Struct CpuPlacementBoundEager 

pub struct CpuPlacementBoundEager { /* private fields */ }

Placement-selected CPU view of one EagerRuntime.

The view snapshots the runtime’s CPU coordinator/provider bundle and the immutable runtime registration metadata when EagerRuntime::on_cpu is called. It holds no resource permit while idle and enters one backend session only while Self::with_eager_session runs. The session exposes core BackendSession operations on concrete Tensor values. This bridge deliberately does not expose the eager runtime’s linalg, FFT, einsum, or extension-runtime registries.

The value is intentionally not Clone: mutable use makes concurrent session ownership explicit without adding another backend mutex.

Examples

use tenferro_ad::EagerRuntime;
use tenferro_cpu::CpuPlacement;

let runtime = EagerRuntime::new()?;
let cpu = runtime.on_cpu(CpuPlacement::Auto)?;
assert_eq!(cpu.runtime_id(), runtime.id());

Implementations

impl CpuPlacementBoundEager

pub fn runtime_id(&self) -> ContextId

Return the identity of the original eager runtime.

Examples

use tenferro_ad::EagerRuntime;
use tenferro_cpu::CpuPlacement;

let runtime = EagerRuntime::new()?;
let cpu = runtime.on_cpu(CpuPlacement::Auto)?;
assert_eq!(cpu.runtime_id(), runtime.id());

pub fn placement(&self) -> CpuPlacement

Return the placement requested when this view was created.

Examples

use tenferro_ad::EagerRuntime;
use tenferro_cpu::CpuPlacement;

let runtime = EagerRuntime::new()?;
let cpu = runtime.on_cpu(CpuPlacement::Auto)?;
assert_eq!(cpu.placement(), CpuPlacement::Auto);

pub fn with_eager_session<T: Send, E: From<Error> + Send>( &mut self, f: impl FnOnce(&mut dyn BackendSession) -> Result<T, E> + Send, ) -> Result<T, E>

Enter one CPU backend session and run core operations through it.

One call creates one backend session. Tenferro-managed CPU executors enter once around the closure and core operations reuse that compatible execution scope. The closure may borrow stack data and need not be 'static.

This phase-2 bridge accepts only core BackendSession operations. It does not lock or dispatch the eager runtime’s linalg, FFT, einsum, or extension registries.

Examples

use tenferro_ad::{EagerRuntime, Error};
use tenferro_cpu::CpuPlacement;
use tenferro_tensor::{Tensor, TensorRead};

let runtime = EagerRuntime::new()?;
let mut cpu = runtime.on_cpu(CpuPlacement::Auto)?;
let lhs = Tensor::from_vec_col_major(vec![1], vec![1.0_f64])?;
let rhs = Tensor::from_vec_col_major(vec![1], vec![2.0_f64])?;
let output = cpu.with_eager_session(|session| {
    session
        .add_read(TensorRead::from_tensor(&lhs), TensorRead::from_tensor(&rhs))
        .map_err(Error::from)
})?;
assert_eq!(output.as_slice::<f64>().unwrap(), &[3.0]);

Errors

Returns the callback’s error unchanged. Core backend operations may report validation, unsupported capability, backend, or runtime-state failures through that error. Returns E::from(Error::SessionEntry) without running the callback when the backend cannot admit the session, for example when it is called from inside another session on this thread (tenferro_tensor::SessionEntryError::Reentered), and E::from the runtime-selection error when the CPU placement cannot be refreshed. Use only the borrowed session for work inside the scope.

Trait Implementations

impl Debug for CpuPlacementBoundEager

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

Formats the value using the given formatter.