Skip to content

Instantly share code, notes, and snippets.

@yuxuanzhuang

yuxuanzhuang/benchmark_on_multi_threads.ipynb

Last active October 13, 2020 20:35
Show Gist options
  • Save yuxuanzhuang/17ed6def63b08248db59f2c44e3e0419 to your computer and use it in GitHub Desktop.
Save yuxuanzhuang/17ed6def63b08248db59f2c44e3e0419 to your computer and use it in GitHub Desktop.
Download ZIP
Display the source blob
Display the rendered blob
Raw
benchmark_on_multi_threads.ipynb
Loading
Sorry, something went wrong. Reload?
Sorry, we cannot display this file.
Sorry, this file is invalid so it cannot be displayed.
@yuxuanzhuang
Copy link
Author

@orbeckst Sorry something is wrong with gist, cannot display the old one properly.
It's a 6-core CPU. I have added the benchmark for 6 threads--don't seem to differ much from 1 thread.
So it's most likely a hyperthreading issue. What confused me is, subscribing to 12 threads doesn't hurt the performance without Transformations.

@yuxuanzhuang
Copy link
Author

Note the last three cells that self._xdr.read() is marginally slower, which I assume have nothing to do with Transformations, but still only happens with Transformations present.
https://gist.github.com/yuxuanzhuang/4918cd1b5d8d62de79eab9df40de4bb7

@orbeckst
Copy link

It is admittedly weird that the presence of transformations affect the read performance.

Is this reproducible?

(Actually, will continue this specific discussion on the other gist.)

Sign up for free to join this conversation on GitHub. Already have an account? Sign in to comment