Hacker News Reader

Comment by shmerl | Hacker News Reader

Any ETA on Linux support?

Given that Linux gaming is perpetually 3-5 years behind Windows/console gaming, I'd say 3-5 years is probably a good bet for it running in any sort of usable capacity (i.e. more than a tech demo someone reverse engineered to prove it's possible).

Before you down vote; I'm really not being snarky here -- games on Windows typically run 30-40% faster on the same hardware than they do on Linux [1] because there just isn't the market there to make it worth doing the optimization.

[1] http://arstechnica.com/gaming/2015/11/ars-benchmarks-show-si...

> games on Windows typically run 30-40% faster on the same hardware than they do on Linux

Typically many game developers have no clue about Linux programming either. So poor performance doesn't surprise me. When developers know what they are doing, performance is good (example: http://knockoutgames.co).

Also, OpenGL is a major mess for several reasons (barely compatible implementations not complying to the standard and so on).

Now there is Vulkan however, so VR developers should have easier time with it.

> Typically many game developers have no clue about Linux programming either. So poor performance doesn't surprise me.

But ultimately, this is a weakness of the Linux platform.

Most games today are cross-platform, so there is a core set of assets and code that has to remain the same across every platform to ensure the game plays the same way. This means that (for right now, anyway - Vulkan was designed to help with this situation) most hardware-specific workarounds/optimizations have to be done in the drivers -- and there are LOTS of them. Game-specific things like "run this shader code, but this model of card has graphical bugs when features x, y and z are active together so stealth disable feature x" or "this card supports this transform algorithm but its output is inconsistent with the version we developed it with, so disable it for this hardware" happen at the driver level. Same thing happens on consoles, but they usually have more direct control over the hardware so console developers have gotten used to this type of driver optimization.

This optimization doesn't happen on Linux (or Mac for that matter), so any games that get ported are just unoptimized.

Wouldn't that happen if you're running the proprietary drivers? I thought in general the proprietary drivers are built from the same core codebase across multiple platforms -- once you've gotten past the WGL, GLX, etc. bindings, and you have some system call for getting data to and from the card, the rest of the driver should be basically the same across platforms.

And similarly the games themselves are the same across platforms, apart from their end of the WGL or GLX or whatever bindings. So anything in the driver that says `if (!strcmp(name, "Portal 2"))` is going to be equally applicable regardless of whether it's a Portal 2 compiled for Windows, Mac, or Linux.

It's certainly a weakness of the all-free-software subset of the Linux platform (at least on Nvidia or AMD), but that's not the only option.

Overall i'm not agree with statement, but what he was talking about is that Windows drivers have tons of hacks to optimize certain D3D code paths that used by most games and sometimes even fully replace whole shaders with GPU vendor one.

Why wouldn't those hacks happen on Linux? I guess the reason is that they're specific to Direct3D and not OpenGL, but for those games that do OpenGL on Windows (which I think is somewhat common), wouldn't the same hacks be present in the Linux OpenGL drivers as the Windows OpenGL ones?

Of course they would and that is reason why I don't agree with the statement. Even AMD had profile for Valve games that improved performance on their blob and Nvidia had several different profiles some of which come from Windows drivers: to be exact one for id Software games that was always using OpenGL on Windows.

Of course there is open source drivers that will likely never support game-specific optimizations since for them codebase quality and overall performance is more important than few tricks that make specific games faster.

OpenGL is actually not all that common on Windows anymore -- D3D is more performant, and having a a DX11 compatible card isn't as big of a hurdle these days. Also, the big engines all support both for cross-platform reasons, but most devs will default to D3D on Windows because it's more likely to work consistently across various permutations of hardware/OS/peripherals (IMO this is more because Windows sucks; but Microsoft is well versed in the very specific ways that Windows sucks so they can work around them).

: When I say "Windows sucks", I'm just talking about the complexity inherent in dealing with a consumer operating system designed to run on arbitrary 3rd party hardware. APIs like DirectX/Direct3D provide an abstraction layer that make the complexity more manageable; and they have a benefit in that they are provided by Microsoft.

They present on Windows because gpu (driver) vendors choose spend tons of their money to make them happen. And they spend their money exactly why? Because they see their profit in Windows pc market.

Intel, AMD and Nvidia work on Linux Vulkan support as well. They aren't tied to Windows.

I would argue that Linux Vulkan support from Intel/AMD/NVidia is likely aimed more at Android than traditional Linux desktop distros.

Also, the whole point of Vulkan is to move these card-specific optimizations outside of the core driver while still letting games use kernel space to interface directly with the hardware (and doubtless platform vendors will put an abstraction layer on top of that like Metal on iOS). So the whole situation I'm describing has been a huge pain for a long time, but things are starting to get better.

> But ultimately, this is a weakness of the Linux platform.

No, it's a weakness of developers who do something without knowing what they are doing. As I said, not everyone is like that.

> This optimization doesn't happen on Linux

It now does, with Vulkan.

VR is interesting for a lot more than just games.

There's no indication I'm aware of that Oculus/Facebook is interested in supporting Linux. Your best bet will be using Wine, but that requires both the Rift games and the Oculus display server to work in Wine, which is a tall order. But it's something we're looking at.

> There's no indication I'm aware of that Oculus/Facebook is interested in supporting Linux.

They said explicitly that they plan it[1][2]. But I saw no ETA.

[1]: https://www.gamingonlinux.com/articles/oculus-rift-vr-headse...

[2]: https://twitter.com/PalmerLuckey/status/674311865023918080

Cool, news to me! I'd love to be wrong.

What is this Slashdot? ;-)

Why Slashdot?

I think he was referring to the fact that on Slashdot, commenters always ask for Linux support for every newly released product.

Why is it so specific to Slashdot though? If some product doesn't support Linux - you can expect such questions everywhere.

Only on forums with a Linux user population of over, say, 1%. Which is a very tiny number of forums.

We aren't talking about random forums. Most of them aren't focused on technology or gaming technology as an example. So it's not unexpected if they are (like here).

Reply on news.ycombinator.com