Struct BufferPool 

pub struct BufferPool { /* private fields */ }

Reusable GPU buffer pool.

Implementations

impl BufferPool

pub fn new(device: Device, queue: Queue, config: &DispatchConfig) -> Self

docs

pub fn with_tiering( device: Device, queue: Queue, config: &DispatchConfig, tiers: Vec<CacheTier>, ) -> Result<Self, BackendError>

Opt-in hot/cold tiered caching on top of the power-of-two pool.

Returns a new BufferPool that shares the underlying device and queue but wraps every acquire/release in a TieredCache governed by the supplied tiers + policy. Consumers that batch many small dispatches (inference servers, streaming scanners, batched probes) use this to keep hot allocations resident and demote/evict cold ones via per-tier O(1) LRU.

The tiers vector is ordered coldest-first; policy controls promotion/eviction. Defaults at TieredCache::try_new(vec![CacheTier::try_new("hot", 1 << 24)?, CacheTier::try_new("cold", 1 << 30)?]) are a reasonable starting point for 16 MiB hot / 1 GiB cold.

pub fn queue(&self) -> &Queue

docs

pub fn device(&self) -> &Device

docs

pub fn acquire( &self, len: u64, usage: BufferUsages, ) -> Result<GpuBufferHandle, BackendError>

docs

pub fn release(&self, handle: GpuBufferHandle)

docs

pub fn stats(&self) -> BufferPoolStats

docs

Trait Implementations

impl Clone for BufferPool

fn clone(&self) -> BufferPool

Returns a duplicate of the value.

fn clone_from(&mut self, source: &Self)

Performs copy-assignment from source.

impl Debug for BufferPool

fn fmt(&self, f: &mut Formatter<'_>) -> Result

Formats the value using the given formatter.