performance - capn-freako/PyBERT GitHub Wiki
PyBERT Performance by Platform
Here we tabulate the PyBERT performance (Msmpls./min.) observed by different users on different platforms.
Note: The following data are for:
- "Standard" Python installation for platform (i.e. - downloaded from
python.org
) - PyBERT v4.0.0
- PyIBIS-AMI v4.0.2
installed via: pip install pipbert
. (Note the new Pip package name!)
Platform | David Banas | David Patterson | Stephen Newberry | ? |
---|---|---|---|---|
MacOS ARM (M1/M2) | (2022 Air M2 16/512GB) 7.5 | |||
MacOS Intel | (2012 Air 4/256GB) 1.7 | (2015 MacBook 2.7Ghz/i5/8GB) 2.9 | ||
Linux | (64-core Threadripper PRO 1TB/?) 3.6 | |||
Windows 11 | (Dell XPS 15 7590 32/1024GB) 0.6 | |||
Windows 10 | (i7-6700k/16GB) 1.1 | |||
Windows X |
Please, send us your numbers and we'll add them to the list. :)
General Notes
-
I'm unable to test under Linux, as the newest version of Python available under WSL2 Ubuntu is 3.6, but we require >=3.8 for both PyBERT and PyIBIS-AMI.
-
Windows performance can probably be improved dramatically, by using the
intel-numpy
package. However, we've got some versioning tangles that need to be unwrapped. :( See this comment on PR115. -
Performance number is computed by running
from pybert.pybert import PyBERT thePyBERT = PyBERT(gui=False) print(f"Performance: {60*thePyBERT.total_perf/1e6:4.1f} Msmpls./min.")