-
Notifications
You must be signed in to change notification settings - Fork 323
New issue
Have a question about this project? Sign up for a free GitHub account to open an issue and contact its maintainers and the community.
By clicking “Sign up for GitHub”, you agree to our terms of service and privacy statement. We’ll occasionally send you account related emails.
Already on GitHub? Sign in to your account
Improving Multi-processor Performance #959
Comments
I have 2 open questions:
If the answer to both questions is a "No", I assume that this paper is not useful for stumpy. |
@JaKasb These are good questions. I don't think Certainly, I'm not looking for 3x speedup but even a modest 20% speedup might be worth exploring if there isn't too much added complexity by, say, switching over to tiles for example. @JaKasb Do you see anything else that we can do to improve the speed of our CPU or GPU matrix profile calculations? Any low hanging fruits? |
All that advanced tiling stuff goes over my head. Readable code also has intrinsic worth. In my opinion, the tradeoff is not worth the effort. Furthermore, for higher speed one can use approximate matrixprofile algorithms or use multiple GPUs. |
This new paper titled "Exploring Multiprocessor Approaches to Time Series Analysis" claims to significantly improve the performance of matrix profile calculations. We should consider looking into this.
Additionally, we should consider if it may be practical to refactor some parts of our code a la cache oblivious algorithms
The text was updated successfully, but these errors were encountered: