strided-basic

Shared typed CPU primitives and concrete copy, concatenation and reduction execution. Generic map/zip, SIMD, indexing plans and their validation stay with the traversal implementation. Concrete ordinary arithmetic/indexing dispatch lives in strided-kernel; runtime-DAG fusion lives in strided-fused. Neither of those packages is a dependency of this crate.

use strided_basic::{copy_into, Exec, StridedArray};
let src = StridedArray::<f64>::from_fn_col_major(&[2, 3], |i| (i[0] + 2*i[1]) as f64);
let mut dst = StridedArray::<f64>::col_major(&[2, 3]);
copy_into(&mut dst.view_mut(), &src.view(), Exec::serial()).unwrap();
assert_eq!(dst.get(&[1, 2]), 5.0);

copy_into_uninit copies into StridedViewMut<MaybeUninit<T>> without first initializing the destination. Success initializes every logical element, not unreachable holes.

Every operation that can fan out takes a threadteam::Exec placement token as its last parameter (re-exported as strided_basic::Exec, with Pool). Exec::serial() runs on the calling thread; Exec::new(&pool, budget) splits the work within budget slots of a pool the caller owns. The width is the token's decision: the serial time of the region (elements times the throughput of its execution::KernelClass) against the dispatch cost of the sections it opens, so a region too small to repay a fork stays on the calling thread whatever the budget. This crate starts no task of its own and never reaches a global pool. Narrow a token with Exec::limit to bound one operation more tightly than its pool.

simd is enabled by default. parallel lets pooled tokens run on a private pool; without it every token runs inline. Features select implementation capabilities, not operation families.

use strided_basic::{map_into, Exec, Pool, StridedArray};
let pool = Pool::builder().threads(2).build().unwrap();
let src = StridedArray::<f64>::from_fn_col_major(&[4], |i| i[0] as f64);
let mut dst = StridedArray::<f64>::col_major(&[4]);
map_into(&mut dst.view_mut(), &src.view(), |x| x + 1.0, Exec::new(&pool, 2).unwrap()).unwrap();
assert_eq!(dst.get(&[3]), 4.0);

The execution module is a documented low-level kernel-extension contract. Prevalidated entry points are unsafe, and a layout marker alone does not satisfy their requirements. Application code should use the checked root APIs.

See kernel boundaries and the packaged NOTICE / THIRD-PARTY-LICENSES for the Strided.jl lineage.

This new package has not yet been published; use the matching workspace checkout.