Module parallel_iter 

Parallel iterators over slices and owned Vecs.

All functions in this module use OS threads (no Rayon dependency) and fall back to sequential execution for small inputs or when only one CPU is available.

Provided operations

Function	Description
parallel_map	Apply f to every element; preserves order.
parallel_reduce	Reduce with a commutative binary op; associativity required.
parallel_filter	Retain elements matching a predicate.
parallel_scan	Inclusive/exclusive prefix scan.
parallel_merge_sort	In-place parallel merge sort.
parallel_for_each	Execute a closure for each element (side-effects).
parallel_partition	Partition elements into two Vecs based on a predicate.
parallel_prefix_sum	Specialised f64 prefix-sum using Blelloch’s algorithm.

Example

use scirs2_core::concurrent::parallel_iter::{parallel_map, parallel_reduce, parallel_scan, ScanMode};

let data: Vec<i64> = (1..=8).collect();
let doubled = parallel_map(&data, |&x| x * 2, 0).expect("map");
assert_eq!(doubled, vec![2, 4, 6, 8, 10, 12, 14, 16]);

let sum = parallel_reduce(&data, 0i64, |a, b| a + b, |a, b| a + b, 0).expect("reduce");
assert_eq!(sum, 36);

let prefix = parallel_scan(&data, 0i64, |a, b| a + b, ScanMode::Inclusive, 0).expect("scan");
assert_eq!(prefix, vec![1, 3, 6, 10, 15, 21, 28, 36]);

Enums

ScanMode

Scan mode (inclusive or exclusive).

Functions

parallel_filter

Retain elements matching pred in parallel.

parallel_for_each

Execute f for each element of data in parallel (fire-and-forget, side-effects only). Order of execution is not guaranteed.

parallel_map

Apply f to every element of data in parallel, returning results in the same order as the input.

parallel_merge_sort

Parallel merge sort.

parallel_partition

Partition data into (matching, non_matching) in parallel.

parallel_prefix_sum

Specialised parallel prefix sum for f64 slices.

parallel_reduce

Parallel reduction using a chunk-local reduce followed by a sequential combine across chunks.

parallel_scan

Parallel prefix scan (generalised prefix sum) using Blelloch’s algorithm.

resolve_threads

Resolve n_threads to a concrete thread count (0 → hardware concurrency).