Valve's L4D2 Is Faster On Linux Than Windows

[entropy]

http://blogs.valvesoftware.com/linux.../#comment-4341

Valve Linux team says:
August 2, 2012 at 3:10 pm

We are already in contact with other game developers and building a list
of those with interest in porting existing games in their current catalog to Linux.
Some of these companies already have Linux versions available.

Sounds promising. Somehow...

[RealNC]

Might Valve be trying to pull off something similar to the Amiga->PC case, where several publishers came together and decided to drop the current de-facto standard gaming platform and create another because they were unhappy with the current platform? Or is this just wishful thinking?

[Scali]

Might Valve be trying to pull off something similar to the Amiga->PC case, where several publishers came together and decided to drop the current de-facto standard gaming platform and create another because they were unhappy with the current platform? Or is this just wishful thinking?

Amiga the 'de facto standard' gaming platform? Not really. It may have been the best platform, but it still had quite a lot of competition from C64, Atari, MS-DOS and the various consoles.
PC simply took over because gaming evolved where the Amiga didn't. Games like Wolfenstein 3D and Doom changed everything, and the Amiga hardware was not updated to handle such games. The higher cost of PCs for gaming was also not a problem, so gamers moved from Amiga to PC to play the new 3D games.
Not really a move of any publishers in particular, just evolution. Much like how MS-DOS was abandoned in favour of Windows at some point.

I doubt that Valve has enough of an influence to move the gaming world from Windows to linux to be honest, so I'd say 'wishful thinking', yes.
I mean, after all is said and done, it's still just an old DirectX 9 game that is being ported. Now if something like Crysis 3 or Battlefield 4 were announced for linux, it's slightly different, but I don't think Valve's engine and games have 'killer app' status at this point.

[AJSB]

I don't think Valve rewrote Source Engine from Dx to OpenGL....and i quote:

---------------------------------------------
Valve Linux team says:
August 2, 2012 at 3:00 pm

We use a modified abstraction layer (based on the original Mac OS X work) that translates Direct3D calls to the proper OpenGL calls. This layer has received the most work but changes have also been made above this layer that resulted in improved performance.
Reply
---------------------------------------------


Am i wrong or they "simply" add an abstraction layer that translates Dx - > OGL ?

If it is so, *that* makes even more impressive their results because any abstration layer increases always , even for a small amount, the time it takes to calculate and transmit to the GPU....or i'm just saying silly things here ?

[elanthis]

if valve adopt let say Qt or GTK/ISO C++[x revision]/opengl/webkit/etc you can maintain 99% of your codebase plataform agnostic that will greatly reduce support cost and will help to implement fixes a freaking lot faster

It would be cross platform between three PC OSes, two of which are close to inconsequential. There are no magic set of libraries to be portable to the platforms that game companies care about. Yet.

Portability is a bigger thing today than it ever has been. The target platforms just aren't PC. They're consoles, mobile phones, the Web, etc.

If you really care about maximum portability, you're either writing everything in HTML5 or using a tool that compiles to HTML5.

On a side note, Valve already has adopted most of the tech you listed. Not Qt or GTK+, but I have a feeling those are on their way out anyway. The world seems to be moving towards a more HTML like UI (which Valve has ready done) rather than using the rigid strict widget libraries of years past. The rest of what GTK and Qt do are better done in smaller, more focused libraries.

[kwahoo]

Now if something like Crysis 3 or Battlefield 4 were announced for linux, it's slightly different, but I don't think Valve's engine and games have 'killer app' status at this point.

Forget about Crysis 3 and Battlefield 4, we all are waiting for Half-Life 3
Source-based games are still very popular and Dota 2 (while still beta) is the most popular game at Steam.

[Scali]

Am i wrong or they "simply" add an abstraction layer that translates Dx - > OGL ?

When they released the OS X version, they were quick to point out that they weren't using any kind of translation/wrapper/emulation, but wrote native OpenGL code...
Which is what I'd expect.
Direct3D is very lowlevel, so you'd normally write a layer of code around that, to manage all your resources, content etc. If you then take OpenGL and do the same, and make sure the layers are mostly compatible, then the rest of the engine can make use of it without a lot of modifications.
That is how I do it anyway.
It's also what they mention in the blog: "The second category would include reducing overhead in calling OpenGL, and extending our renderer with new interfaces for better encapsulation of OpenGL and Direct3D."
They have a layer that encapsulates OpenGL and Direct3D (at the same level, not encapsulating OpenGL in a Direct3D wrapper, which then is encapsulated in the higher level layer).

I think that's what they meant to say, but somehow it came out wrong.

[Hamish Wilson]

It seems you just have a strange idea that porting an application would in any way imply that the code has to be in any way portable at all. That is not the case, as you can see here: http://en.wikipedia.org/wiki/Porting

As for the triviality, I beg to differ. Firstly, games were developed by much smaller teams as well, back then (usually just one or two programmers). Secondly, where Windows and linux run on identical hardware, these older platforms often had quite different hardware characteristics, and it would sometimes require signifcant effort to reinvent the same graphics on a completely different platform.
I suggest you read this blog, to get an idea: http://popc64.blogspot.com

My point being: it's pretty sad that developers don't put in any effort anymore to support other platforms, or even different OSes on the same platform, where one day they would bother porting games to many different platforms, even if it meant rewriting the code from scratch, and even recreating the content (the original version of Prince of Persia was on Apple II, later versions such as MS-DOS and Amiga had updated graphics and sound effects).

On this I fully agree.

I doubt that Valve has enough of an influence to move the gaming world from Windows to linux to be honest, so I'd say 'wishful thinking', yes.
I mean, after all is said and done, it's still just an old DirectX 9 game that is being ported. Now if something like Crysis 3 or Battlefield 4 were announced for linux, it's slightly different, but I don't think Valve's engine and games have 'killer app' status at this point.

When it comes to this though, I am getting kind of sick of the argument that "it is only game A we are getting, and not game B, which is the only one that can really change things". Where are these arbitrary ratings and standards of what makes an important game coming from? I would have thought Valve would be AAA enough (another vague qualifier, but anyway) to appease most of these kind of people, but I guess I was wrong. Now we are getting that we need to get game "X" or "Y" instead.

[Scali]

When it comes to this though, I am getting kind of sick of the argument that "it is only game A we are getting, and not game B, which is the only one that can really change things". Where are these arbitrary ratings and standards of what makes an important game coming from? I would have thought Valve would be AAA enough (another vague qualifier, but anyway) to appease most of these kind of people, but I guess I was wrong. Now we are getting that we need to get game "X" or "Y" instead.

I guess you missed my point, which is:
Valve's efforts don't show parity with cutting-edge Windows games. The Source engine is based on 10+-year old technology (which is why they are getting 300+ fps in the first place, it does not really stress a modern system).
It would be more interesting if a full DX11+ whizzbang engine/game (Crysis 3 and Battlefield 4 being metaphors of that, note that these are imaginary future game titles, not current games) were ported to linux (not for me, don't get me wrong. I'm not even a gamer... but those are the kind of games surrounded with the most hype, and those generally get sold on their technical merits, even when they're not actually all that fun to play. They'd give linux gaming more of a boost).

[entropy]

I would have thought Valve would be AAA enough (another vague qualifier, but anyway) to appease most of these kind of people, but I guess I was wrong. Now we are getting that we need to get game "X" or "Y" instead.

I'd even go that far to claim that Valve is one of the very few big companies that also deliver AAA content besides AAA technology and an AAA budget.