OpenCL & CUDA GPU support
It would be nice to have (relatively) easy access to the dedicated GPU's on our Win10 PC's, and OpenCL support for the AMD Drivers & CUDA for NVidia drivers etc.
What I am thinking on specifically (although there are many many uses for it and CUDA support) is hashing functions that are lackluster within windows itself, but benefit immensely within Linux and Unix.
Cryptocurrency mining for example, is known to be much faster on Linux systems than Windows. Do you see any support for OpenCL and/or CUDA in the future?
Hey WSL users. We hear you!
We’re actively investigating this ask but don’t have any firm plans to announce just yet. Rest assured that we are as keen to deliver this ask as you are ;)
Please ping me on twitter with any questions: https://twitter.com/tara_msft
Yunbo WANG commented
This is really needed as we are using intensivelly GPU for my work. Due to this limitation, I had to uninstall WIndows and install Unbuntu to run my projects.
People prefer to not dualboot. For many, GPU support in WSL is why they still dualboot, and why they build machines without Windows to avoid dualboot.
Vincent Li commented
+1 looking forward to use pytorch on WSL/WSL 2.
Please enable GPU support in Ubuntu from Windows 10 store! Thanks.
Great work on this so far! Working with TensorFlow and would love to enable some more features with GPU support. Chose Windows over alternatives for this dev, please support!
Working with TensorFlow and would love to enable some more features with GPU support. Chose Windows over alternatives for this dev, please support!
ust some basic GPU support like CUDA would be great as well...
Jannie Kirsten commented
Just some basic GPU support would be great as well...
I had in here a more detailed comment but for some reason it was deleted.
1) It is based on Hyper-V based on Microsoft's blogs (you can search "WSL 2 Hyper-V").
2) It seems that Virtualbox 6 can run with Hyper-V, so it may not longer be an issue?
Kenneth Benson commented
That is something I need to know. Is it going to use Hyper-V? If it is that's going to make it rough on using some of the Python/Conda and other VM's out there that require Virtualbox. <sigh>
You'd think most of the pieces are now in place:
1) WSL 2 is now contained in Hyper-V VM
2) Windows (at least Server) supports Discrete Device Assignment (DDA) for GPU pass-through with Hyper-V
So perhaps some limited GPU support can now be offered (if you are willing to pass an entire discrete graphics card to WSL2)
Saad Khan commented
Now that WSL 2 has been announced and will run a full linux kernel, will this feature be possible? I couldn't find any information about presence or absence of GPU support in WSL 2.
To use Tensorflow and similar deep learning framework is a pain in Windows because most libraries / examples are generated and tested only for Linux.. to have this feature should be a great solution to continue to work with Windows.. otherwise the only solution should be to switch definitively on Linux
+1..looking forward for this
I asked for update on https://github.com/Microsoft/WSL/issues/3789 and they silenced it. I think the WSL team is doing great job addressing the huge number of incoming issues every day. But this might be too complicated for them to address so my guess is that we might not see any GPU support on WSL coming in the near future. I do hope I am wrong though.
Zeeshan Ali Sayyed commented
This SHOULD NOT BE on the backlog. This should be implemented as soon as possible. This looks like it is the most requested feature of all. Every time I think of not installing Linux on a new machine I get, I come and check this and go install Linux anyway.
Rather a reason why Microsoft always looses its users as they never deliver on the most wanted thing and when competitors release the most wanted feature, they release a broken fix for the feature and then they start working on it. A classic Microsoft move which has them loose on so much market absolutely ridiculous and a clear move to miss out on the market opportunity
SHIJIE MAO commented
Looking forward to it.
Alejandro Saucedo commented
This would be very helpful, perhaps to explore further the discussion around open sourcing parts of this project, drivers would certainly be an area where cross collaboration with other communities would really help. There are quite a lot of community maintained and supplier-maintained drivers.
Any walkaround on this? I want my primary desktop in windows and linux as the deep learning environment.