Struct GpuDevicePool 

pub struct GpuDevicePool { /* private fields */ }

Pool of CUDA devices for multi-GPU operations

Manages multiple devices and provides round-robin scheduling.

Implementations

impl GpuDevicePool

pub fn new(device_count: usize) -> Result<Self>

Create a new device pool with the specified number of devices

pub fn device_count(&self) -> usize

Get the number of devices in the pool

pub fn get_device(&self, idx: usize) -> Option<&Arc<CudaDevice>>

Get a specific device by index

pub fn next_device_idx(&self) -> usize

Get the next device index using round-robin scheduling

pub fn next_device(&self) -> &Arc<CudaDevice>

Get the next device using round-robin scheduling

pub fn synchronize_all(&self) -> Result<()>

Synchronize all devices

pub fn devices(&self) -> &[Arc<CudaDevice>]

Get all devices