Struct ManagedMemory 

pub struct ManagedMemory { /* private fields */ }

Backend-aware unified memory that routes allocation through the active backend

When a GPU backend is available, this allocates memory that is accessible from both host and device via the backend’s memory management. Falls back to host-only allocation when no GPU backend is present.

Implementations

impl ManagedMemory

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

Allocate managed memory (tries backend, falls back to host)

pub fn is_backend_registered(&self) -> bool

Check if memory is registered with a GPU backend

pub fn as_unified(&self) -> &UnifiedMemory

Get the underlying unified memory

pub fn as_unified_mut(&mut self) -> &mut UnifiedMemory

Get a mutable reference to the underlying unified memory

pub fn size(&self) -> usize

Get size

pub fn copy_from_slice(&mut self, data: &[u8]) -> Result<()>

Copy from host slice

pub fn copy_to_slice(&self, data: &mut [u8]) -> Result<()>

Copy to host slice

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

Prefetch to the device (hint for the runtime; no-op in CPU mode)

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

Prefetch to the host (hint for the runtime; no-op in CPU mode)