Fixed#938: Speed up njit-decorated function for sliding dot product [WIP]

[NimaSarajpoor]

This PR addresses the issue raised in #938. We want to create a new function for sliding_dot_product that:
(1) has a performance that is close to the performance of core.sliding_dot_product
(2) can be called by a njit-decorated function

Currently, the existing alternative to core.sliding_dot_product is core._sliding_dot_product; however, in contrast to core.sliding_dot_product, its performance is NOT independent of m, the length of query.

As suggested by @seanlaw, we are going to try OTFFT to implement fft and ifft. Then, we use them to create a new function for sliding_dot_product.

[review-notebook-app]

Check out this pull request on

See visual diffs & provide feedback on Jupyter Notebooks.

---

Powered by ReviewNB

[codecov]

Codecov Report

All modified and coverable lines are covered by tests ✅

Comparison is base (857063c) 98.93% compared to head (2c5dd4d) 98.93%.
Report is 4 commits behind head on main.

Additional details and impacted files

@@ Coverage Diff @@## main #939 +/- ##
=======================================
Coverage 98.93% 98.93% =======================================
Files 84 84 Lines 14292 14292 =======================================
Hits 14140 14140 Misses 152 152 

☔ View full report in Codecov by Sentry.
📢 Have feedback on the report? Share it here.

[NimaSarajpoor]

@seanlaw
I think the previous notebook got too messy and there were different versions there. I did not remove it (I just kept it for our future reference). Instead, I just pushed a new notebook (see c2a9b43). My goal is to add functions step by step based on the discussions we have had so far on this PR and #938.

For now I just resolved the previously- provided comments (by you and me) that were regarding the initial notebook. I will consider them in this new notebook though.

[NimaSarajpoor]

Line #20. x[:] = scipy.fft.fft(x) # we will implement our fft shortly!

We will later work on implementing this fft function using 6-step / 8-step algorithm. We then change this line and the caller function accordingly.

Reply via ReviewNB

[NimaSarajpoor]

I believe this PR is not needed anymore. See:

• https://github.com/stumpy-dev/stumpy/blob/main/stumpy/sdp.py
• https://github.com/stumpy-dev/sliding_dot_product
• Fixed #1111 Add module for sliding dot product; Include pyfftw as (soft) dependency #1118 (comment)