24343, "mppf", "Improve the performance of slices and rank change operations", "2024-02-06T15:05:32Z"
The performance of array slice and rank change operations is not what users expect. Users who are coming from other languages (at least Fortran and I think numpy) expect these slice operations to be fast but they have high overhead in Chapel today. I think we need to prioritize improving this situation as it has been a show-stopper for some potential users.
Related issues / posts:
- #8203
- #10763
- #22756
- Werid performance slow-down
- parallel processing - Is there a way to customize the default parallelization behavior of whole-array statements in Chapel? - Stack Overflow
- Array slicing performance in Chapel - Stack Overflow
Implementation ideas & workarounds:
- #5368
- #14530
- #16133
- #20749
- #23890
- https://github.com/Cray/chapel-private/issues/1382