Jeff Long

Thanks for looking into this. A single PR for all CtrlPort fixes would be good. You can use multiple commits in the PR if you think the might be applied/reverted/bisected...

This sounds like it could be related to #7009, which was just merged on `main`.

The unhandled exception has something to do with `pkg_resources`, part of `setuptools`. Still seems unrelated to anything in this PR.

There could be cases where there is no efficient SIMD method for a processor and std is slower than lookup. Getting the same value result (maybe with some noise) from...

We should put some energy into fixing this up. Do you think there would be any improvement in fxpt if `gen_sine_table.py` did proper rounding of both x and y values?...

That's the speed difference between fxpt single value and fxpt vec. Thus the disbelief. ``` $ ./src/gnuradio/build/gr-blocks/tests/benchmark_nco +——————————————————————————————————————————————————————————————————+ | fxpt sine time: 2.7000e-07 s throughput: 7.407e+13 it/s | | fxpt...

Yah, the performance stuff was a side trip. However, 40-50 MS/s isn't going to be a great thing. Affected blocks could have a parameter for `Prefer [Precision | Performance]`.

I was thinking that a few more bits in the table, and either rounding or the interpolation method above, would be interesting to try out. The table has 2**10 entries...

@VladislavF curious whether increasing NBITS has any effect on the error you are seeing. Try this patch if you get a chance. The API is still `float`. ``` diff --git...

The table stores y=f(int(x)) and m, then the lookup does y + dx * m or something similar, for interpolation. Agree that the "whole" solution is not just to increase...