Struct KernelInvocationBuilder

pub struct KernelInvocationBuilder<'a, 'b> { /* private fields */ }

A builder-pattern constructor for kernel invocations.

Implementations

impl<'a, 'b> KernelInvocationBuilder<'a, 'b>

pub fn new( queue: &'b mut KernelInvocationQueue<'a>, pipeline: &'a ComputePipeline, ) -> Self

Initiates the creation of a kernel invocation for the given compute pipeline.

Note that the invocation will not be queued into queue until KernelInvocationBuilder::queue or KernelInvocationBuilder::queue_indirect is called.

pub fn bind0<const INPUTS: usize>(self, inputs: [&Buffer; INPUTS]) -> Self

Binds INPUTS consecutive buffers to the bind group with id 0.

This method is less versatile than Self::bind and Self::bind_at but covers one of the most common cases. This will bind ipunts[i] to the storage binding i of bind group 0.

pub fn bind<const INPUTS: usize>( self, bind_group_id: u32, inputs: [&Buffer; INPUTS], ) -> Self

Binds INPUTS consecutive buffers to the bind group with id bind_group_id.

This method is more versatile than Self::bind0, but less than Self::bind_at. This will bind ipunts[i] to the storage binding i of bind group bind_group_id.

pub fn bind_at<const INPUTS: usize>( self, bind_group_id: u32, inputs: [(&Buffer, u32); INPUTS], ) -> Self

Binds INPUTS buffers with arbitrary storage binding ids, to the bind group with id bind_group_id.

This method is more versatile than Self::bind0, and Self::bind. This will bind inputs[i].0 to the storage binding inputs[i].1 of bind group bind_group_id.

pub fn queue(self, workgroups: impl WorkgroupSize)

Queues the kernel invocation into the queue that was given to KernelInvocationQueue::new.

The invocation will be configured with the given workgroups size (typically specified as a single u32 or a [u32; 3]).

pub fn queue_indirect(self, workgroups: Arc<Buffer>)

Queues the indirect kernel invocation into the queue that was given to KernelInvocationQueue::new.

The invocation will be configured with an indirect workgroups size specified with a Buffer that must contain exactly one instance of wgpu::util::DispatchIndirectArgs.