Trait RuntimeBackend

pub trait RuntimeBackend {
    // Required methods
    fn is_evaluated(&self, x: Id) -> bool;
    fn is_free_id(&self, x: Id) -> bool;
    fn remove(&mut self, x: Id) -> Result<(), ZyxError>;
    fn store<T: Scalar, IT>(&mut self, x: Id, iter: IT) -> Result<(), ZyxError>
       where IT: IntoIterator<Item = T>,
             IT::IntoIter: ExactSizeIterator;
    fn load<T: Scalar>(
        &mut self,
        x: Id,
        numel: usize,
    ) -> Result<Vec<T>, ZyxError>;
    fn evaluate(
        &mut self,
        rcs: BTreeMap<Id, u32>,
        order: &[Id],
        nodes: &[Node],
    ) -> Result<(), ZyxError>;
}

RuntimeBackend is a good plug in point for backend developers. Use Runtime::new(YourOwnStructThatImplementsRuntimeBackend::new()) to write your own backend which needs to implement only evaluation of graph. Used by torch and native backends.

Required Methods

fn is_evaluated(&self, x: Id) -> bool

Is tensor x evaluated?

fn is_free_id(&self, x: Id) -> bool

Check if there are no more buffers on id x

fn remove(&mut self, x: Id) -> Result<(), ZyxError>

Delete all memory used by tensor x.

fn store<T: Scalar, IT>(&mut self, x: Id, iter: IT) -> Result<(), ZyxError>

where IT: IntoIterator<Item = T>, IT::IntoIter: ExactSizeIterator,

Store iterator into runtime backend

fn load<T: Scalar>(&mut self, x: Id, numel: usize) -> Result<Vec<T>, ZyxError>

Load evaluated tensor x.

fn evaluate( &mut self, rcs: BTreeMap<Id, u32>, order: &[Id], nodes: &[Node], ) -> Result<(), ZyxError>

Evaluate tensors to_eval with given graph of nodes and recommended order of evaluation.

Dyn Compatibility

This trait is not dyn compatible.

In older versions of Rust, dyn compatibility was called "object safety", so this trait is not object safe.

Implementors