Add advanced compilation chapter

ray_tracing_advanced_compilation/CMakeLists.txt

@@ -0,0 +1,78 @@
+#*****************************************************************************
+# Copyright 2020 NVIDIA Corporation. All rights reserved.
+#*****************************************************************************
+
+cmake_minimum_required(VERSION 3.9.6 FATAL_ERROR)
+
+#--------------------------------------------------------------------------------------------------
+# Project setting
+get_filename_component(PROJNAME ${CMAKE_CURRENT_SOURCE_DIR} NAME)
+set(PROJNAME vk_${PROJNAME}_KHR)
+project(${PROJNAME} LANGUAGES C CXX)
+message(STATUS "-------------------------------")
+message(STATUS "Processing Project ${PROJNAME}:")
+
+
+#--------------------------------------------------------------------------------------------------
+# C++ target and defines
+set(CMAKE_CXX_STANDARD 17)
+add_executable(${PROJNAME})
+_add_project_definitions(${PROJNAME})
+
+
+#--------------------------------------------------------------------------------------------------
+# Source files for this project
+#
+file(GLOB SOURCE_FILES *.cpp *.hpp *.inl *.h *.c)
+file(GLOB EXTRA_COMMON ${TUTO_KHR_DIR}/common/*.*)
+list(APPEND COMMON_SOURCE_FILES ${EXTRA_COMMON})
+include_directories(${TUTO_KHR_DIR}/common)
+
+
+#--------------------------------------------------------------------------------------------------
+# GLSL to SPIR-V custom build
+compile_glsl_directory(
+	SRC "${CMAKE_CURRENT_SOURCE_DIR}/shaders" 
+	DST "${CMAKE_CURRENT_SOURCE_DIR}/spv" 
+	VULKAN_TARGET "vulkan1.2"
+	)
+
+
+#--------------------------------------------------------------------------------------------------
+# Sources
+target_sources(${PROJNAME} PUBLIC ${SOURCE_FILES} ${HEADER_FILES})
+target_sources(${PROJNAME} PUBLIC ${COMMON_SOURCE_FILES})
+target_sources(${PROJNAME} PUBLIC ${PACKAGE_SOURCE_FILES})
+target_sources(${PROJNAME} PUBLIC ${GLSL_SOURCES} ${GLSL_HEADERS})
+
+
+#--------------------------------------------------------------------------------------------------
+# Sub-folders in Visual Studio
+#
+source_group("Common"       FILES ${COMMON_SOURCE_FILES} ${PACKAGE_SOURCE_FILES})
+source_group("Sources"      FILES ${SOURCE_FILES})
+source_group("Headers"      FILES ${HEADER_FILES})
+source_group("Shader_Files" FILES ${GLSL_SOURCES} ${GLSL_HEADERS})
+
+
+#--------------------------------------------------------------------------------------------------
+# Linkage
+#
+target_link_libraries(${PROJNAME} ${PLATFORM_LIBRARIES} nvpro_core)
+
+foreach(DEBUGLIB ${LIBRARIES_DEBUG})
+  target_link_libraries(${PROJNAME} debug ${DEBUGLIB})
+endforeach(DEBUGLIB)
+
+foreach(RELEASELIB ${LIBRARIES_OPTIMIZED})
+  target_link_libraries(${PROJNAME} optimized ${RELEASELIB})
+endforeach(RELEASELIB)
+
+#--------------------------------------------------------------------------------------------------
+# copies binaries that need to be put next to the exe files (ZLib, etc.)
+#
+_finalize_target( ${PROJNAME} )
+
+
+install(FILES ${SPV_OUTPUT} CONFIGURATIONS Release DESTINATION "bin_${ARCH}/${PROJNAME}/spv")
+install(FILES ${SPV_OUTPUT} CONFIGURATIONS Debug DESTINATION "bin_${ARCH}_debug/${PROJNAME}/spv")

ray_tracing_advanced_compilation/README.md

@@ -0,0 +1,248 @@
+# Advanced Compilation
+
+![](images/advanced_compilation.png)
+
+The simplest way of defining ray tracing pipelines is by using monolithic `VkRayTracingPipelineCreateInfoKHR` structures specifying the pipeline stages and shader groups. Even if pipelines share some stages, they are compiled separately. Furthermore, by default `vkCreateRayTracingPipelinesKHR` blocks the current thread until completion:
+![](images/high_level_regular_compilation.png)
+
+ However, when using multiple pipelines, some of them may have some shader stages in common. Large pipelines may also take a long time to compile on a single thread.
+
+This sample introduces the [VK_KHR_pipeline_library](https://www.khronos.org/registry/vulkan/specs/1.2-extensions/man/html/VK_KHR_pipeline_library.html) extension to create shader libraries that can be separately compiled and reused in ray tracing pipelines. The compilation of the final pipeline is carried out on multiple threads using the [VK_KHR_deferred_host_operations](https://www.khronos.org/registry/vulkan/specs/1.2-extensions/man/html/VK_KHR_deferred_host_operations.html) extension.
+
+The code below is based on the [`ray_tracing_specialization`](../ray_tracing_specialization) sample, which introduces compile-time variations of a hit shader. 
+
+
+## Pipeline Library
+
+
+Using monolithic pipeline definitions all stages are compiled for each pipeline, regardless of potential reuse. The shader groups are then referenced in the Shader Binding Table using their indices:
+![](images/regular_pipeline.png)
+
+Pipeline libraries are `VkPipeline` objects that cannot be bound directly. Instead, they can be compiled once and linked into as many pipelines as necessary. The Shader Binding Table of the resulting pipeline references the shader groups of the library as if they had been appended to the groups and stages in the main `VkRayTracingPipelineCreateInfo`
+![](images/library.png)
+
+ We start by adding a new member in the `HelloVulkan` class:
+~~~~ C
+  // Ray tracing shader library
+  VkPipeline                                        m_rtShaderLibrary;
+~~~~
+
+In `HelloVulkan::createRtPipeline()` the `StageIndices` enumeration describes the indices of the stages defined in the pipeline creation structure. The hit groups will be moved to our library, hence we remove them from the enumeration:
+~~~~ C
+  enum StageIndices
+  {
+    eRaygen,
+    eMiss,
+    eMiss2,
+    eShaderGroupCount
+  };
+~~~~
+
+The shader modules will be referenced partly in the main pipeline, and partly in the pipeline library. To ensure proper deletion of the modules after use, we will store their handles in
+~~~~ C
+  // Store the created modules for later cleanup
+  std::vector<VkShaderModule>                                    modules;
+~~~~
+Then, after each call to `nvvk::createShaderModule` we store the resulting module:
+~~~~ C
+  modules.push_back(stage.module);
+~~~~
+
+The specialization constants sample creates one shader module per specialization. Instead, we load that module once and reuse it for each specialization. Those specializations are then stored in the stages of the pipeline library:
+~~~~ C
+  // Hit Group - Closest Hit
+  // Create many variation of the closest hit
+  stage.module = nvvk::createShaderModule(
+    m_device, nvh::loadFile("spv/raytrace.rchit.spv", true, defaultSearchPaths, true));
+
+  modules.push_back(stage.module);
+  // Store the hit groups for compilation in a separate pipeline library object
+  std::vector<VkPipelineShaderStageCreateInfo> libraryStages{};
+  for(uint32_t s = 0; s < (uint32_t)specializations.size(); s++)
+  {
+    stage.stage               = VK_SHADER_STAGE_CLOSEST_HIT_BIT_KHR;
+    stage.pSpecializationInfo = specializations[s].getSpecialization();
+    libraryStages.push_back(stage);
+  }
+~~~~
+
+Similarly, the hit groups will be stored in the library by replacing the storage of the hit groups in `m_rtShaderGroups` by:
+~~~~ C
+  // Shader groups for the pipeline library containing the closest hit shaders
+  std::vector<VkRayTracingShaderGroupCreateInfoKHR> libraryShaderGroups;
+
+  VkRayTracingShaderGroupCreateInfoKHR libraryGroup{VK_STRUCTURE_TYPE_RAY_TRACING_SHADER_GROUP_CREATE_INFO_KHR};
+  libraryGroup.anyHitShader       = VK_SHADER_UNUSED_KHR;
+  libraryGroup.closestHitShader   = VK_SHADER_UNUSED_KHR;
+  libraryGroup.generalShader      = VK_SHADER_UNUSED_KHR;
+  libraryGroup.intersectionShader = VK_SHADER_UNUSED_KHR;
+  libraryGroup.type               = VK_RAY_TRACING_SHADER_GROUP_TYPE_TRIANGLES_HIT_GROUP_KHR;
+
+  // Hit Group - Closest Hit + AnyHit
+  // Creating many Hit groups, one for each specialization
+  for(uint32_t s = 0; s < (uint32_t)specializations.size(); s++)
+  {
+    // The indices of the stages are local to the pipeline library
+    libraryGroup.closestHitShader = s;  // Using variation of the closest hit
+    libraryShaderGroups.push_back(libraryGroup);
+  }
+~~~~
+It is important to note that the stage indices are local to the pipeline library, regardless of where they will be used in the final pipeline. Those indices will be later offset depending on the contents of the pipeline. 
+
+Once the groups and stages are defined we can create the pipeline library. After the creation of the ray tracing pipeline layout, we define the base of the library creation information:
+~~~~ C
+  // Creation of the pipeline library object
+  VkRayTracingPipelineCreateInfoKHR pipelineLibraryInfo{VK_STRUCTURE_TYPE_RAY_TRACING_PIPELINE_CREATE_INFO_KHR};
+  // Flag the object as a pipeline library, which is a specific object that cannot be used directly.
+  pipelineLibraryInfo.flags = VK_PIPELINE_CREATE_LIBRARY_BIT_KHR;
+  // Use the same layout as the target pipeline
+  pipelineLibraryInfo.layout = m_rtPipelineLayout;
+  // As for the interface the maximum recursion depth must also be consistent across the pipeline
+  pipelineLibraryInfo.maxPipelineRayRecursionDepth = 2;
+~~~~
+Pipeline libraries are technically independent from the pipeline they will be linked into. However, linking can only be achieved by enforcing strong consistency constraints, such as having the same pipeline layout and maximum recursion depth. If the recursion depth differs the compilation of the final pipeline will fail. 
+
+In addition, the pipeline libraries need to have the same pipeline interface. This interface defines the maximum amount of data passed across stages:
+~~~~ C
+  // Pipeline libraries need to define an interface, defined by the maximum hit attribute size (typically 2 for
+  // the built-in triangle intersector) and the maximum payload size (3 floating-point values in this sample).
+  // Pipeline libraries can be linked into a final pipeline only if their interface matches
+  VkRayTracingPipelineInterfaceCreateInfoKHR pipelineInterface{VK_STRUCTURE_TYPE_RAY_TRACING_PIPELINE_INTERFACE_CREATE_INFO_KHR};
+  pipelineInterface.maxPipelineRayHitAttributeSize = sizeof(nvmath::vec2f);
+  pipelineInterface.maxPipelineRayPayloadSize      = sizeof(nvmath::vec3f);
+  pipelineLibraryInfo.pLibraryInterface            = &pipelineInterface;
+~~~~
+
+Finally we provide the stage and shader groups information to the library creation information, and create the pipeline library in the same way as any other pipeline:
+~~~~ C
+  // Shader groups and stages for the library
+  pipelineLibraryInfo.groupCount = static_cast<uint32_t>(libraryShaderGroups.size());
+  pipelineLibraryInfo.pGroups    = libraryShaderGroups.data();
+  pipelineLibraryInfo.stageCount = static_cast<uint32_t>(libraryStages.size());
+  pipelineLibraryInfo.pStages    = libraryStages.data();
+
+  // Creation of the pipeline library
+  vkCreateRayTracingPipelinesKHR(m_device, {}, {}, 1, &pipelineLibraryInfo, nullptr, &m_rtShaderLibrary);
+~~~~
+The pipeline library is now created, but the application cannot run yet: we still need to indicate that the final pipeline will link with our library. Before calling `vkCreateRayTracingPipelinesKHR` for the final pipeline, we insert the reference to the library:
+~~~~ C
+  // The library will be linked into the final pipeline by specifying its handle and shared interface
+  VkPipelineLibraryCreateInfoKHR inputLibrary{VK_STRUCTURE_TYPE_PIPELINE_LIBRARY_CREATE_INFO_KHR};
+  inputLibrary.libraryCount         = 1;
+  inputLibrary.pLibraries           = &m_rtShaderLibrary;
+  rayPipelineInfo.pLibraryInfo      = &inputLibrary;
+  rayPipelineInfo.pLibraryInterface = &pipelineInterface;
+~~~~
+
+The pipeline is now built from the specified shader groups and stages as well as the library containing the hit groups. The groups and stages are linked together using indices, and indices are local to each library. To avoid collisions the pipeline creation will consider the stages of the libraries as if they had been appended to the stage list of the original `VkRayTracingPipelineCreateInfoKHR`, in the order in which the libraries are defined in `pLibraries`.
+
+Therefore, the Shader Binding Table needs to be updated accordingly, by making the wrapper aware of the contents of the library:
+~~~~ C
+  // The Shader Binding Table is built accounting for the entire pipeline, including the
+  // stages contained in the library. Passing the library information allows the wrapper
+  // to shift the shader group indices accordingly
+  m_sbtWrapper.create(m_rtPipeline, rayPipelineInfo, {pipelineLibraryInfo});
+~~~~
+
+At the end of the function we destroy the shader modules using our vector of modules instead of iterating over the stages of the main pipeline:
+~~~~ C
+  // Destroy all the created modules, for both libraries and main pipeline
+  for(auto& m : modules)
+    vkDestroyShaderModule(m_device, m, nullptr);
+~~~~
+
+The pipeline library has the same lifetime as the pipeline that uses it. The final step of this Section is the destruction of the library in the `HelloVulkan::destroy()` method:
+~~~~ C
+  // Pipeline libraries have the same lifetime as the pipelines that uses them
+  vkDestroyPipeline(m_device, m_rtShaderLibrary, nullptr);
+~~~~
+
+As an exercise, it is possible to create another library containing the other shader stages, and link those libraries together into the pipeline. 
+
+## Parallel Compilation Using Deferred Host Operations
+
+By default pipelines compilations are blocking the calling thread until completion:
+ ![Pipeline compilation on a single thread](images/single_threaded_compilation.png)
+
+Ray tracing pipelines are often complex, and can benefit from multithreaded compilation.  This can drastically reduce the compilation times of complex pipelines. The deferred host operations extension allows the work involved in `vkCreateRayTracingPipelinesKHR` to be split into multiple threads provided by the application:
+ ![Deferred Host Operations use app-provided threads to parallelize the compilation](images/deferred_host_operations.png)
+
+We start by including the support of C++ threading using `std::async` at the beginning of the source file:
+~~~~ C
+#include <future>
+~~~~
+
+In this sample we will distribute the compilation of the final ray tracing pipeline using a `VkDeferredOperation`, created just before calling `vkCreateRayTracingPipelinesKHR`:
+~~~~ C
+  // Deferred operations allow the driver to parallelize the pipeline compilation on several threads
+  // Create a deferred operation
+  VkDeferredOperationKHR deferredOperation;
+  VkResult               result = vkCreateDeferredOperationKHR(m_device, nullptr, &deferredOperation);
+  assert(result == VK_SUCCESS);
+~~~~
+
+Then we modify the pipeline creation to indicate we defer the operation:
+~~~~ C
+  // The pipeline creation is called with the deferred operation. Instead of blocking until
+  // the compilation is done, the call returns immediately
+  vkCreateRayTracingPipelinesKHR(m_device, deferredOperation, {}, 1, &rayPipelineInfo, nullptr, &m_rtPipeline);
+~~~~
+Instead of immediately launching the compilation and blocking execution until completion, this call will return immediately with value `VK_OPERATION_DEFERRED_KHR` if deferred operations are supported by the system. 
+
+Threading control is left to the application. Therefore, our application will allocate a number of threads for compilation:
+~~~~ C
+  // The compilation will be split into a maximum of 8 threads, or the maximum supported by the
+  // driver for that operation
+  uint32_t maxThreads{8};
+  uint32_t threadCount = std::min(vkGetDeferredOperationMaxConcurrencyKHR(m_device, deferredOperation), maxThreads);
+~~~~
+
+We then launch those threads using `std::async`:
+~~~~ C
+  std::vector<std::future<void>> joins;
+  for(uint32_t i = 0; i < threadCount; i++)
+  {
+    VkDevice device{m_device};
+    joins.emplace_back(std::async(std::launch::async, [device, deferredOperation]() {
+      // Wait until the thread has finished its work
+      VkResult result = vkDeferredOperationJoinKHR(device, deferredOperation);
+      // A return value of SUCCESS means the pipeline compilation is done.
+      // THREAD_DONE indicates that thread has no work to do for this task
+      // (e.g. the operation could not be split into that many threads)
+      // THREAD_IDLE indicates the thread has finished its task, but the overall pipeline
+      // compilation is not finished.
+      // In the last two cases, more work could be performed by the thread, such as waiting
+      // for another deferred operation
+      assert(result == VK_SUCCESS || result == VK_THREAD_DONE_KHR || result == VK_THREAD_IDLE_KHR);
+    }));
+  }
+~~~~
+Each thread executes a blocking function taking care of a subset of the compilation. When a thread has finished its task, the pipeline compilation may be complete (`VK_SUCCESS`) or there may be no more work for this thread. In this case one could consider executing more work using those threads, such as compiling another pipeline.
+
+Since there is only one pipeline to compile, we wait for all threads to finish and check whether the pipeline compilation succeeded:
+~~~~ C
+  // Wait for all threads to finish
+  for(auto& f : joins)
+  {
+    f.get();
+  }
+  // Once the deferred operation is complete, check for compilation success
+  result = vkGetDeferredOperationResultKHR(m_device, deferredOperation);
+  assert(result == VK_SUCCESS);
+~~~~
+
+Once the compilation is finished we can destroy the deferred operation:
+~~~~ C
+  // Destroy the deferred operation
+  vkDestroyDeferredOperationKHR(m_device, deferredOperation, nullptr);
+~~~~
+
+Congratulations! The ray tracing pipeline is now built using explicit stages and a pipeline library, and the final compilation is executed on multiple threads. As an exercise, the pipeline library described at the beginning of this tutorial can also be compiled in parallel. 
+This approach can be extended to compile multiple pipelines sharing some components using multiple threads:
+![](images/high_level_advanced_compilation.png)
+
+## References
+
+ * [VK_KHR_pipeline_library](https://www.khronos.org/registry/vulkan/specs/1.2-extensions/man/html/VK_KHR_pipeline_library.html) 
+ * [VK_KHR_deferred_host_operations](https://www.khronos.org/registry/vulkan/specs/1.2-extensions/man/html/VK_KHR_deferred_host_operations.html) 
+

ray_tracing_advanced_compilation/hello_vulkan.h

@@ -0,0 +1,159 @@
+/*
+ * Copyright (c) 2014-2021, NVIDIA CORPORATION.  All rights reserved.
+ *
+ * Licensed under the Apache License, Version 2.0 (the "License");
+ * you may not use this file except in compliance with the License.
+ * You may obtain a copy of the License at
+ *
+ *     http://www.apache.org/licenses/LICENSE-2.0
+ *
+ * Unless required by applicable law or agreed to in writing, software
+ * distributed under the License is distributed on an "AS IS" BASIS,
+ * WITHOUT WARRANTIES OR CONDITIONS OF ANY KIND, either express or implied.
+ * See the License for the specific language governing permissions and
+ * limitations under the License.
+ *
+ * SPDX-FileCopyrightText: Copyright (c) 2014-2021 NVIDIA CORPORATION
+ * SPDX-License-Identifier: Apache-2.0
+ */
+
+#pragma once
+
+#include "nvvk/appbase_vk.hpp"
+#include "nvvk/debug_util_vk.hpp"
+#include "nvvk/descriptorsets_vk.hpp"
+#include "nvvk/memallocator_dma_vk.hpp"
+#include "nvvk/resourceallocator_vk.hpp"
+
+// #VKRay
+#include "nvvk/raytraceKHR_vk.hpp"
+#include "nvvk/sbtwrapper_vk.hpp"
+
+//--------------------------------------------------------------------------------------------------
+// Simple rasterizer of OBJ objects
+// - Each OBJ loaded are stored in an `ObjModel` and referenced by a `ObjInstance`
+// - It is possible to have many `ObjInstance` referencing the same `ObjModel`
+// - Rendering is done in an offscreen framebuffer
+// - The image of the framebuffer is displayed in post-process in a full-screen quad
+//
+class HelloVulkan : public nvvk::AppBaseVk
+{
+public:
+  void setup(const VkInstance& instance, const VkDevice& device, const VkPhysicalDevice& physicalDevice, uint32_t queueFamily) override;
+  void createDescriptorSetLayout();
+  void createGraphicsPipeline();
+  void loadModel(const std::string& filename, nvmath::mat4f transform = nvmath::mat4f(1));
+  void updateDescriptorSet();
+  void createUniformBuffer();
+  void createSceneDescriptionBuffer();
+  void createTextureImages(const VkCommandBuffer& cmdBuf, const std::vector<std::string>& textures);
+  void updateUniformBuffer(const VkCommandBuffer& cmdBuf);
+  void onResize(int /*w*/, int /*h*/) override;
+  void destroyResources();
+  void rasterize(const VkCommandBuffer& cmdBuff);
+
+  // The OBJ model
+  struct ObjModel
+  {
+    uint32_t     nbIndices{0};
+    uint32_t     nbVertices{0};
+    nvvk::Buffer vertexBuffer;    // Device buffer of all 'Vertex'
+    nvvk::Buffer indexBuffer;     // Device buffer of the indices forming triangles
+    nvvk::Buffer matColorBuffer;  // Device buffer of array of 'Wavefront material'
+    nvvk::Buffer matIndexBuffer;  // Device buffer of array of 'Wavefront material'
+  };
+
+  // Instance of the OBJ
+  struct ObjInstance
+  {
+    uint32_t      objIndex{0};     // Reference to the `m_objModel`
+    uint32_t      txtOffset{0};    // Offset in `m_textures`
+    nvmath::mat4f transform{1};    // Position of the instance
+    nvmath::mat4f transformIT{1};  // Inverse transpose
+  };
+
+  // Information pushed at each draw call
+  struct ObjPushConstant
+  {
+    nvmath::vec3f lightPosition{10.f, 15.f, 8.f};
+    int           instanceId{0};  // To retrieve the transformation matrix
+    float         lightIntensity{100.f};
+    int           lightType{0};       // 0: point, 1: infinite
+    int           specialization{7};  // all in use
+  };
+  ObjPushConstant m_pushConstant;
+
+  // Array of objects and instances in the scene
+  std::vector<ObjModel>    m_objModel;
+  std::vector<ObjInstance> m_objInstance;
+
+  // Graphic pipeline
+  VkPipelineLayout            m_pipelineLayout;
+  VkPipeline                  m_graphicsPipeline;
+  nvvk::DescriptorSetBindings m_descSetLayoutBind;
+  VkDescriptorPool            m_descPool;
+  VkDescriptorSetLayout       m_descSetLayout;
+  VkDescriptorSet             m_descSet;
+
+  nvvk::Buffer               m_cameraMat;  // Device-Host of the camera matrices
+  nvvk::Buffer               m_sceneDesc;  // Device buffer of the OBJ instances
+  std::vector<nvvk::Texture> m_textures;   // vector of all textures of the scene
+
+
+  nvvk::ResourceAllocatorDma m_alloc;  // Allocator for buffer, images, acceleration structures
+  nvvk::DebugUtil            m_debug;  // Utility to name objects
+
+
+  // #Post
+  void createOffscreenRender();
+  void createPostPipeline();
+  void createPostDescriptor();
+  void updatePostDescriptorSet();
+  void drawPost(VkCommandBuffer cmdBuf);
+
+  nvvk::DescriptorSetBindings m_postDescSetLayoutBind;
+  VkDescriptorPool            m_postDescPool{VK_NULL_HANDLE};
+  VkDescriptorSetLayout       m_postDescSetLayout{VK_NULL_HANDLE};
+  VkDescriptorSet             m_postDescSet{VK_NULL_HANDLE};
+  VkPipeline                  m_postPipeline{VK_NULL_HANDLE};
+  VkPipelineLayout            m_postPipelineLayout{VK_NULL_HANDLE};
+  VkRenderPass                m_offscreenRenderPass{VK_NULL_HANDLE};
+  VkFramebuffer               m_offscreenFramebuffer{VK_NULL_HANDLE};
+  nvvk::Texture               m_offscreenColor;
+  nvvk::Texture               m_offscreenDepth;
+  VkFormat                    m_offscreenColorFormat{VK_FORMAT_R32G32B32A32_SFLOAT};
+  VkFormat                    m_offscreenDepthFormat{VK_FORMAT_X8_D24_UNORM_PACK32};
+
+  // #VKRay
+  void initRayTracing();
+  auto objectToVkGeometryKHR(const ObjModel& model);
+  void createBottomLevelAS();
+  void createTopLevelAS();
+  void createRtDescriptorSet();
+  void updateRtDescriptorSet();
+  void createRtPipeline();
+  void raytrace(const VkCommandBuffer& cmdBuf, const nvmath::vec4f& clearColor);
+
+
+  VkPhysicalDeviceRayTracingPipelinePropertiesKHR m_rtProperties{VK_STRUCTURE_TYPE_PHYSICAL_DEVICE_RAY_TRACING_PIPELINE_PROPERTIES_KHR};
+  nvvk::RaytracingBuilderKHR                      m_rtBuilder;
+  nvvk::DescriptorSetBindings                     m_rtDescSetLayoutBind;
+  VkDescriptorPool                                m_rtDescPool;
+  VkDescriptorSetLayout                           m_rtDescSetLayout;
+  VkDescriptorSet                                 m_rtDescSet;
+  std::vector<VkRayTracingShaderGroupCreateInfoKHR> m_rtShaderGroups;
+  VkPipelineLayout                                  m_rtPipelineLayout;
+  VkPipeline                                        m_rtPipeline;
+  nvvk::SBTWrapper                                  m_sbtWrapper;
+  // Ray tracing shader library
+  VkPipeline                                        m_rtShaderLibrary;
+
+  struct RtPushConstant
+  {
+    nvmath::vec4f clearColor;
+    nvmath::vec3f lightPosition;
+    float         lightIntensity{100.0f};
+    int           lightType{0};
+    int           specialization{7};
+  } m_rtPushConstants;
+};

ray_tracing_advanced_compilation/main.cpp

@@ -0,0 +1,308 @@
+/*
+ * Copyright (c) 2014-2021, NVIDIA CORPORATION.  All rights reserved.
+ *
+ * Licensed under the Apache License, Version 2.0 (the "License");
+ * you may not use this file except in compliance with the License.
+ * You may obtain a copy of the License at
+ *
+ *     http://www.apache.org/licenses/LICENSE-2.0
+ *
+ * Unless required by applicable law or agreed to in writing, software
+ * distributed under the License is distributed on an "AS IS" BASIS,
+ * WITHOUT WARRANTIES OR CONDITIONS OF ANY KIND, either express or implied.
+ * See the License for the specific language governing permissions and
+ * limitations under the License.
+ *
+ * SPDX-FileCopyrightText: Copyright (c) 2014-2021 NVIDIA CORPORATION
+ * SPDX-License-Identifier: Apache-2.0
+ */
+
+
+// ImGui - standalone example application for Glfw + Vulkan, using programmable
+// pipeline If you are new to ImGui, see examples/README.txt and documentation
+// at the top of imgui.cpp.
+
+#include <array>
+
+#include "backends/imgui_impl_glfw.h"
+#include "imgui.h"
+
+#include "hello_vulkan.h"
+#include "imgui/imgui_camera_widget.h"
+#include "nvh/cameramanipulator.hpp"
+#include "nvh/fileoperations.hpp"
+#include "nvpsystem.hpp"
+#include "nvvk/commands_vk.hpp"
+#include "nvvk/context_vk.hpp"
+
+
+//////////////////////////////////////////////////////////////////////////
+#define UNUSED(x) (void)(x)
+//////////////////////////////////////////////////////////////////////////
+
+// Default search path for shaders
+std::vector<std::string> defaultSearchPaths;
+
+
+// GLFW Callback functions
+static void onErrorCallback(int error, const char* description)
+{
+  fprintf(stderr, "GLFW Error %d: %s\n", error, description);
+}
+
+// Extra UI
+void renderUI(HelloVulkan& helloVk)
+{
+  ImGuiH::CameraWidget();
+  if(ImGui::CollapsingHeader("Light"))
+  {
+    ImGui::RadioButton("Point", &helloVk.m_pushConstant.lightType, 0);
+    ImGui::SameLine();
+    ImGui::RadioButton("Infinite", &helloVk.m_pushConstant.lightType, 1);
+
+    ImGui::SliderFloat3("Position", &helloVk.m_pushConstant.lightPosition.x, -20.f, 20.f);
+    ImGui::SliderFloat("Intensity", &helloVk.m_pushConstant.lightIntensity, 0.f, 150.f);
+  }
+
+  // Specialization
+  ImGui::SliderInt("Specialization", &helloVk.m_pushConstant.specialization, 0, 7);
+  int s = helloVk.m_pushConstant.specialization;
+  int a = ((s >> 2) % 2) == 1;
+  int b = ((s >> 1) % 2) == 1;
+  int c = ((s >> 0) % 2) == 1;
+  ImGui::Checkbox("Use Diffuse", (bool*)&a);
+  ImGui::Checkbox("Use Specular", (bool*)&b);
+  ImGui::Checkbox("Trace shadow", (bool*)&c);
+  helloVk.m_pushConstant.specialization = (a << 2) + (b << 1) + c;
+}
+
+//////////////////////////////////////////////////////////////////////////
+//////////////////////////////////////////////////////////////////////////
+//////////////////////////////////////////////////////////////////////////
+static int const SAMPLE_WIDTH  = 1280;
+static int const SAMPLE_HEIGHT = 720;
+
+
+//--------------------------------------------------------------------------------------------------
+// Application Entry
+//
+int main(int argc, char** argv)
+{
+  UNUSED(argc);
+
+  // Setup GLFW window
+  glfwSetErrorCallback(onErrorCallback);
+  if(!glfwInit())
+  {
+    return 1;
+  }
+  glfwWindowHint(GLFW_CLIENT_API, GLFW_NO_API);
+  GLFWwindow* window = glfwCreateWindow(SAMPLE_WIDTH, SAMPLE_HEIGHT, PROJECT_NAME, nullptr, nullptr);
+
+  // Setup camera
+  CameraManip.setWindowSize(SAMPLE_WIDTH, SAMPLE_HEIGHT);
+  CameraManip.setLookat(nvmath::vec3f(5, 4, -4), nvmath::vec3f(0, 1, 0), nvmath::vec3f(0, 1, 0));
+
+  // Setup Vulkan
+  if(!glfwVulkanSupported())
+  {
+    printf("GLFW: Vulkan Not Supported\n");
+    return 1;
+  }
+
+  // setup some basic things for the sample, logging file for example
+  NVPSystem system(PROJECT_NAME);
+
+  // Search path for shaders and other media
+  defaultSearchPaths = {
+      NVPSystem::exePath() + PROJECT_RELDIRECTORY,
+      NVPSystem::exePath() + PROJECT_RELDIRECTORY "..",
+      std::string(PROJECT_NAME),
+  };
+
+  // Requesting Vulkan extensions and layers
+  nvvk::ContextCreateInfo contextInfo;
+  contextInfo.setVersion(1, 2);
+  contextInfo.addInstanceLayer("VK_LAYER_LUNARG_monitor", true);
+  contextInfo.addInstanceExtension(VK_EXT_DEBUG_UTILS_EXTENSION_NAME, true);
+  contextInfo.addInstanceExtension(VK_KHR_SURFACE_EXTENSION_NAME);
+#ifdef _WIN32
+  contextInfo.addInstanceExtension(VK_KHR_WIN32_SURFACE_EXTENSION_NAME);
+#else
+  contextInfo.addInstanceExtension(VK_KHR_XLIB_SURFACE_EXTENSION_NAME);
+  contextInfo.addInstanceExtension(VK_KHR_XCB_SURFACE_EXTENSION_NAME);
+#endif
+  contextInfo.addInstanceExtension(VK_KHR_GET_PHYSICAL_DEVICE_PROPERTIES_2_EXTENSION_NAME);
+  contextInfo.addDeviceExtension(VK_KHR_SWAPCHAIN_EXTENSION_NAME);
+  contextInfo.addDeviceExtension(VK_KHR_DEDICATED_ALLOCATION_EXTENSION_NAME);
+  contextInfo.addDeviceExtension(VK_KHR_GET_MEMORY_REQUIREMENTS_2_EXTENSION_NAME);
+  contextInfo.addDeviceExtension(VK_EXT_DESCRIPTOR_INDEXING_EXTENSION_NAME);
+  contextInfo.addDeviceExtension(VK_EXT_SCALAR_BLOCK_LAYOUT_EXTENSION_NAME);
+
+  // #VKRay: Activate the ray tracing extension
+  VkPhysicalDeviceAccelerationStructureFeaturesKHR accelFeature{VK_STRUCTURE_TYPE_PHYSICAL_DEVICE_ACCELERATION_STRUCTURE_FEATURES_KHR};
+  contextInfo.addDeviceExtension(VK_KHR_ACCELERATION_STRUCTURE_EXTENSION_NAME, false, &accelFeature);
+  VkPhysicalDeviceRayTracingPipelineFeaturesKHR rtPipelineFeature{VK_STRUCTURE_TYPE_PHYSICAL_DEVICE_RAY_TRACING_PIPELINE_FEATURES_KHR};
+  contextInfo.addDeviceExtension(VK_KHR_RAY_TRACING_PIPELINE_EXTENSION_NAME, false, &rtPipelineFeature);
+  contextInfo.addDeviceExtension(VK_KHR_MAINTENANCE3_EXTENSION_NAME);
+  contextInfo.addDeviceExtension(VK_KHR_PIPELINE_LIBRARY_EXTENSION_NAME);
+  contextInfo.addDeviceExtension(VK_KHR_DEFERRED_HOST_OPERATIONS_EXTENSION_NAME);
+  contextInfo.addDeviceExtension(VK_KHR_BUFFER_DEVICE_ADDRESS_EXTENSION_NAME);
+
+  // Creating Vulkan base application
+  nvvk::Context vkctx{};
+  vkctx.initInstance(contextInfo);
+  // Find all compatible devices
+  auto compatibleDevices = vkctx.getCompatibleDevices(contextInfo);
+  assert(!compatibleDevices.empty());
+  // Use a compatible device
+  vkctx.initDevice(compatibleDevices[0], contextInfo);
+
+  // Create example
+  HelloVulkan helloVk;
+
+  // Window need to be opened to get the surface on which to draw
+  const VkSurfaceKHR surface = helloVk.getVkSurface(vkctx.m_instance, window);
+  vkctx.setGCTQueueWithPresent(surface);
+
+  helloVk.setup(vkctx.m_instance, vkctx.m_device, vkctx.m_physicalDevice, vkctx.m_queueGCT.familyIndex);
+  helloVk.createSwapchain(surface, SAMPLE_WIDTH, SAMPLE_HEIGHT);
+  helloVk.createDepthBuffer();
+  helloVk.createRenderPass();
+  helloVk.createFrameBuffers();
+
+  // Setup Imgui
+  helloVk.initGUI(0);  // Using sub-pass 0
+
+  // Creation of the example
+  helloVk.loadModel(nvh::findFile("media/scenes/Medieval_building.obj", defaultSearchPaths, true));
+  helloVk.loadModel(nvh::findFile("media/scenes/plane.obj", defaultSearchPaths, true));
+
+
+  helloVk.createOffscreenRender();
+  helloVk.createDescriptorSetLayout();
+  helloVk.createGraphicsPipeline();
+  helloVk.createUniformBuffer();
+  helloVk.createSceneDescriptionBuffer();
+  helloVk.updateDescriptorSet();
+
+  // #VKRay
+  helloVk.initRayTracing();
+  helloVk.createBottomLevelAS();
+  helloVk.createTopLevelAS();
+  helloVk.createRtDescriptorSet();
+  helloVk.createRtPipeline();
+
+  helloVk.createPostDescriptor();
+  helloVk.createPostPipeline();
+  helloVk.updatePostDescriptorSet();
+
+
+  nvmath::vec4f clearColor   = nvmath::vec4f(1, 1, 1, 1.00f);
+  bool          useRaytracer = true;
+
+
+  helloVk.setupGlfwCallbacks(window);
+  ImGui_ImplGlfw_InitForVulkan(window, true);
+
+  // Main loop
+  while(!glfwWindowShouldClose(window))
+  {
+    glfwPollEvents();
+    if(helloVk.isMinimized())
+      continue;
+
+    // Start the Dear ImGui frame
+    ImGui_ImplGlfw_NewFrame();
+    ImGui::NewFrame();
+
+
+    // Show UI window.
+    if(helloVk.showGui())
+    {
+      ImGuiH::Panel::Begin();
+      ImGui::ColorEdit3("Clear color", reinterpret_cast<float*>(&clearColor));
+      ImGui::Checkbox("Ray Tracer mode", &useRaytracer);  // Switch between raster and ray tracing
+
+      renderUI(helloVk);
+      ImGui::Text("Application average %.3f ms/frame (%.1f FPS)", 1000.0f / ImGui::GetIO().Framerate, ImGui::GetIO().Framerate);
+      ImGuiH::Control::Info("", "", "(F10) Toggle Pane", ImGuiH::Control::Flags::Disabled);
+      ImGuiH::Panel::End();
+    }
+
+    // Start rendering the scene
+    helloVk.prepareFrame();
+
+    // Start command buffer of this frame
+    auto                   curFrame = helloVk.getCurFrame();
+    const VkCommandBuffer& cmdBuf   = helloVk.getCommandBuffers()[curFrame];
+
+    VkCommandBufferBeginInfo beginInfo{VK_STRUCTURE_TYPE_COMMAND_BUFFER_BEGIN_INFO};
+    beginInfo.flags = VK_COMMAND_BUFFER_USAGE_ONE_TIME_SUBMIT_BIT;
+    vkBeginCommandBuffer(cmdBuf, &beginInfo);
+
+    // Updating camera buffer
+    helloVk.updateUniformBuffer(cmdBuf);
+
+    // Clearing screen
+    std::array<VkClearValue, 2> clearValues{};
+    clearValues[0].color        = {{clearColor[0], clearColor[1], clearColor[2], clearColor[3]}};
+    clearValues[1].depthStencil = {1.0f, 0};
+
+    // Offscreen render pass
+    {
+      VkRenderPassBeginInfo offscreenRenderPassBeginInfo{VK_STRUCTURE_TYPE_RENDER_PASS_BEGIN_INFO};
+      offscreenRenderPassBeginInfo.clearValueCount = 2;
+      offscreenRenderPassBeginInfo.pClearValues    = clearValues.data();
+      offscreenRenderPassBeginInfo.renderPass      = helloVk.m_offscreenRenderPass;
+      offscreenRenderPassBeginInfo.framebuffer     = helloVk.m_offscreenFramebuffer;
+      offscreenRenderPassBeginInfo.renderArea      = {{0, 0}, helloVk.getSize()};
+
+      // Rendering Scene
+      if(useRaytracer)
+      {
+        helloVk.raytrace(cmdBuf, clearColor);
+      }
+      else
+      {
+        vkCmdBeginRenderPass(cmdBuf, &offscreenRenderPassBeginInfo, VK_SUBPASS_CONTENTS_INLINE);
+        helloVk.rasterize(cmdBuf);
+        vkCmdEndRenderPass(cmdBuf);
+      }
+    }
+
+    // 2nd rendering pass: tone mapper, UI
+    {
+      VkRenderPassBeginInfo postRenderPassBeginInfo{VK_STRUCTURE_TYPE_RENDER_PASS_BEGIN_INFO};
+      postRenderPassBeginInfo.clearValueCount = 2;
+      postRenderPassBeginInfo.pClearValues    = clearValues.data();
+      postRenderPassBeginInfo.renderPass      = helloVk.getRenderPass();
+      postRenderPassBeginInfo.framebuffer     = helloVk.getFramebuffers()[curFrame];
+      postRenderPassBeginInfo.renderArea      = {{0, 0}, helloVk.getSize()};
+
+      // Rendering tonemapper
+      vkCmdBeginRenderPass(cmdBuf, &postRenderPassBeginInfo, VK_SUBPASS_CONTENTS_INLINE);
+      helloVk.drawPost(cmdBuf);
+      // Rendering UI
+      ImGui::Render();
+      ImGui_ImplVulkan_RenderDrawData(ImGui::GetDrawData(), cmdBuf);
+      vkCmdEndRenderPass(cmdBuf);
+    }
+
+    // Submit for display
+    vkEndCommandBuffer(cmdBuf);
+    helloVk.submitFrame();
+  }
+
+  // Cleanup
+  vkDeviceWaitIdle(helloVk.getDevice());
+
+  helloVk.destroyResources();
+  helloVk.destroy();
+  vkctx.deinit();
+
+  glfwDestroyWindow(window);
+  glfwTerminate();
+
+  return 0;
+}

ray_tracing_advanced_compilation/shaders/frag_shader.frag

@@ -0,0 +1,98 @@
+/*
+ * Copyright (c) 2019-2021, NVIDIA CORPORATION.  All rights reserved.
+ *
+ * Licensed under the Apache License, Version 2.0 (the "License");
+ * you may not use this file except in compliance with the License.
+ * You may obtain a copy of the License at
+ *
+ *     http://www.apache.org/licenses/LICENSE-2.0
+ *
+ * Unless required by applicable law or agreed to in writing, software
+ * distributed under the License is distributed on an "AS IS" BASIS,
+ * WITHOUT WARRANTIES OR CONDITIONS OF ANY KIND, either express or implied.
+ * See the License for the specific language governing permissions and
+ * limitations under the License.
+ *
+ * SPDX-FileCopyrightText: Copyright (c) 2019-2021 NVIDIA CORPORATION
+ * SPDX-License-Identifier: Apache-2.0
+ */
+ 
+#version 450
+#extension GL_ARB_separate_shader_objects : enable
+#extension GL_EXT_nonuniform_qualifier : enable
+#extension GL_GOOGLE_include_directive : enable
+#extension GL_EXT_scalar_block_layout : enable
+
+#include "wavefront.glsl"
+
+
+layout(push_constant) uniform shaderInformation
+{
+  vec3  lightPosition;
+  uint  instanceId;
+  float lightIntensity;
+  int   lightType;
+}
+pushC;
+
+// clang-format off
+// Incoming 
+//layout(location = 0) flat in int matIndex;
+layout(location = 1) in vec2 fragTexCoord;
+layout(location = 2) in vec3 fragNormal;
+layout(location = 3) in vec3 viewDir;
+layout(location = 4) in vec3 worldPos;
+// Outgoing
+layout(location = 0) out vec4 outColor;
+// Buffers
+layout(binding = 1, scalar) buffer MatColorBufferObject { WaveFrontMaterial m[]; } materials[];
+layout(binding = 2, scalar) buffer ScnDesc { sceneDesc i[]; } scnDesc;
+layout(binding = 3) uniform sampler2D[] textureSamplers;
+layout(binding = 4, scalar) buffer MatIndex { int i[]; } matIdx[];
+
+// clang-format on
+
+
+void main()
+{
+  // Object of this instance
+  int objId = scnDesc.i[pushC.instanceId].objId;
+
+  // Material of the object
+  int               matIndex = matIdx[nonuniformEXT(objId)].i[gl_PrimitiveID];
+  WaveFrontMaterial mat      = materials[nonuniformEXT(objId)].m[matIndex];
+
+  vec3 N = normalize(fragNormal);
+
+  // Vector toward light
+  vec3  L;
+  float lightIntensity = pushC.lightIntensity;
+  if(pushC.lightType == 0)
+  {
+    vec3  lDir     = pushC.lightPosition - worldPos;
+    float d        = length(lDir);
+    lightIntensity = pushC.lightIntensity / (d * d);
+    L              = normalize(lDir);
+  }
+  else
+  {
+    L = normalize(pushC.lightPosition - vec3(0));
+  }
+
+
+  // Diffuse
+  vec3 diffuse = computeDiffuse(mat, L, N);
+  if(mat.textureId >= 0)
+  {
+    int  txtOffset  = scnDesc.i[pushC.instanceId].txtOffset;
+    uint txtId      = txtOffset + mat.textureId;
+    vec3 diffuseTxt = texture(textureSamplers[nonuniformEXT(txtId)], fragTexCoord).xyz;
+    diffuse *= diffuseTxt;
+  }
+
+  // Specular
+  vec3 specular = computeSpecular(mat, viewDir, L, N);
+
+  // Result
+  outColor = vec4(lightIntensity * (diffuse + specular), 1);
+}

ray_tracing_advanced_compilation/shaders/passthrough.vert

@@ -0,0 +1,34 @@
+/*
+ * Copyright (c) 2019-2021, NVIDIA CORPORATION.  All rights reserved.
+ *
+ * Licensed under the Apache License, Version 2.0 (the "License");
+ * you may not use this file except in compliance with the License.
+ * You may obtain a copy of the License at
+ *
+ *     http://www.apache.org/licenses/LICENSE-2.0
+ *
+ * Unless required by applicable law or agreed to in writing, software
+ * distributed under the License is distributed on an "AS IS" BASIS,
+ * WITHOUT WARRANTIES OR CONDITIONS OF ANY KIND, either express or implied.
+ * See the License for the specific language governing permissions and
+ * limitations under the License.
+ *
+ * SPDX-FileCopyrightText: Copyright (c) 2019-2021 NVIDIA CORPORATION
+ * SPDX-License-Identifier: Apache-2.0
+ */
+ 
+#version 450
+layout (location = 0) out vec2 outUV;
+
+
+out gl_PerVertex
+{
+  vec4 gl_Position;
+};
+
+
+void main()
+{
+  outUV = vec2((gl_VertexIndex << 1) & 2, gl_VertexIndex & 2);
+  gl_Position = vec4(outUV * 2.0f - 1.0f, 1.0f, 1.0f);
+}

ray_tracing_advanced_compilation/shaders/post.frag

@@ -0,0 +1,37 @@
+/*
+ * Copyright (c) 2019-2021, NVIDIA CORPORATION.  All rights reserved.
+ *
+ * Licensed under the Apache License, Version 2.0 (the "License");
+ * you may not use this file except in compliance with the License.
+ * You may obtain a copy of the License at
+ *
+ *     http://www.apache.org/licenses/LICENSE-2.0
+ *
+ * Unless required by applicable law or agreed to in writing, software
+ * distributed under the License is distributed on an "AS IS" BASIS,
+ * WITHOUT WARRANTIES OR CONDITIONS OF ANY KIND, either express or implied.
+ * See the License for the specific language governing permissions and
+ * limitations under the License.
+ *
+ * SPDX-FileCopyrightText: Copyright (c) 2019-2021 NVIDIA CORPORATION
+ * SPDX-License-Identifier: Apache-2.0
+ */
+ 
+#version 450
+layout(location = 0) in vec2 outUV;
+layout(location = 0) out vec4 fragColor;
+
+layout(set = 0, binding = 0) uniform sampler2D noisyTxt;
+
+layout(push_constant) uniform shaderInformation
+{
+  float aspectRatio;
+}
+pushc;
+
+void main()
+{
+  vec2  uv    = outUV;
+  float gamma = 1. / 2.2;
+  fragColor   = pow(texture(noisyTxt, uv).rgba, vec4(gamma));
+}

ray_tracing_advanced_compilation/shaders/raycommon.glsl

@@ -0,0 +1,23 @@
+/*
+ * Copyright (c) 2019-2021, NVIDIA CORPORATION.  All rights reserved.
+ *
+ * Licensed under the Apache License, Version 2.0 (the "License");
+ * you may not use this file except in compliance with the License.
+ * You may obtain a copy of the License at
+ *
+ *     http://www.apache.org/licenses/LICENSE-2.0
+ *
+ * Unless required by applicable law or agreed to in writing, software
+ * distributed under the License is distributed on an "AS IS" BASIS,
+ * WITHOUT WARRANTIES OR CONDITIONS OF ANY KIND, either express or implied.
+ * See the License for the specific language governing permissions and
+ * limitations under the License.
+ *
+ * SPDX-FileCopyrightText: Copyright (c) 2019-2021 NVIDIA CORPORATION
+ * SPDX-License-Identifier: Apache-2.0
+ */
+
+struct hitPayload
+{
+  vec3 hitValue;
+};

ray_tracing_advanced_compilation/shaders/raytrace.rchit

@@ -0,0 +1,169 @@
+/*
+ * Copyright (c) 2019-2021, NVIDIA CORPORATION.  All rights reserved.
+ *
+ * Licensed under the Apache License, Version 2.0 (the "License");
+ * you may not use this file except in compliance with the License.
+ * You may obtain a copy of the License at
+ *
+ *     http://www.apache.org/licenses/LICENSE-2.0
+ *
+ * Unless required by applicable law or agreed to in writing, software
+ * distributed under the License is distributed on an "AS IS" BASIS,
+ * WITHOUT WARRANTIES OR CONDITIONS OF ANY KIND, either express or implied.
+ * See the License for the specific language governing permissions and
+ * limitations under the License.
+ *
+ * SPDX-FileCopyrightText: Copyright (c) 2019-2021 NVIDIA CORPORATION
+ * SPDX-License-Identifier: Apache-2.0
+ */
+
+#version 460
+#extension GL_EXT_ray_tracing : require
+#extension GL_EXT_nonuniform_qualifier : enable
+#extension GL_EXT_scalar_block_layout : enable
+#extension GL_GOOGLE_include_directive : enable
+#include "raycommon.glsl"
+#include "wavefront.glsl"
+
+hitAttributeEXT vec2 attribs;
+
+// clang-format off
+layout(location = 0) rayPayloadInEXT hitPayload prd;
+layout(location = 1) rayPayloadEXT bool isShadowed;
+
+layout(binding = 0, set = 0) uniform accelerationStructureEXT topLevelAS;
+layout(binding = 1, set = 1, scalar) buffer MatColorBufferObject { WaveFrontMaterial m[]; } materials[];
+layout(binding = 2, set = 1, scalar) buffer ScnDesc { sceneDesc i[]; } scnDesc;
+layout(binding = 3, set = 1) uniform sampler2D textureSamplers[];
+layout(binding = 4, set = 1)  buffer MatIndexColorBuffer { int i[]; } matIndex[];
+layout(binding = 5, set = 1, scalar) buffer Vertices { Vertex v[]; } vertices[];
+layout(binding = 6, set = 1) buffer Indices { uint i[]; } indices[];
+
+
+layout(constant_id = 0) const int USE_DIFFUSE = 1;
+layout(constant_id = 1) const int USE_SPECULAR = 1;
+layout(constant_id = 2) const int TRACE_SHADOW = 1;
+
+// clang-format on
+
+layout(push_constant) uniform Constants
+{
+  vec4  clearColor;
+  vec3  lightPosition;
+  float lightIntensity;
+  int   lightType;
+  int   specialization;
+}
+pushC;
+
+
+void main()
+{
+  // Object of this instance
+  uint objId = scnDesc.i[gl_InstanceCustomIndexEXT].objId;
+
+  // Indices of the triangle
+  ivec3 ind = ivec3(indices[nonuniformEXT(objId)].i[3 * gl_PrimitiveID + 0],   //
+                    indices[nonuniformEXT(objId)].i[3 * gl_PrimitiveID + 1],   //
+                    indices[nonuniformEXT(objId)].i[3 * gl_PrimitiveID + 2]);  //
+  // Vertex of the triangle
+  Vertex v0 = vertices[nonuniformEXT(objId)].v[ind.x];
+  Vertex v1 = vertices[nonuniformEXT(objId)].v[ind.y];
+  Vertex v2 = vertices[nonuniformEXT(objId)].v[ind.z];
+
+  const vec3 barycentrics = vec3(1.0 - attribs.x - attribs.y, attribs.x, attribs.y);
+
+  // Computing the normal at hit position
+  vec3 normal = v0.nrm * barycentrics.x + v1.nrm * barycentrics.y + v2.nrm * barycentrics.z;
+  // Transforming the normal to world space
+  normal = normalize(vec3(scnDesc.i[gl_InstanceCustomIndexEXT].transfoIT * vec4(normal, 0.0)));
+
+
+  // Computing the coordinates of the hit position
+  vec3 worldPos = v0.pos * barycentrics.x + v1.pos * barycentrics.y + v2.pos * barycentrics.z;
+  // Transforming the position to world space
+  worldPos = vec3(scnDesc.i[gl_InstanceCustomIndexEXT].transfo * vec4(worldPos, 1.0));
+
+  // Vector toward the light
+  vec3  L;
+  float lightIntensity = pushC.lightIntensity;
+  float lightDistance  = 100000.0;
+  // Point light
+  if(pushC.lightType == 0)
+  {
+    vec3 lDir      = pushC.lightPosition - worldPos;
+    lightDistance  = length(lDir);
+    lightIntensity = pushC.lightIntensity / (lightDistance * lightDistance);
+    L              = normalize(lDir);
+  }
+  else  // Directional light
+  {
+    L = normalize(pushC.lightPosition - vec3(0));
+  }
+
+  // Material of the object
+  int               matIdx = matIndex[nonuniformEXT(objId)].i[gl_PrimitiveID];
+  WaveFrontMaterial mat    = materials[nonuniformEXT(objId)].m[matIdx];
+
+
+  // Diffuse
+  vec3 diffuse = vec3(0);
+  if(USE_DIFFUSE == 1)
+  {
+    diffuse = computeDiffuse(mat, L, normal);
+    if(mat.textureId >= 0)
+    {
+      uint txtId    = mat.textureId + scnDesc.i[gl_InstanceCustomIndexEXT].txtOffset;
+      vec2 texCoord = v0.texCoord * barycentrics.x + v1.texCoord * barycentrics.y
+                      + v2.texCoord * barycentrics.z;
+      diffuse *= texture(textureSamplers[nonuniformEXT(txtId)], texCoord).xyz;
+    }
+  }
+
+  vec3  specular    = vec3(0);
+  float attenuation = 1;
+
+  // Tracing shadow ray only if the light is visible from the surface
+  if(dot(normal, L) > 0)
+  {
+    if(TRACE_SHADOW == 1)
+    {
+      float tMin   = 0.001;
+      float tMax   = lightDistance;
+      vec3  origin = gl_WorldRayOriginEXT + gl_WorldRayDirectionEXT * gl_HitTEXT;
+      vec3  rayDir = L;
+      uint  flags  = gl_RayFlagsTerminateOnFirstHitEXT | gl_RayFlagsOpaqueEXT
+                   | gl_RayFlagsSkipClosestHitShaderEXT;
+      isShadowed = true;
+      traceRayEXT(topLevelAS,  // acceleration structure
+                  flags,       // rayFlags
+                  0xFF,        // cullMask
+                  0,           // sbtRecordOffset
+                  0,           // sbtRecordStride
+                  1,           // missIndex
+                  origin,      // ray origin
+                  tMin,        // ray min range
+                  rayDir,      // ray direction
+                  tMax,        // ray max range
+                  1            // payload (location = 1)
+      );
+    }
+    else
+      isShadowed = false;
+
+    if(isShadowed)
+    {
+      attenuation = 0.3;
+    }
+    else
+    {
+      // Specular
+      if(USE_SPECULAR == 1)
+      {
+        specular = computeSpecular(mat, gl_WorldRayDirectionEXT, L, normal);
+      }
+    }
+  }
+
+  prd.hitValue = vec3(lightIntensity * attenuation * (diffuse + specular));
+}

ray_tracing_advanced_compilation/shaders/raytrace.rgen

@@ -0,0 +1,77 @@
+/*
+ * Copyright (c) 2019-2021, NVIDIA CORPORATION.  All rights reserved.
+ *
+ * Licensed under the Apache License, Version 2.0 (the "License");
+ * you may not use this file except in compliance with the License.
+ * You may obtain a copy of the License at
+ *
+ *     http://www.apache.org/licenses/LICENSE-2.0
+ *
+ * Unless required by applicable law or agreed to in writing, software
+ * distributed under the License is distributed on an "AS IS" BASIS,
+ * WITHOUT WARRANTIES OR CONDITIONS OF ANY KIND, either express or implied.
+ * See the License for the specific language governing permissions and
+ * limitations under the License.
+ *
+ * SPDX-FileCopyrightText: Copyright (c) 2019-2021 NVIDIA CORPORATION
+ * SPDX-License-Identifier: Apache-2.0
+ */
+
+#version 460
+#extension GL_EXT_ray_tracing : require
+#extension GL_GOOGLE_include_directive : enable
+#include "raycommon.glsl"
+
+layout(binding = 0, set = 0) uniform accelerationStructureEXT topLevelAS;
+layout(binding = 1, set = 0, rgba32f) uniform image2D image;
+
+layout(location = 0) rayPayloadEXT hitPayload prd;
+
+layout(binding = 0, set = 1) uniform CameraProperties
+{
+  mat4 view;
+  mat4 proj;
+  mat4 viewInverse;
+  mat4 projInverse;
+}
+cam;
+
+layout(push_constant) uniform Constants
+{
+  vec4  clearColor;
+  vec3  lightPosition;
+  float lightIntensity;
+  int   lightType;
+  int   specialization;
+}
+pushC;
+
+void main()
+{
+  const vec2 pixelCenter = vec2(gl_LaunchIDEXT.xy) + vec2(0.5);
+  const vec2 inUV        = pixelCenter / vec2(gl_LaunchSizeEXT.xy);
+  vec2       d           = inUV * 2.0 - 1.0;
+
+  vec4 origin    = cam.viewInverse * vec4(0, 0, 0, 1);
+  vec4 target    = cam.projInverse * vec4(d.x, d.y, 1, 1);
+  vec4 direction = cam.viewInverse * vec4(normalize(target.xyz), 0);
+
+  uint  rayFlags = gl_RayFlagsOpaqueEXT;
+  float tMin     = 0.001;
+  float tMax     = 10000.0;
+
+  traceRayEXT(topLevelAS,            // acceleration structure
+              rayFlags,              // rayFlags
+              0xFF,                  // cullMask
+              pushC.specialization,  // sbtRecordOffset
+              0,                     // sbtRecordStride
+              0,                     // missIndex
+              origin.xyz,            // ray origin
+              tMin,                  // ray min range
+              direction.xyz,         // ray direction
+              tMax,                  // ray max range
+              0                      // payload (location = 0)
+  );
+
+  imageStore(image, ivec2(gl_LaunchIDEXT.xy), vec4(prd.hitValue, 1.0));
+}

ray_tracing_advanced_compilation/shaders/raytrace.rmiss

@@ -0,0 +1,35 @@
+/*
+ * Copyright (c) 2019-2021, NVIDIA CORPORATION.  All rights reserved.
+ *
+ * Licensed under the Apache License, Version 2.0 (the "License");
+ * you may not use this file except in compliance with the License.
+ * You may obtain a copy of the License at
+ *
+ *     http://www.apache.org/licenses/LICENSE-2.0
+ *
+ * Unless required by applicable law or agreed to in writing, software
+ * distributed under the License is distributed on an "AS IS" BASIS,
+ * WITHOUT WARRANTIES OR CONDITIONS OF ANY KIND, either express or implied.
+ * See the License for the specific language governing permissions and
+ * limitations under the License.
+ *
+ * SPDX-FileCopyrightText: Copyright (c) 2019-2021 NVIDIA CORPORATION
+ * SPDX-License-Identifier: Apache-2.0
+ */
+ 
+#version 460
+#extension GL_EXT_ray_tracing : require
+#extension GL_GOOGLE_include_directive : enable
+#include "raycommon.glsl"
+
+layout(location = 0) rayPayloadInEXT hitPayload prd;
+
+layout(push_constant) uniform Constants
+{
+  vec4 clearColor;
+};
+
+void main()
+{
+  prd.hitValue = clearColor.xyz * 0.8;
+}

ray_tracing_advanced_compilation/shaders/raytraceShadow.rmiss

@@ -0,0 +1,28 @@
+/*
+ * Copyright (c) 2019-2021, NVIDIA CORPORATION.  All rights reserved.
+ *
+ * Licensed under the Apache License, Version 2.0 (the "License");
+ * you may not use this file except in compliance with the License.
+ * You may obtain a copy of the License at
+ *
+ *     http://www.apache.org/licenses/LICENSE-2.0
+ *
+ * Unless required by applicable law or agreed to in writing, software
+ * distributed under the License is distributed on an "AS IS" BASIS,
+ * WITHOUT WARRANTIES OR CONDITIONS OF ANY KIND, either express or implied.
+ * See the License for the specific language governing permissions and
+ * limitations under the License.
+ *
+ * SPDX-FileCopyrightText: Copyright (c) 2019-2021 NVIDIA CORPORATION
+ * SPDX-License-Identifier: Apache-2.0
+ */
+ 
+#version 460
+#extension GL_EXT_ray_tracing : require
+
+layout(location = 1) rayPayloadInEXT bool isShadowed;
+
+void main()
+{
+  isShadowed = false;
+}

ray_tracing_advanced_compilation/shaders/vert_shader.vert

@@ -0,0 +1,80 @@
+/*
+ * Copyright (c) 2019-2021, NVIDIA CORPORATION.  All rights reserved.
+ *
+ * Licensed under the Apache License, Version 2.0 (the "License");
+ * you may not use this file except in compliance with the License.
+ * You may obtain a copy of the License at
+ *
+ *     http://www.apache.org/licenses/LICENSE-2.0
+ *
+ * Unless required by applicable law or agreed to in writing, software
+ * distributed under the License is distributed on an "AS IS" BASIS,
+ * WITHOUT WARRANTIES OR CONDITIONS OF ANY KIND, either express or implied.
+ * See the License for the specific language governing permissions and
+ * limitations under the License.
+ *
+ * SPDX-FileCopyrightText: Copyright (c) 2019-2021 NVIDIA CORPORATION
+ * SPDX-License-Identifier: Apache-2.0
+ */
+ 
+#version 450
+#extension GL_ARB_separate_shader_objects : enable
+#extension GL_EXT_scalar_block_layout : enable
+#extension GL_GOOGLE_include_directive : enable
+
+#include "wavefront.glsl"
+
+// clang-format off
+layout(binding = 2, set = 0, scalar) buffer ScnDesc { sceneDesc i[]; } scnDesc;
+// clang-format on
+
+layout(binding = 0) uniform UniformBufferObject
+{
+  mat4 view;
+  mat4 proj;
+  mat4 viewI;
+}
+ubo;
+
+layout(push_constant) uniform shaderInformation
+{
+  vec3  lightPosition;
+  uint  instanceId;
+  float lightIntensity;
+  int   lightType;
+}
+pushC;
+
+layout(location = 0) in vec3 inPosition;
+layout(location = 1) in vec3 inNormal;
+layout(location = 2) in vec3 inColor;
+layout(location = 3) in vec2 inTexCoord;
+
+
+//layout(location = 0) flat out int matIndex;
+layout(location = 1) out vec2 fragTexCoord;
+layout(location = 2) out vec3 fragNormal;
+layout(location = 3) out vec3 viewDir;
+layout(location = 4) out vec3 worldPos;
+
+out gl_PerVertex
+{
+  vec4 gl_Position;
+};
+
+
+void main()
+{
+  mat4 objMatrix   = scnDesc.i[pushC.instanceId].transfo;
+  mat4 objMatrixIT = scnDesc.i[pushC.instanceId].transfoIT;
+
+  vec3 origin = vec3(ubo.viewI * vec4(0, 0, 0, 1));
+
+  worldPos     = vec3(objMatrix * vec4(inPosition, 1.0));
+  viewDir      = vec3(worldPos - origin);
+  fragTexCoord = inTexCoord;
+  fragNormal   = vec3(objMatrixIT * vec4(inNormal, 0.0));
+  //  matIndex     = inMatID;
+
+  gl_Position = ubo.proj * ubo.view * vec4(worldPos, 1.0);
+}

ray_tracing_advanced_compilation/shaders/wavefront.glsl

@@ -0,0 +1,77 @@
+/*
+ * Copyright (c) 2019-2021, NVIDIA CORPORATION.  All rights reserved.
+ *
+ * Licensed under the Apache License, Version 2.0 (the "License");
+ * you may not use this file except in compliance with the License.
+ * You may obtain a copy of the License at
+ *
+ *     http://www.apache.org/licenses/LICENSE-2.0
+ *
+ * Unless required by applicable law or agreed to in writing, software
+ * distributed under the License is distributed on an "AS IS" BASIS,
+ * WITHOUT WARRANTIES OR CONDITIONS OF ANY KIND, either express or implied.
+ * See the License for the specific language governing permissions and
+ * limitations under the License.
+ *
+ * SPDX-FileCopyrightText: Copyright (c) 2019-2021 NVIDIA CORPORATION
+ * SPDX-License-Identifier: Apache-2.0
+ */
+
+struct Vertex
+{
+  vec3 pos;
+  vec3 nrm;
+  vec3 color;
+  vec2 texCoord;
+};
+
+struct WaveFrontMaterial
+{
+  vec3  ambient;
+  vec3  diffuse;
+  vec3  specular;
+  vec3  transmittance;
+  vec3  emission;
+  float shininess;
+  float ior;       // index of refraction
+  float dissolve;  // 1 == opaque; 0 == fully transparent
+  int   illum;     // illumination model (see http://www.fileformat.info/format/material/)
+  int   textureId;
+};
+
+struct sceneDesc
+{
+  int  objId;
+  int  txtOffset;
+  mat4 transfo;
+  mat4 transfoIT;
+};
+
+
+vec3 computeDiffuse(WaveFrontMaterial mat, vec3 lightDir, vec3 normal)
+{
+  // Lambertian
+  float dotNL = max(dot(normal, lightDir), 0.0);
+  vec3  c     = mat.diffuse * dotNL;
+  if(mat.illum >= 1)
+    c += mat.ambient;
+  return c;
+}
+
+vec3 computeSpecular(WaveFrontMaterial mat, vec3 viewDir, vec3 lightDir, vec3 normal)
+{
+  if(mat.illum < 2)
+    return vec3(0);
+
+  // Compute specular only if not in shadow
+  const float kPi        = 3.14159265;
+  const float kShininess = max(mat.shininess, 4.0);
+
+  // Specular
+  const float kEnergyConservation = (2.0 + kShininess) / (2.0 * kPi);
+  vec3        V                   = normalize(-viewDir);
+  vec3        R                   = reflect(-lightDir, normal);
+  float       specular            = kEnergyConservation * pow(max(dot(V, R), 0.0), kShininess);
+
+  return vec3(mat.specular * specular);
+}