Hacker News Board Hacker News, browsed like a classic message board
Live data from Hacker News

CUDA on Windows Subsystem for Linux 2

devblogs.nvidia.com

Post reply on HN
Print view View original

91–100 of 221 posts

Re: CUDA on Windows Subsystem for Linux 2

#91

Post by SXX »

mycall wrote: post #76
SXX wrote: post #50

Earlier quoted context omitted.

WSL2 and all these features don't have anything to do with open source either. It's basically a shim for communication of two proprietary bits of software on both sides. Fortunately all these graphics and compute shims will never get merged into upstream kernel so they'll be second-class citizens like forever. Dave Airlie who is maintainer of Linux graphics subsystem will never allow to merge it since it's can only b…

Is this the wrong repository for what you are referring to, as this appear open source? https://github.com/microsoft/WSL2-Linux-Kernel

This custom kernel only contain shim that needed to pass commands from proprietary component on Linux to proprietary drivers on Windows.

When it's come to graphics drivers Linux upstream don't accept any code that can only be used for proprietary components. So all this Direct3D and CUDA shims never gonna be merged.

Re: CUDA on Windows Subsystem for Linux 2

#92

Post by FridgeSeal »

anonymousDan wrote: post #12
benbristow wrote: post #9

Earlier quoted context omitted.

Does kind of feel that way. Having a full Windows Desktop with good graphics drivers and software like Microsoft Office, Photoshop, commercial games etc. and Linux for the programming/server-software side of things. I have literally no reason to dual-boot with Ubuntu anymore. WSL2 and Visual Studio Code fulfils all my Linux coding requirements. Maybe (probably?) a bad thing but it's damn convenient.

Yeah I just started using Windows 10/WSL yesterday and I have to say I'm impressed. Are there any downsides/painpoints I should be aware of?

I/O performance is pretty horrific.

Re: CUDA on Windows Subsystem for Linux 2

#94

Post by 0xFFC »

deno wrote: post #51
withinrafael wrote: post #32

Throwing out a random data point: Until now, NVIDIA has been very anti-virtualization (on the consumer side), going as far as to engineer its drivers to detect in-use virtualization (Hyper-V, Xen, QEMU, etc.) and fail on purpose [1]. I'm curious to see how they now handle this scenario (given WSLv2 runs in a virtual machine). Perhaps they just commented those checks out in their 'specialized' drivers, an interesting…

It’s not GPU pass-through, it’s paravirtualization.

Can you explain about paravirtualization?

Re: CUDA on Windows Subsystem for Linux 2

#95

Post by jfim »

jchw wrote: post #87
anaisbetts wrote: post #21

Having set this up, this is going to be absolutely huge for ML. A lot of nonsense is getting cut down by this: 1. You don't have to install special CUDA-specific drivers that are behind the normal gaming drivers anymore. You'll soon be able to just use the Regular nVidia Drivers (and even now, all you need to do is install a Beta driver version). That's huge for someone just starting out, they don't have to have "the…

> (aka the Wayland Nightmare and DPI issues everywhere) It’s more of a pick-one. Wayland has pretty good per monitor scaling; your X11 apps will be blurry but that’s about all I’ve experienced. The Windows DPI nightmare long surpassed it anyways. On a typical Wayland desktop the usual biggest problem is the DPI being set wrong for a monitor. Even on X11 the worst you get is applications that don’t scale correctly. On…

> And that’s not all. I close my laptop lid when my laptop is docked. Windows helpfully sleeps the laptop even though external monitors, a mouse, and a keyboard are connected. There is no fix.

Right click the battery icon in the system tray > Power Options > Choose what closing the lid does > Do nothing.

Re: CUDA on Windows Subsystem for Linux 2

#96

Post by smichel17 »

dman wrote: post #68
patrec wrote: post #65

Earlier quoted context omitted.

> I personally don't get why any developer would prefer Windows over Linux Because windows now literally has all the stuff Linux has plus a lot more. The only reason for a well paid developer to prefer linux over windows for a desktop OS is that you don't need any of the extra stuff (a decent desktop experience, hardware support and a commercial end user software eco system) or that you have some fundamental objectio…

I hope they stop at extend this time.

I think "extend" is a good safety buffer / canary in the coal mine, so I'd be happier if they stuck to "embrace".

Re: CUDA on Windows Subsystem for Linux 2

#97

Post by smichel17 »

Skunkleton wrote: post #45
alerighi wrote: post #36

Earlier quoted context omitted.

First a developer you don't do only programming: you maybe have to work with documents, do a presentation, a spreadsheet and thus use Office. You maybe have to make some icons or logos and thus you need Adobe Illustrator. You maybe need to connect to a printer like the one that we have at my office that good luck doing that on Linux. You maybe need a CAD software to designa PCB, you need whatever other software that…

Just to add to this: Linux is happy to work in whatever way is convenient to you. WSL2 is case and point here. Windows is less flexible. If I can get everything I want out of one environment (stable gui, business software, and a native-like linux shell), that will save me the hassle of multiple machines, traditional VMs, or dual booting. I don't think we will see the Linux Subsystem for Windows 2 ever. I also am not…

The Linux Subsystem for Windows 1, on the other hand, is called Wine and works impressively well. Better than WSL1 did, at any rate.

Re: CUDA on Windows Subsystem for Linux 2

#98

Post by emsal »

black_puppydog wrote: post #5
FiReaNG3L wrote: post #2

This is BIG for machine learning adoption! 2 years ago it was a nightmare to setup everything on a Windows box, and wasn't working in any VMs (WSL or VirtualBox). Very excited about this.

This feels like it's gonna be terrible for people already doing ML. Now, the following has more than a bit of elitism, I'm aware of that, so please don't comment to point that bit out... I'm afraid that searching for answers to specific ML questions will start feeling like trying to google some Windows problems (every few years when I make the mistake of trying to help someone out) where it's all "download this, then…

What's scary about Googling for specific ML questions beginning to look like googling some Windows problems? This kind of dichotomy already exists between people who interact with ML only through keras and those who use a more flexible library. If you fall into the latter category it's still easy enough to find the answers to ML questions that mostly pertain to the deeper stuff. Don't even get me started on the difference between people who do and don't know statistics theory. But the point is that all of these groups do ML and they coexist just fine.

On another note. I don't think technical proficiency (or whatever your elitism-metric is) is very correlated with using Windows as the primary platform. You can Google stuff related to a Windows problem and through that learn how to work with the registry, and yes, some of the things you need to do require you to RTFM. So this elitism also suffers from being off the mark.

Re: CUDA on Windows Subsystem for Linux 2

#99

Post by claudeganon »

epmaybe wrote: post #79
claudeganon wrote: post #75

Earlier quoted context omitted.

Anyone recommend a good guide for quick setup and noodling on one of these platforms? I tried a Digital Ocean machine learning droplet awhile back with one of their guides and nothing on it worked.

I liked using colab to test certain things out, but it depends on whether or not you enjoy using tensorflow. Edit: I should also add that if you enjoyed the colab experience but want to scale up, you can attach the same notebooks to a Google cloud compute VM and run them on that.

Thanks! I'll check it out.

Re: CUDA on Windows Subsystem for Linux 2

#100

Post by est31 »

patrec wrote: post #65
jacquesm wrote: post #10

Pity. CUDA was/is proof positive that Linux on the desktop is perfectly feasible and that you can use it to both do a UI and do meaningful computation on the same machine without getting tied down into all kinds of licensing schemes. Opening this further up to Windows gives fewer people a really good reason to try out Linux as their daily driver. I personally don't get why any developer would prefer Windows over Linu…

> I personally don't get why any developer would prefer Windows over Linux Because windows now literally has all the stuff Linux has plus a lot more. The only reason for a well paid developer to prefer linux over windows for a desktop OS is that you don't need any of the extra stuff (a decent desktop experience, hardware support and a commercial end user software eco system) or that you have some fundamental objectio…

According to the Stack Overflow survey, Linux Desktop marketshare for developer has been growing constantly over the years. 23% in 2018's "primary desktop OS" question, 26% in 2019, 27% in 2020. Prior to that the question was asked differently, allowing multiple choices, but Linux was even lower back then. WSL was released in 2016 so much of this growth has happened while it was out. There might be some delay in the effects though, and maybe it would have been a much faster trend had it not been for WSL.

https://insights.stackoverflow.com/survey/2016#technology-de...

https://insights.stackoverflow.com/survey/2018#technology-_-...

https://insights.stackoverflow.com/survey/2019#technology-_-...

https://insights.stackoverflow.com/survey/2020#technology-de...

Post reply on HN