Major updates with PR 7:
- `OpenCilk` support
- Switched to `meson` build
Further details:
[OpenCilk]
- Replace Intel Cilk Plus array slice notation (aka vector extensions) with corresponding loops; fuse when possible.
- Remove utilities for dynamically setting of Cilk workers.
- Update build config to allow both OpenCilk and Intel Cilk Plus.
[Build]
- Add full-fledged Meson build scripts.
- Update `configure`/`make` scripts and fix some incorrect flags.
[Dependencies]
- Fix library dependencies (remove OpenMP; add LZ4, used by FLANN).
- Allow selection of single-/multi-threaded FFTW library.
[Performance]
- Use single-threaded FFTW by default. Multi-threaded FFTW scales poorly (even becomes slower) with more cores for non-huge grid sizes.
[README]
- Add meson build instructions.
- Update references.
- Misc edits.
[Bug fix]
- Correct mismatched `new`-`free` and `malloc`-`delete` calls.