Skip to content
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

No multiprocessing floating point improvement between WSL1 and WSL2 #4135

Closed
kimrj opened this issue Jun 14, 2019 · 8 comments
Closed

No multiprocessing floating point improvement between WSL1 and WSL2 #4135

kimrj opened this issue Jun 14, 2019 · 8 comments

Comments

@kimrj
Copy link

kimrj commented Jun 14, 2019

See here: http://www.simxon.com/articles/20190614-wsl2-and-still-waiting.html

@benhillis
Copy link
Member

Can you please provide a summary of the issues you are hitting?

@kimrj
Copy link
Author

kimrj commented Jun 14, 2019 via email

@benhillis
Copy link
Member

benhillis commented Jun 14, 2019

Filling out the issue template is a good start. I'd prefer GitHub issues to not be links to external blogs.

@therealkenc
Copy link
Collaborator

therealkenc commented Jun 14, 2019

I'd prefer GitHub issues to not be links to external blogs.

Or just a repro. The blog worms to some previous article, which worms on something about (something called) OpenFOAM, which links to a "build guide". All very interesting stuff I'm sure (really).

If this format does not suit you, how would you like me to present my information, then?

CLI repro steps from a clean install of some WSL distro from the Store. A script or github repository that can be cloned is nice, but a list copy-pasteable command line steps will often do.

Once you have those lined up, try running those same steps in a bog-standard Hyper-V instance and see if the results are different than WSL 2. Certainly if there is a diverge from Hyper-V it should be looked at. If performance is the same, there are whole communities for discussion about running numerical apps in vitrual machines. That might be OpenFOAM on Hyper-V. Or lots of other scenarios; I dunno AV1 trans-coding on Windows QEMU maybe. Lots of improvements being made to VM solutions all the time. Sounds like you are pretty wired into that community so you'll probably see them as they happen.

[ed] Belatedly looking at the timing at the very bottom of your blog post, interesting that your use-case (whatever it may be) is running almost exactly half speed. That could be coincidental, but this kind of screams dupe #4137.

@kimrj
Copy link
Author

kimrj commented Jun 15, 2019 via email

@therealkenc
Copy link
Collaborator

therealkenc commented Jun 15, 2019

due to my ignorance concerning Microsoft procedures

No worries. Those are in the template that you deleted.

In that context, how can I hand over, say, 5 GB, to you?

Any way you like so long as it is accessible on the Interwebs. Although the odds of someone downloading 5GB, absent a lot of evidence that whatever it is you are reporting (which remains unclear), is required to demonstrate a diverge against bog-standard Hyper-V behavior is probably pretty small. You want a targeted repro.

I do not understand as it refers to concepts which I have never heard of. Will you please clarify?

There is information on Hyper-V here. Information on QEMU is here. Information on transcoding is here. AV1 here. You asked.

But if I've skimmed your blog post, you've already run on Windows Docker and it is all the same tech as WSL2. I am doubting you'll learn anything new. If it were me I'd subscribe to #4137 because Occam's Razor. But tests that demonstrate novel errant behavior are always welcome and encouraged here. Bonne chance.

@kimrj
Copy link
Author

kimrj commented Jun 19, 2019

Now I have tried bare-metal Hyper-V: http://www.simxon.com/articles/20190618-wsl2-and-still-waiting-ii.html

@therealkenc
Copy link
Collaborator

therealkenc commented Jun 19, 2019

Microsoft has claimed that they would soon be able to run Linux binaries as efficiently as a native GNU/Linux installation

That line lacks citation. [The WSL2 announcement does not.]

There is no WSL2 actionable here. Thanks for sharing. If you do come up with a tight repro explaining the ~50% fall-off versus metal (I suspect a short 50-liner ought do it) do feel encouraged to submit an issue following the template. In the mean time subscribe #4137.

Sign up for free to join this conversation on GitHub. Already have an account? Sign in to comment
Projects
None yet
Development

No branches or pull requests

3 participants