|
Diligent Engine
|
|
|
Diligent Engine
|
|
Diligent Core is a modern, cross-platform low-level graphics API that forms the foundation of Diligent Engine. It provides implementations for Direct3D11, Direct3D12, OpenGL, OpenGLES, Vulkan, and WebGPU rendering backends. A Metal backend is also available for commercial clients. In addition, the module includes essential platform-specific utilities. Diligent Core is fully self-contained and can be built independently of the rest of the engine. For details on supported platforms, features, and build instructions, please refer to the main repository.
| Platform | Build Status |
|---|---|
 Win32 |  |
 Universal Windows | |
 Linux |  |
 Android |  |
 MacOS |  |
 iOS | |
 tvOS | |
 Web |  |
To get the repository and all submodules, use the following command:
To build the module, see build instructions in the master repository.
Before you can use any functionality provided by the engine, you need to create a render device, an immediate context and a swap chain.
On Win32 platform, you can create OpenGL, Direct3D11, Direct3D12 or Vulkan device as shown below:
On Windows, the engine can be statically linked to the application or built as a separate DLL. In the first case, factory functions GetEngineFactoryOpenGL(), GetEngineFactoryD3D11(), GetEngineFactoryD3D12(), and GetEngineFactoryVk() can be called directly. In the second case, you need to load the DLL into the process's address space using LoadGraphicsEngineOpenGL(), LoadGraphicsEngineD3D11(), LoadGraphicsEngineD3D12(), or LoadGraphicsEngineVk() function. Each function loads appropriate dynamic library and imports the functions required to initialize the engine. You need to include the following headers:
You also need to add the following directories to the include search paths:
DiligentCore/Graphics/GraphicsEngineD3D11/interfaceDiligentCore/Graphics/GraphicsEngineD3D12/interfaceDiligentCore/Graphics/GraphicsEngineOpenGL/interfaceDiligentCore/Graphics/GraphicsEngineVulkan/interfaceAs an alternative, you may only add the path to the root folder and then use include paths relative to it.
Enable Diligent namespace:
IEngineFactoryD3D11::CreateDeviceAndContextsD3D11(), IEngineFactoryD3D12::CreateDeviceAndContextsD3D12(), and IEngineFactoryVk::CreateDeviceAndContextsVk() functions can also create a specified number of immediate and deferred contexts, which can be used for asynchronous rendering and multi-threaded command recording. The contexts may only be created during the initialization of the engine. The function populates an array of pointers to the contexts, where the immediates contexts go first, followed by all deferred contexts.
For more details, take a look at Tutorial00_HelloWin32.cpp file.
On Universal Windows Platform, you can create Direct3D11 or Direct3D12 device. Initialization is performed the same way as on Win32 Platform. The difference is that you first create the render device and device contexts by calling IEngineFactoryD3D11::CreateDeviceAndContextsD3D11() or IEngineFactoryD3D12::CreateDeviceAndContextsD3D12(). The swap chain is created later by a call to IEngineFactoryD3D11::CreateSwapChainD3D11() or IEngineFactoryD3D12::CreateSwapChainD3D12(). Please look at SampleAppUWP.cpp file for more details.
On Linux platform, the engine supports OpenGL and Vulkan backends. Initialization of GL context on Linux is tightly coupled with window creation. As a result, Diligent Engine does not initialize the context, but attaches to the one initialized by the app. An example of the engine initialization on Linux can be found in Tutorial00_HelloLinux.cpp.
On MacOS, Diligent Engine supports OpenGL, Vulkan and Metal backends. Initialization of GL context on MacOS is performed by the application, and the engine attaches to the context created by the app; see GLView.mm for details. Vulkan backend is initialized similar to other platforms. See MetalView.mm.
On Android, you can create OpenGLES or Vulkan device. The following code snippet shows an example:
If the engine is built as dynamic library, the library needs to be loaded by the native activity. The following code shows one possible way:
iOS implementation supports OpenGLES, Vulkan and Metal backend. Initialization of GL context on iOS is performed by the application, and the engine attaches to the context initialized by the app; see EAGLView.mm for details.
On the Web, you can create OpenGLES or WebGPU device. The following code snippet shows an example:
If you are using SDL or GLFW with existing context, you can provide null as the native window handle: EngineCI.Window = NativeWindow{nullptr}
The engine performs automatic reference counting and shuts down when the last reference to an engine object is released.
Device resources are created by the render device. The two main resource types are buffers, which represent linear memory, and textures, which use memory layouts optimized for fast filtering. To create a buffer, you need to populate BufferDesc structure and call IRenderDevice::CreateBuffer(). The following code creates a uniform (constant) buffer:
Similar, to create a texture, populate TextureDesc structure and call IRenderDevice::CreateTexture() as in the following example:
There is only one function CreateTexture() that is capable of creating all types of textures. Type, format, array size and all other parameters are specified by the members of the TextureDesc structure.
For every bind flag specified during the texture creation time, the texture object creates a default view. Default shader resource view addresses the entire texture, default render target and depth stencil views reference all array slices in the most detailed mip level, and unordered access view references the entire texture. To get a default view from the texture, use ITexture::GetDefaultView() function. Note that this function does not increment the reference counter of the returned interface. You can create additional texture views using ITexture::CreateView(). Use IBuffer::CreateView() to create additional views of a buffer.
To create a shader, populate ShaderCreateInfo structure:
There are three ways to create a shader. The first way is to provide a pointer to the shader source code through ShaderCreateInfo::Source member. The second way is to provide a file name. The third way is to provide a pointer to the compiled byte code through ShaderCreateInfo::ByteCode member. Graphics Engine is entirely decoupled from the platform. Since the host file system is platform-dependent, the structure exposes ShaderCreateInfo::pShaderSourceStreamFactory member that is intended to give the engine access to the file system. If you provided the source file name, you must also provide a non-null pointer to the shader source stream factory. If the shader source contains any #include directives, the source stream factory will also be used to load these files. The engine provides default implementation for every supported platform that should be sufficient in most cases. You can however define your own implementation.
An important member is ShaderCreateInfo::SourceLanguage. The following are valid values for this member:
SHADER_SOURCE_LANGUAGE_DEFAULT - The shader source format matches the underlying graphics API: HLSL for D3D11 or D3D12 mode, and GLSL for OpenGL, OpenGLES, and Vulkan modes.SHADER_SOURCE_LANGUAGE_HLSL - The shader source is in HLSL. For OpenGL and OpenGLES modes, the source code will be converted to GLSL. In Vulkan back-end, the code will be compiled to SPIRV directly.SHADER_SOURCE_LANGUAGE_GLSL - The shader source is in GLSL.SHADER_SOURCE_LANGUAGE_GLSL_VERBATIM - The shader source language is GLSL and should be compiled verbatim.SHADER_SOURCE_LANGUAGE_MSL - The source language is Metal Shading Language.Other members of the ShaderCreateInfo structure define the shader include search directories, shader macro definitions, shader entry point and other parameters.
When everything is ready, call IRenderDevice::CreateShader() to create the shader object:
Diligent Engine follows Direct3D12/Vulkan style to configure the graphics/compute pipeline. One monolithic Pipelines State Object (PSO) encompasses all required states (all shader stages, input layout description, depth stencil, rasterizer and blend state descriptions etc.). To create a graphics pipeline state object, define an instance of GraphicsPipelineStateCreateInfo structure:
Describe the pipeline specifics such as the number and format of render targets as well as depth-stencil format:
Initialize depth-stencil state description DepthStencilStateDesc. Note that the constructor initializes the members with default values and you may only set the ones that are different from default.
Initialize blend state description BlendStateDesc:
Initialize rasterizer state description RasterizerStateDesc:
Initialize input layout description InputLayoutDesc:
Define primitive topology and set shader pointers:
Pipeline resource layout informs the engine how the application is going to use different shader resource variables. To allow grouping of resources based on the expected frequency of resource bindings changes, Diligent Engine introduces classification of shader variables:
SHADER_RESOURCE_VARIABLE_TYPE_STATIC) are variables that are expected to be set only once. They may not be changed once a resource is bound to the variable. Such variables are intended to hold global constants such as camera attributes or global light attributes constant buffers. Note that it is the resource binding that may not change, while the contents of the resource is allowed to change according to its usage.SHADER_RESOURCE_VARIABLE_TYPE_MUTABLE) define resources that are expected to change on a per-material frequency. Examples may include diffuse textures, normal maps etc. Again updates to the contents of the resource are orthogobal to the binding changes.SHADER_RESOURCE_VARIABLE_TYPE_DYNAMIC) are expected to change frequently and randomly.To define variable types, prepare an array of ShaderResourceVariableDesc structures and initialize PSODesc.ResourceLayout.Variables and PSODesc.ResourceLayout.NumVariables members. Also PSODesc.ResourceLayout.DefaultVariableType can be used to set the type that will be used if a variable name is not provided.
When creating a pipeline state, textures can be permanently assigned immutable samplers. If an immutable sampler is assigned to a texture, it will always be used instead of the one initialized in the texture shader resource view. To define immutable samplers, prepare an array of ImmutableSamplerDesc structures and initialize PSODesc.ResourceLayout.ImmutableSamplers and PSODesc.ResourceLayout.NumImmutableSamplers members. Notice that immutable samplers can be assigned to a texture variable of any type, not necessarily static, so that the texture binding can be changed at run-time, while the sampler will stay immutable. It is highly recommended to use immutable samplers whenever possible.
This document provides a detailed information about working with texture samplers.
When all required fields of PSO description structure are set, call IRenderDevice::CreateGraphicsPipelineState() to create the PSO object:
As mentioned above, shader resource binding in Diligent Engine is based on grouping variables in 3 different groups (static, mutable and dynamic). Static variables are variables that are expected to be set only once. They may not be changed once a resource is bound to the variable. Such variables are intended to hold global constants such as camera attributes or global light attributes constant buffers. They are bound directly to the Pipeline State Object:
Mutable and dynamic variables are bound via a new object called Shader Resource Binding (SRB), which is created by the pipeline state (IPipelineState::CreateShaderResourceBinding()), or pipeline resource signature in advanced use cases:
The second parameter tells the system to initialize internal structures in the SRB object that reference static variables in the PSO.
Dynamic and mutable resources are then bound through the SRB object:
The difference between mutable and dynamic resources is that mutable resources can only be set once per instance of a shader resource binding. Dynamic resources can be set multiple times. It is important to properly set the variable type as this affects performance. Static and mutable variables are more efficient. Dynamic variables are more expensive and introduce some run-time overhead.
An alternative way to bind shader resources is to create an IResourceMapping interface that maps resource literal names to the actual resources:
The resource mapping can then be used to bind all static resources in a pipeline state (IPipelineState::BindStaticResources()):
or all mutable and dynamic resources in a shader resource binding (IShaderResourceBinding::BindResources()):
The last parameter to all BindResources() functions defines how resources should be resolved:
BIND_SHADER_RESOURCES_UPDATE_STATIC - Indicates that static variable bindings are to be updated.BIND_SHADER_RESOURCES_UPDATE_MUTABLE - Indicates that mutable variable bindings are to be updated.BIND_SHADER_RESOURCES_UPDATE_DYNAMIC -Indicates that dynamic variable bindings are to be updated.BIND_SHADER_RESOURCES_UPDATE_ALL - Indicates that all variable types (static, mutable and dynamic) are to be updated. Note that if none of BIND_SHADER_RESOURCES_UPDATE_STATIC, BIND_SHADER_RESOURCES_UPDATE_MUTABLE, and BIND_SHADER_RESOURCES_UPDATE_DYNAMIC flags are set, all variable types are updated as if BIND_SHADER_RESOURCES_UPDATE_ALL was specified.BIND_SHADER_RESOURCES_KEEP_EXISTING - If this flag is specified, only unresolved bindings will be updated. All existing bindings will keep their original values. If this flag is not specified, every shader variable will be updated if the mapping contains corresponding resource.BIND_SHADER_RESOURCES_VERIFY_ALL_RESOLVED - If this flag is specified, all shader bindings are expected be resolved after the call. If this is not the case, an error will be reported.BindResources() may be called several times with different resource mappings to bind resources. However, it is recommended to use one large resource mapping as the size of the mapping does not affect element search time.
The engine performs run-time checks to verify that correct resources are being bound. For example, if you try to bind a constant buffer to a shader resource view variable, an error will be output to the debug console.
Before any draw command can be invoked, all required vertex and index buffers as well as the pipeline state should be bound to the device context:
All methods that may need to perform resource state transitions take RESOURCE_STATE_TRANSITION_MODE enum as parameter. The enum defines the following modes:
RESOURCE_STATE_TRANSITION_MODE_NONE - Perform no resource state transitions.RESOURCE_STATE_TRANSITION_MODE_TRANSITION - Transition resources to the states required by the command.RESOURCE_STATE_TRANSITION_MODE_VERIFY - Do not transition, but verify that states are correct.The final step is to commit shader resources to the device context. This is accomplished by the IDeviceContext::CommitShaderResources() method:
If the method is not called, the engine will detect that resources are not committed and output a debug message. Note that the last parameter tells the system to transition resources to correct states. If this flag is not specified, the resources must be explicitly transitioned to required states by a call to IDeviceContext::TransitionShaderResources():
Note that the method requires pointer to the pipeline state that created the shader resource binding.
When all required states and resources are bound, IDeviceContext::DrawIndexed() can be used to execute a draw command or IDeviceContext::DispatchCompute() can be used to execute a compute command. Note that for a draw command, a graphics pipeline must be bound, and for a dispatch command, a compute pipeline must be bound. DrawIndexed() takes DrawIndexedAttribs structure as an argument, for example:
DRAW_FLAG_VERIFY_STATES flag instructs the engine to verify that vertex and index buffers used by the draw command are transitioned to proper states.
DispatchCompute() takes DispatchComputeAttribs structure that defines compute grid dimensions:
You can learn more about the engine API by studying samples and tutorials.
Diligent Engine extensively supports interoperability with underlying low-level APIs. The engine can be initialized by attaching to existing D3D11/D3D12 device or OpenGL/GLES context and provides access to the underlying native API objects. Refer to the following pages for more information:
Follow the following steps to build the NuGet package:
| Argument | Description | Required |
|---|---|---|
-c (configuration) | Native dynamic libraries build configuration (e.g. Debug, Release, etc.) | Yes |
-d (root-dir) | The path to the root directory of DiligentCore | Yes |
-s (settings) | The path to the settings file | No |
dotnet-tests | Flag indicating whether to run .NET tests | No |
dotnet-publish | Flag indicating whether to publish the package to NuGet Gallery | No |
free-memory | Use this argument if you encounter insufficient memory during the build process | No |
You can override the default settings using a settings file (check the default_settings dictionary in dotnet-build-package.py)
See Apache 2.0 license.
This project has some third-party dependencies, each of which may have independent licensing:
To contribute your code, submit a Pull Request to this repository. Diligent Engine is distributed under the Apache 2.0 license that guarantees that content in the DiligentCore repository is free of Intellectual Property encumbrances. In submitting any content to this repository, you license that content under the same terms, and you agree that the content is free of any Intellectual Property claims and you have the right to license it under those terms.
Diligent Engine uses clang-format to ensure consistent source code style throughout the code base. The format is validated by CI for each commit and pull request, and the build will fail if any code formatting issue is found. Please refer to this page for instructions on how to set up clang-format and automatic code formatting.
See Release History
