Struct rafx_framework::PooledResourceAllocator

pub struct PooledResourceAllocator<T: PooledResourceImpl> { /* fields omitted */ }

This handles waiting for N frames to pass before resetting the pool. "Restting" could mean different things depending on the resource. This allocator also has a callback for allocating new pools for use. A maximum pool count should be provided so that an unbounded leak of pools can be detected.

Implementations

impl<T: PooledResourceImpl> PooledResourceAllocator<T>

pub fn new<F: Fn(&RafxDeviceContext) -> RafxResult<T> + Send + Sync + 'static>(
    device_context: &RafxDeviceContext,
    max_in_flight_frames: u32,
    max_pool_count: u32,
    allocate_fn: F
) -> Self

Create a pool allocator that will reset resources after N frames. Keep in mind that if for example you want to push a single resource per frame, up to N+1 resources will exist in the sink. If max_in_flight_frames is 2, then you would have a resource that has likely not been submitted to the GPU yet, plus a resource per the N frames that have been submitted

pub fn allocate_pool(&mut self) -> RafxResult<T>

Allocate a pool - either reusing an old one that has been reset or creating a new one. Will assert that we do not exceed max_pool_count. The pool is allowed to exist until retire_pool is called. After this point, we will wait for N frames before restting it.

pub fn retire_pool(&mut self, pool: T)

Schedule the pool to reset after we complete N frames

pub fn update(&mut self) -> RafxResult<()>

Call when we are ready to reset another set of resources, most likely when a frame is presented or a new frame begins

pub fn destroy(&mut self) -> RafxResult<()>

Immediately destroy everything. We assume the device is idle and nothing is in flight. Calling this function when the device is not idle could result in a deadlock

Trait Implementations

impl<T: PooledResourceImpl> Drop for PooledResourceAllocator<T>

fn drop(&mut self)

Executes the destructor for this type.