Documentation

Unity Engine


User Manual

Script Reference

Unity Engine


DispatchIndirect

Execute a compute shader.
Read time 2 minutesLast updated 5 days ago

Definition

  • Type: Method
  • Namespace: UnityEngine
  • Assembly: UnityEngine.CoreModule

DispatchIndirect(int, ComputeBuffer, uint)

Execute a compute shader.
public void DispatchIndirect(int kernelIndex, ComputeBuffer argsBuffer, uint argsOffset)

Parameters

kernelIndex

Which kernel to execute. A single compute shader asset can have multiple kernel entry points.

argsBuffer

Buffer with dispatch arguments.

argsOffset

The byte offset into the buffer, where the draw arguments start.

Remarks

This function "runs" the compute shader, with the given work size read directly from the GPU. Typical use case is generating arbitrary amount of data from a ComputeShader and then dispatching that, without requiring a readback to the CPU.
Buffer with arguments,
argsBuffer
, has to have three integer numbers at given
argsOffset
offset: number of work groups in X dimension, number of work groups in Y dimension, number of work groups in Z dimension.
Within each work group, a number of shader invocations ("threads") are done. The work group size is specified in the compute shader itself (using "numthreads" HLSL attribute), and the total amount of compute shader invocations is thus group count multiplied by the thread group size. Work group size can be queried using ComputeShader.GetKernelThreadGroupSizes function.
This very much maps to Direct3D11 DispatchIndirect, OpenGL glDispatchComputeIndirect and equivalent functions on other graphics APIs.

DispatchIndirect(int, GraphicsBuffer, uint)

Execute a compute shader.
public void DispatchIndirect(int kernelIndex, GraphicsBuffer argsBuffer, uint argsOffset)

Parameters

kernelIndex

Which kernel to execute. A single compute shader asset can have multiple kernel entry points.

argsBuffer

Buffer with dispatch arguments.

argsOffset

The byte offset into the buffer, where the draw arguments start.

Remarks

This function "runs" the compute shader, with the given work size read directly from the GPU. Typical use case is generating arbitrary amount of data from a ComputeShader and then dispatching that, without requiring a readback to the CPU.
Buffer with arguments,
argsBuffer
, has to have three integer numbers at given
argsOffset
offset: number of work groups in X dimension, number of work groups in Y dimension, number of work groups in Z dimension.
Within each work group, a number of shader invocations ("threads") are done. The work group size is specified in the compute shader itself (using "numthreads" HLSL attribute), and the total amount of compute shader invocations is thus group count multiplied by the thread group size. Work group size can be queried using ComputeShader.GetKernelThreadGroupSizes function.
This very much maps to Direct3D11 DispatchIndirect, OpenGL glDispatchComputeIndirect and equivalent functions on other graphics APIs.