Python often gets a bad wrap for performance, specifically regarding numeric computations. In this presentation I've dug in to Numba and how I got around an error I was having, solving a performance bottleneck I had.
2. Brief Problem Statement
Reading DIRECTLY off the slides
Never posts the slides
Video is available
Automatically grab all the slides
Quickly underwhelmed with speed
How to speed this up
Speed has never been an issue previously
3. Python isn’t the Problem*
The fault dear programmer isn’t in Python
But sometimes it is
Get around it
Async
C extension
Pypy (JIT everything)
Numba (just JIT certain methods)
5. Initial Code/Data
Reading the frames:
Data:
1920 x 1080 x 3
All pixels recorded in the video, then 3 values for each pixel (RGB)
Test for validation:
14. What I learned
Follow the hierarchy of needs!
If you don’t you will get burned
Tests/initial benchmarks are key
Slowest code first
Incrementally solve the next bit
Very rarely, break out specialized tools
Know your tools
Has anyone ran in to the problem I had with broadcasting?