Compare the OpenGL ES 2.0 shader pipeline to Direct3D

[This article is for Windows 8.x and Windows Phone 8.x developers writing Windows Runtime apps. If you’re developing for Windows 10, see the latest documentation]

Conceptually, the Direct3D 11 shader pipeline is very similar to the one in OpenGL ES 2.0. In terms of API design, however, the major components for creating and managing the shader stages are parts of two primary interfaces, ID3D11Device1 and ID3D11DeviceContext1. This topic attempts to map common OpenGL ES 2.0 shader pipeline API patterns to the Direct3D 11 equivalents in these interfaces.

Reviewing the Direct3D 11 shader pipeline

The shader objects are created with methods on the ID3D11Device1 interface, such as ID3D11Device1::CreateVertexShader and ID3D11Device1::CreatePixelShader.

The Direct3D 11 graphics pipeline is managed by instances of the ID3D11DeviceContext1 interface, and has the following stages:

  • Input-Assembler Stage. The input-assembler stage supplies data (triangles, lines and points) to the pipeline. ID3D11DeviceContext1 methods that support this stage are prefixed with "IA".
  • Vertex-Shader Stage - The vertex-shader stage processes vertices, typically performing operations such as transformations, skinning, and lighting. A vertex shader always takes a single input vertex and produces a single output vertex. ID3D11DeviceContext1 methods that support this stage are prefixed with "VS".
  • Stream-Output Stage - The stream-output stage streams primitive data from the pipeline to memory on its way to the rasterizer. Data can be streamed out and/or passed into the rasterizer. Data streamed out to memory can be recirculated back into the pipeline as input data or read-back from the CPU. ID3D11DeviceContext1 methods that support this stage are prefixed with "SO".
  • Rasterizer Stage - The rasterizer clips primitives, prepares primitives for the pixel shader, and determines how to invoke pixel shaders. You can disable rasterization by telling the pipeline there is no pixel shader (set the pixel shader stage to NULL with ID3D11DeviceContext::PSSetShader), and disabling depth and stencil testing (set DepthEnable and StencilEnable to FALSE in D3D11_DEPTH_STENCIL_DESC). While disabled, rasterization-related pipeline counters will not update.
  • Pixel-Shader Stage - The pixel-shader stage receives interpolated data for a primitive and generates per-pixel data such as color. ID3D11DeviceContext1 methods that support this stage are prefixed with "PS".
  • Output-Merger Stage - The output-merger stage combines various types of output data (pixel shader values, depth and stencil information) with the contents of the render target and depth/stencil buffers to generate the final pipeline result. ID3D11DeviceContext1 methods that support this stage are prefixed with "OM".

(There are also stages for geometry shaders, hull shaders, tesselators, and domain shaders, but since they have no analogues in OpenGL ES 2.0, we won't discuss them here.)

For a complete list of the methods for these stages, refer to the ID3D11DeviceContext and ID3D11DeviceContext1 reference pages. ID3D11DeviceContext1 extends ID3D11DeviceContext for Direct3D 11.

Creating a shader

In Direct3D, shader resources are not created before compiling and loading them; rather, the resource is created when the HLSLis loaded. Therefore, there is no directly analogous function to glCreateShader, which creates an initialized shader resource of a specific type (such as GL_VERTEX_SHADER or GL_FRAGMENT_SHADER). Rather, shaders are created after the HLSL is loaded with specific functions like ID3D11Device1::CreateVertexShader and ID3D11Device1::CreatePixelShader, and which take the type and the compiled HLSL as parameters.

OpenGL ES 2.0 Direct3D 11
glCreateShader Call ID3D11Device1::CreateVertexShader and ID3D11Device1::CreatePixelShader after successfully loading the compiled shader object, passing them the CSO as a buffer.

 

Compiling a shader

Direct3D haders must be precompiled as Compiled Shader Object (.cso) files in Windows Store apps and loaded using one of the Windows Runtime file APIs. (Desktop apps can compile the shaders from text files or string at run-time.) The CSO files are built from any .hlsl files that are part of your Microsoft Visual Studio project, and retain the same names, only with a .cso file extension. Ensure that they are included with your package when you ship!

OpenGL ES 2.0 Direct3D 11
glCompileShader N/A. Compile the shaders to .cso files in Visual Studio and include them in your package.
Using glGetShaderiv for compile status N/A. See the compilation output from Visual Studio's FX Compiler (FXC) if there are errors in compilation. If compilation is successful, a corresponding CSO file is created.

 

Loading a shader

As noted in the section on creating a shader, Direct3D 11 creates the shader when the corresponding CSO file is loaded into a buffer and passed to one of the methods in the following table.

OpenGL ES 2.0 Direct3D 11
ShaderSource Call ID3D11Device1::CreateVertexShader and ID3D11Device1::CreatePixelShader after successfully loading the compiled shader object.

 

Setting up the pipeline

OpenGL ES 2.0 has the "shader program" object, which contains multiple shaders for execution. Individual shaders are attached to the shader program object. However, in Direct3D 11, you work with the rendering context (ID3D11DeviceContext1) directly and create shaders on it.

OpenGL ES 2.0 Direct3D 11
glCreateProgram N/A. Direct3D 11 does not use the shader program object abstraction.
glLinkProgram N/A. Direct3D 11 does not use the shader program object abstraction.
glUseProgram N/A. Direct3D 11 does not use the shader program object abstraction.
glGetProgramiv Use the reference you created to ID3D11DeviceContext1.

 

Create an instance of ID3D11DeviceContext1 and ID3D11Device1 with the static D3D11CreateDevice method.

Microsoft::WRL::ComPtr<ID3D11Device1>         m_d3dDevice;
Microsoft::WRL::ComPtr<ID3D11DeviceContext1>  m_d3dContext;

// ...

D3D11CreateDevice(
  nullptr, // Specify nullptr to use the default adapter.
  D3D_DRIVER_TYPE_HARDWARE,
  nullptr,
  creationFlags, // Set set debug and Direct2D compatibility flags.
  featureLevels, // List of feature levels this app can support.
  ARRAYSIZE(featureLevels),
  D3D11_SDK_VERSION, // Always set this to D3D11_SDK_VERSION for Windows Store apps.
  &device, // Returns the Direct3D device created.
  &m_featureLevel, // Returns feature level of device created.
  &m_d3dContext // Returns the device's immediate context.
);

Setting the viewport(s)

Setting a viewport in Direct3D 11 is very similar to how you set a viewport in OpenGL ES 2.0. In Direct3D 11, call ID3D11DeviceContext::RSSetViewports with a configured CD3D11_VIEWPORT.

Direct3D 11: Setting a viewport.

CD3D11_VIEWPORT viewport(
        0.0f,
        0.0f,
        m_d3dRenderTargetSize.Width,
        m_d3dRenderTargetSize.Height
        );
m_d3dContext->RSSetViewports(1, &viewport);
OpenGL ES 2.0 Direct3D 11
glViewport CD3D11_VIEWPORT, ID3D11DeviceContext::RSSetViewports

 

Configuring the vertex shaders

Configuring a vertex shader in Direct3D 11 is done when the shader is loaded. Uniforms are passed as constant buffers using ID3D11DeviceContext1::VSSetConstantBuffers1.

OpenGL ES 2.0 Direct3D 11
glAttachShader ID3D11Device1::CreateVertexShader
glGetShaderiv, glGetShaderSource ID3D11DeviceContext1::VSGetShader
glGetUniformfv, glGetUniformiv ID3D11DeviceContext1::VSGetConstantBuffers1.

 

Configuring the pixel shaders

Configuring a pixel shader in Direct3D 11 is done when the shader is loaded. Uniforms are passed as constant buffers using ID3D11DeviceContext1::PSSetConstantBuffers1.

OpenGL ES 2.0 Direct3D 11
glAttachShader ID3D11Device1::CreatePixelShader
glGetShaderiv, glGetShaderSource ID3D11DeviceContext1::PSGetShader
glGetUniformfv, glGetUniformiv ID3D11DeviceContext1::PSGetConstantBuffers1.

 

Generating the final results

When the pipeline completes, you draw the results of the shader stages into the back buffer. In Direct3D 11, just as it is with Open GL ES 2.0, this involves calling a draw command to output the results as a color map in the back buffer, and thensending that back buffer to the display.

OpenGL ES 2.0 Direct3D 11
glDrawElements ID3D11DeviceContext1::Draw, ID3D11DeviceContext1::DrawIndexed (or other Draw* methods on ID3D11DeviceContext1).
eglSwapBuffers IDXGISwapChain1::Present1

 

Porting GLSL to HLSL

GLSL and HLSL are not very different beyond complex type support and syntax some overall syntax. Many developers find it easiest to port between the two by aliasing common OpenGL ES 2.0 instructions and definitions to their HLSL equivalent. Note that Direct3D uses the Shader Model version to express the feature set of the HLSL supported by a graphics interface; OpenGL has a different version specification for HLSL. The following table attempts to give you some approximate idea of the shader language feature sets defined for Direct3D 11 and OpenGL ES 2.0 in the terms of the other's version.

Shader language GLSL feature version Direct3D Shader Model
Direct3D 11 HLSL ~4.30. SM 5.0
GLSL ES for OpenGL ES 2.0 1.40. Older implementations of GLSL ES for OpenGL ES 2.0 may use 1.10 through 1.30. Check your original code with glGetString(GL_SHADING_LANGUAGE_VERSION) or glGetString(SHADING_LANGUAGE_VERSION) to determine it. ~SM 2.0

 

For more details of differences between the two shader languages, as well as common syntax mappings, read the GLSL-to-HLSL reference.

Porting the OpenGL intrinsics to HLSL semantics

Direct3D 11 HLSL semantics are strings that, like a uniform or attribute name, are used to identify a value passed between the app and a shader program. While they can be any of a variety of possible strings, the best practice is to use a string like POSITION or COLOR that indicates the usage. You assign these semantics when you are constructing a constant buffer or buffer input layout. You can also append a number between 0 and 7 to the semantic so that you use separate registers for similar values. For example: COLOR0, COLOR1, COLOR2...

Semantics that are prefixed with "SV_" are system value semantics that are written to by your shader program; your app itself (running on the CPU) cannot modify them. Typically, these contain values that are inputs or outputs from another shader stage in the graphics pipeline, or are generated entirely by the GPU.

Additionally, SV_ semantics have different behaviors when they are used to specify input to or output from a shader stage. For example, SV_POSITION (output) contains the vertex data transformed during the vertex shader stage, and SV_POSITION (input) contains the pixel position values interpolated during rasterization.

Here are a few mappings for common OpenGL ES 2.0 shader instrinsics:

OpenGL system value Use this HLSL Semantic
gl_Position POSITION(n) for vertex buffer data. SV_POSITION provides a pixel position to the pixel shader and cannot be written by your app.,
gl_Normal NORMAL(n) for normal data provided by the vertex buffer.
gl_TexCoord[n] TEXCOORD(n) for texture UV (ST in some OpenGL documentation) coordinate data supplied to a shader.
gl_FragColor COLOR(n) for RGBA color data supplied to a shader. Note that it is treated identically to coordinate data; the semantic simply helps you identify that it is color data.
gl_FragData[n] SV_Target[n] for writing from a pixel shader to a target texture or other pixel buffer.

 

The method by which you code for semantics is not the same as using intrinsics in OpenGL ES 2.0. In OpenGL, you can access many of the intrinsics directly without any configuration or declaration; in Direct3D, you must declare a field in a specific constant buffer to use a particular semantic, or you declare it as the return value for a shader's main() method.

Here's an example of a semantic used in a constant buffer definition:

struct VertexShaderInput
{
  float3 pos : POSITION;
  float3 color : COLOR0;
};

// The position is interpolated to the pixel value by the system. The per-vertex color data is also interpolated and passed through the pixel shader. 
struct PixelShaderInput
{
  float4 pos : SV_POSITION;
  float3 color : COLOR0;
};

This code defines a pair of simple constant buffers

And here's an example of a semantic used to define the value returned by a fragment shader:

// A pass-through for the (interpolated) color data.
float4 main(PixelShaderInput input) : SV_TARGET
{
  return float4(input.color,1.0f);
}

In this case, SV_TARGET is the location of the render target that the pixel color (defined as a vector with four float values) is written to when the shader completes execution.

For more details on the use of semantics with Direct3D, read HLSL Semantics.