Module indirect_dispatch 

Indirect compute dispatch support for GPU-driven workloads.

Indirect dispatch allows the workgroup counts for a compute shader to be stored in a GPU-side buffer (written by a preceding GPU pass), eliminating the need for a CPU readback round-trip before issuing the next dispatch.

Key types

• DispatchIndirectArgs — layout-stable, 12-byte struct that maps directly onto the VkDispatchIndirectCommand / D3D12 / Metal equivalent.
• IndirectDispatchBuffer — a GPU buffer holding one or more dispatch argument slots, initialised from the CPU and updateable at any time.

Helper functions

• workgroup_count_1d / workgroup_count_2d / workgroup_count_3d — compute the ceiling-division workgroup counts for common dispatch dimensions without requiring a GPU context.
• args_for_elements — short-hand for building a 1-D dispatch args struct.
• dispatch_indirect_on_pass — convenience wrapper that sets the pipeline, bind groups, and issues the indirect dispatch in one call.

Structs

DispatchIndirectArgs

Layout-stable representation of workgroup dispatch dimensions stored in GPU memory.

IndirectDispatchBuffer

A GPU buffer that holds one or more DispatchIndirectArgs slots for use with wgpu::ComputePass::dispatch_workgroups_indirect.

Functions

args_for_elements

Build DispatchIndirectArgs for a 1-D dispatch over total elements with the given workgroup_size.

dispatch_indirect_on_pass

Set the pipeline, bind groups, and issue an indirect dispatch in one call.

workgroup_count_1d

Compute the ceiling-division workgroup count for a 1-D dispatch.

workgroup_count_2d

Compute the ceiling-division workgroup counts for a 2-D dispatch.

workgroup_count_3d

Compute the ceiling-division workgroup counts for a 3-D dispatch.