SIMD Parallelism #4
Description
Explicit SIMD can be used to improve Hexe's performance either in throughput or latency depending where it's used.
Architectural Components
Designing Position and State to use the cross-platform types found in std::simd. This would allow for batch position processing. This would also be a very cache-friendly approach.
Individual Components
There are numerous individual areas where parallelism with explicit SIMD can improve performance:
Hexe Main
- ???
Hexe Core
-
Piece::contained_in -
PieceMap-
is_empty -
len -
count_of
-
-
piece::map::{Iter, IterMut}::len -
piece::map::{Iter, IterMut}::next