Benchmarking The Python Optimizations Of Clear Linux Against Ubuntu, Intel Python

[phoronix]

Phoronix: Benchmarking The Python Optimizations Of Clear Linux Against Ubuntu, Intel Python

Stemming from Clear Linux detailing how they optimize Python's performance using various techniques, there's been reader interest in seeing just how their Python build stacks up. Here's a look at the Clear Linux Python performance compared to a few other configurations as well as Ubuntu Linux...

http://www.phoronix.com/scan.php?page=news_item&px=Clear-Linux-2019-Python-Perf

[schmidtbag]

Weird how the supposedly optimized Python performs worse than Clear's default python. Cool to see how Clear's default runs so much better vs Ubuntu though.

I'll be really interested to see the pypy benches. Makes me wonder if Clear's default can outperform pypy on Ubuntu.

[daufinsyd]

It's the contrary.
Clear's default = clear linux + recent optimisations

[darkbasic]

I'll be really interested to see the pypy benches. Makes me wonder if Clear's default can outperform pypy on Ubuntu.

I think it can. Pypy on clear would be very interesting as well.

[wpupkin]

Weird how the supposedly optimized Python performs worse than Clear's default python.

AFAIK Upstream Python should be built using PGO. They use their default tests, which covers all/most possible use cases.

But Intel ...

This change is critical because choosing the proper training task is crucial in the FDO technology. Why? Because each application will generate different block and edge frequency counts. The information in one profile could optimize the performance of one use case, but could also have a negative effect on the performance of other applications at the same time. For this reason, we highly recommend that developers have the option to define the proper training task for their application use case.

... seems chooses "the proper training task for their application use case", e.g. to optimize pybench results.

https://clearlinux.org/blogs/boosting-python-profile-guided-platform-specific-optimizations

https://github.com/clearlinux-pkgs/python3/blob/master/0003-Use-pybench-to-optimize-python.patch

[cybertraveler]

Really interesting, thanks. I'm also looking forward to seeing the PyPy tests!

[hoohoo]

Impressive!

[adude]

Which version of python are are you using in the testing on Ubuntu. That isn't clear since ubuntu has both 2.7 and 3.6. I don't want to assume.

[Michael]

Which version of python are are you using in the testing on Ubuntu. That isn't clear since ubuntu has both 2.7 and 3.6. I don't want to assume.

Python3 everywhere.

[ermo]

Michael

SciKit-Learn was a magnitude faster

should probably be "SciKit-Learn was an order of magnitude faster?