Struct ParallelGradientComputer 

pub struct ParallelGradientComputer { /* private fields */ }

Parallel gradient computer using SciRS2-Core

Implementations

impl ParallelGradientComputer

pub fn new() -> Self

Create a new parallel gradient computer with default configuration

pub fn with_config(config: ParallelConfig) -> Self

Create with custom configuration

pub fn set_config(&mut self, config: ParallelConfig)

Set configuration

pub fn config(&self) -> &ParallelConfig

Get current configuration

pub fn stats(&self) -> &ParallelStats

Get statistics

pub fn reset_stats(&mut self)

Reset statistics

pub fn should_parallelize(&self, tensor_size: usize) -> bool

Check if a tensor should be processed in parallel

pub fn compute_optimal_chunk_size(&self, tensor_size: usize) -> usize

Compute optimal chunk size for a given tensor

pub fn compute_gradients_parallel<T>( &mut self, data: &[T], ) -> AutogradResult<Vec<T>>

where T: Send + Sync + Clone + Copy,

Compute gradients in parallel for multiple tensors

This uses SciRS2-Core’s parallel operations to distribute gradient computation across multiple CPU cores with optimal work distribution.

pub fn parallel_element_wise_op<T, F>( &mut self, data: &[T], op: F, ) -> AutogradResult<Vec<T>>

where T: Send + Sync + Clone, F: Fn(&T) -> T + Send + Sync,

Apply a parallel operation to gradient data

This demonstrates integration with scirs2-core’s parallel_ops for element-wise operations on gradient tensors.

pub fn compute_with_intelligent_chunking<T>( &mut self, data: &[T], grad_fn: impl Fn(&T) -> T + Send + Sync, ) -> AutogradResult<Vec<T>>

where T: Send + Sync + Clone,

Compute gradients using SciRS2’s intelligent chunking

This integrates with SciRS2-Core’s automatic performance optimization through intelligent chunking strategies.

pub fn report_performance(&self) -> String

Report current performance statistics

Trait Implementations

impl Default for ParallelGradientComputer

fn default() -> Self

Returns the “default value” for a type.