OpenMP is extra slow on Mac M3

Issue type

Performance

OS platform and distribution

MacOS Sonoma 14.5

Current behavior?

Current benchmarks which heavily rely on OpenMP are very slow.

Expected Behavior?

It should perform similar to Linux version.

Standalone code or description to reproduce the issue

Run FFTBatch or FRI benchmark and it will result in 2-3x slower performance than expected.

Additional context

We have tested the serial version performs as expected. So it is definitely a performance issue of OpenMP on Mac (only tested on M3).