** @brief Memory property flags to be filled by external source at buffer creation (to query at some later point) */
 		VkMemoryPropertyFlags memoryPropertyFlags;
+		uint64_t deviceAddress;
 		VkResult map(VkDeviceSize size = VK_WHOLE_SIZE, VkDeviceSize offset = 0);
 		void unmap();
 		VkResult bind(VkDeviceSize offset = 0);

base/VulkanFrameBuffer.hpp

@@ -1,7 +1,7 @@
 /*
 * Vulkan framebuffer class
 *
-* Copyright (C) 2016 by Sascha Willems - www.saschawillems.de
+* Copyright (C) 2016-2025 by Sascha Willems - www.saschawillems.de
 *
 * This code is licensed under the MIT license (MIT) (http://opensource.org/licenses/MIT)
 */
@@ -304,23 +304,36 @@ namespace vks
 			}
 
 			// Use subpass dependencies for attachment layout transitions
-			std::array<VkSubpassDependency, 2> dependencies;
+			std::array<VkSubpassDependency, 4> dependencies{};
 
 			dependencies[0].srcSubpass = VK_SUBPASS_EXTERNAL;
 			dependencies[0].dstSubpass = 0;
-			dependencies[0].srcStageMask = VK_PIPELINE_STAGE_BOTTOM_OF_PIPE_BIT;
-			dependencies[0].dstStageMask = VK_PIPELINE_STAGE_COLOR_ATTACHMENT_OUTPUT_BIT;
-			dependencies[0].srcAccessMask = VK_ACCESS_MEMORY_READ_BIT;
-			dependencies[0].dstAccessMask = VK_ACCESS_COLOR_ATTACHMENT_READ_BIT | VK_ACCESS_COLOR_ATTACHMENT_WRITE_BIT;
-			dependencies[0].dependencyFlags = VK_DEPENDENCY_BY_REGION_BIT;
-
-			dependencies[1].srcSubpass = 0;
-			dependencies[1].dstSubpass = VK_SUBPASS_EXTERNAL;
-			dependencies[1].srcStageMask = VK_PIPELINE_STAGE_COLOR_ATTACHMENT_OUTPUT_BIT;
-			dependencies[1].dstStageMask = VK_PIPELINE_STAGE_BOTTOM_OF_PIPE_BIT;
-			dependencies[1].srcAccessMask = VK_ACCESS_COLOR_ATTACHMENT_READ_BIT | VK_ACCESS_COLOR_ATTACHMENT_WRITE_BIT;
-			dependencies[1].dstAccessMask = VK_ACCESS_MEMORY_READ_BIT;
-			dependencies[1].dependencyFlags = VK_DEPENDENCY_BY_REGION_BIT;
+			dependencies[0].srcStageMask = VK_PIPELINE_STAGE_EARLY_FRAGMENT_TESTS_BIT | VK_PIPELINE_STAGE_LATE_FRAGMENT_TESTS_BIT;
+			dependencies[0].dstStageMask = VK_PIPELINE_STAGE_EARLY_FRAGMENT_TESTS_BIT | VK_PIPELINE_STAGE_LATE_FRAGMENT_TESTS_BIT;
+			dependencies[0].srcAccessMask = VK_ACCESS_DEPTH_STENCIL_ATTACHMENT_WRITE_BIT;
+			dependencies[0].dstAccessMask = VK_ACCESS_DEPTH_STENCIL_ATTACHMENT_WRITE_BIT | VK_ACCESS_DEPTH_STENCIL_ATTACHMENT_READ_BIT;
+
+
+			dependencies[1].srcSubpass = VK_SUBPASS_EXTERNAL;
+			dependencies[1].dstSubpass = 0;
+			dependencies[1].srcStageMask = VK_PIPELINE_STAGE_BOTTOM_OF_PIPE_BIT;
+			dependencies[1].dstStageMask = VK_PIPELINE_STAGE_COLOR_ATTACHMENT_OUTPUT_BIT;
+			dependencies[1].srcAccessMask = VK_ACCESS_MEMORY_READ_BIT;
+			dependencies[1].dstAccessMask = VK_ACCESS_COLOR_ATTACHMENT_READ_BIT | VK_ACCESS_COLOR_ATTACHMENT_WRITE_BIT;
+
+			dependencies[2].srcSubpass = 0;
+			dependencies[2].dstSubpass = VK_SUBPASS_EXTERNAL;
+			dependencies[2].srcStageMask = VK_PIPELINE_STAGE_COLOR_ATTACHMENT_OUTPUT_BIT;
+			dependencies[2].dstStageMask = VK_PIPELINE_STAGE_BOTTOM_OF_PIPE_BIT;
+			dependencies[2].srcAccessMask = VK_ACCESS_COLOR_ATTACHMENT_READ_BIT | VK_ACCESS_COLOR_ATTACHMENT_WRITE_BIT;
+			dependencies[2].dstAccessMask = VK_ACCESS_MEMORY_READ_BIT;
+
+			dependencies[3].srcSubpass = 0;
+			dependencies[3].dstSubpass = VK_SUBPASS_EXTERNAL;
+			dependencies[3].srcStageMask = VK_PIPELINE_STAGE_LATE_FRAGMENT_TESTS_BIT;
+			dependencies[3].dstStageMask = VK_PIPELINE_STAGE_FRAGMENT_SHADER_BIT;
+			dependencies[3].srcAccessMask = VK_ACCESS_DEPTH_STENCIL_ATTACHMENT_WRITE_BIT;
+			dependencies[3].dstAccessMask = VK_ACCESS_SHADER_READ_BIT;
 
 			// Create render pass
 			VkRenderPassCreateInfo renderPassInfo = {};
@@ -329,19 +342,19 @@ namespace vks
 			renderPassInfo.attachmentCount = static_cast<uint32_t>(attachmentDescriptions.size());
 			renderPassInfo.subpassCount = 1;
 			renderPassInfo.pSubpasses = &subpass;
-			renderPassInfo.dependencyCount = 2;
+			renderPassInfo.dependencyCount = 4;
 			renderPassInfo.pDependencies = dependencies.data();
 			VK_CHECK_RESULT(vkCreateRenderPass(vulkanDevice->logicalDevice, &renderPassInfo, nullptr, &renderPass));
 
 			std::vector<VkImageView> attachmentViews;
-			for (auto attachment : attachments)
+			for (auto& attachment : attachments)
 			{
 				attachmentViews.push_back(attachment.view);
 			}
 
 			// Find. max number of layers across attachments
 			uint32_t maxLayers = 0;
-			for (auto attachment : attachments)
+			for (auto& attachment : attachments)
 			{
 				if (attachment.subresourceRange.layerCount > maxLayers)
 				{

base/VulkanInitializers.hpp

@@ -160,9 +160,7 @@ namespace vks
 
 		inline VkSubmitInfo submitInfo()
 		{
-			VkSubmitInfo submitInfo {};
-			submitInfo.sType = VK_STRUCTURE_TYPE_SUBMIT_INFO;
-			return submitInfo;
+			return { .sType = VK_STRUCTURE_TYPE_SUBMIT_INFO };
 		}
 
 		inline VkViewport viewport(

base/VulkanRaytracingSample.cpp

@@ -1,7 +1,7 @@
 /*
 * Extended sample base class for ray tracing based samples
 *
-* Copyright (C) 2020-2024 by Sascha Willems - www.saschawillems.de
+* Copyright (C) 2020-2025 by Sascha Willems - www.saschawillems.de
 *
 * This code is licensed under the MIT license (MIT) (http://opensource.org/licenses/MIT)
 */
@@ -139,6 +139,9 @@ void VulkanRaytracingSample::enableExtensions()
 
 	// Required by VK_KHR_spirv_1_4
 	enabledDeviceExtensions.push_back(VK_KHR_SHADER_FLOAT_CONTROLS_EXTENSION_NAME);
+
+	// Format for the storage image is decided at runtime, so we can't explicilily state it in the shader
+	enabledFeatures.shaderStorageImageWriteWithoutFormat = VK_TRUE;
 }
 
 VulkanRaytracingSample::ScratchBuffer VulkanRaytracingSample::createScratchBuffer(VkDeviceSize size)