_phase2gray about slmsuite HOT 2 CLOSED

Hi @LeoKonzett , thanks for this great question.

1. The np.bitwise_and() performs the action of np.mod() in the case when self.phase_scaling == 1 (consider the case of a 10- or 7- bit SLM). This is super fast compared with np.mod(). However, you're completely right that this call is unnecessary in the case where you have an 8-bit or 16-bit SLM, as the type casting already performs the and. I have added a check to not do this in those cases (6ece4bb).
2. The intent was to avoid np.around() overhead. We could accomplish the same thing by adding .5 before type-casting, but I'm uncertain whether it's worth the overhead for this fractional change. I have added documentation (6ece4bb) specifying the floor()ed behavior.
3. In testing, I wasn't able to find a case where unsafe casting didn't behave like floor(), but if you can find one then I agree that a better solution is needed. I agree that the ironic "unsafe" flag correctly implies lack of safety, but I think it's deterministic in our case. Note that this line makes sure that all numbers are negative. The example that you gave is desired behavior, because this is the effect of mod by 256.

To answer your last question, yes the reasoning is speed. While something similar to this could be faster (like the ultimate solution for speed: going full-C and writing directly to some integer buffer and never touching floats), this solution was able to maintain both the 60 Hz operation in testing and the user-friendly 2pi float submission that we wanted to have. np.mod() was a main limitation for other implementations. Of course, we welcome any contributions to make the package better. If you have any ideas for how to improve this while still maintaining speed, feel free to discuss or add something! Let me know if you have any comments on the mentioned changes (6ece4bb), we'll probably pull request this within a week.

from slmsuite.

I'm closing this issue as these changes were pulled under request #24.

Please feel free, @LeoKonzett, to reopen it or to contact us about anything else! Looking forward to see what you do with the package.

from slmsuite.