News for hackers. Stuff that matters.
Re:Microsoft to acquire ZeniMax Media and Bethesda Softworks... by MaxBarraclough on Monday September 21, 2020 @09:49AM (#24543391)
If you want to target Xbox, you need to write a D3D renderer anyway (although the Xbox API has some significant differences, if I understand correctly). There's little point writing an OpenGL renderer if your target platforms are Windows, Xbox, and PlayStation (which has its own graphics API).

Also, my understanding is that on Windows, OpenGL generally runs into more issues with driver bugs than D3D does.

Re:Microsoft to acquire ZeniMax Media and Bethesda Softworks... by gambiting on Monday September 21, 2020 @10:41AM (#24544018)
I work for a large video games company and we usually write two backends - DirectX for Windows and Xbox, and a special one for PlayStation. That is changing slightly with Stadia, because that forces us to write a Vulkan renderer too. But on the few games that have it, the DX12 performance is better on windows than that of the Vulkan renderer backend - so we haven't released it to the public.
Re:Microsoft to acquire ZeniMax Media and Bethesda Softworks... by shrewduser on Monday September 21, 2020 @11:45AM (#24544719)
you should release vulkan for proton / linux users i think
Re:Microsoft to acquire ZeniMax Media and Bethesda Softworks... by KitDuncan on Monday September 21, 2020 @12:01PM (#24544942)
Please do release it to the public! Just for Proton and to experiment/bechmark.
Re:Microsoft to acquire ZeniMax Media and Bethesda Softworks... by monsieurbanana on Monday September 21, 2020 @12:52PM (#24545637)
As much as I'd like that, I can hear from here the outcries from the gaming community.

"Your game is bugged! It doesn't work! I want a refund!"

"But the vulkan renderer is only tested on Stadia, it's not officially supported and the game is free."

"I. Want. A. Refund."

Re:Microsoft to acquire ZeniMax Media and Bethesda Softworks... by MaxBarraclough on Monday September 21, 2020 @12:55PM (#24545674)
Put another way, then: doing so would mean taking on a considerable support burden.
Re:Microsoft to acquire ZeniMax Media and Bethesda Softworks... by jhasse on Tuesday September 22, 2020 @02:58AM (#24551696)
Make it a command line flag ;)
Re:Microsoft to acquire ZeniMax Media and Bethesda Softworks... by MaxBarraclough on Tuesday September 22, 2020 @06:07AM (#24552775)
Not a bad idea, especially if it's something like --experimental-vulkan. Unlikely to get blamed if it goes wrong, that way.
Re:Microsoft to acquire ZeniMax Media and Bethesda Softworks... by gambiting on Tuesday September 22, 2020 @07:05AM (#24553093)
So.....no, because gamer press is ruthless :P We've had one graphical bug in one of our games(which I'm certain everyone on the internet has seen by this point), and in reality it was only happening if you played on one specific GeForce card, and only when using beta Nvidia drivers. But of course no one cared that the bug only happened when using experimental drivers - it was somehow a proof of how broken our games are. If this(the vulkan backend) lead to any graphical bugs in our games, I can almost guarantee that they would be paraded everywhere, no matter how clearly we mark it as experimental and unfinished.
Re:Microsoft to acquire ZeniMax Media and Bethesda Softworks... by rawbot on Tuesday September 22, 2020 @06:29AM (#24552891)
Or a beta branch in Steam.
Re:Microsoft to acquire ZeniMax Media and Bethesda Softworks... by smileybarry on Monday September 21, 2020 @11:05AM (#24544276)
> Also, my understanding is that on Windows, OpenGL generally runs into more issues with driver bugs than D3D does.

I'm not sure if that's the case anymore, but it definitely has a bad rep on Windows. Initially (throughout XP and maybe some of Vista) OpenGL support on Windows was done by a OpenGL -> DirectX translation layer, so performance was always worse in OpenGL mode unless a game's Direct3D implementation was especially awful. This stopped being the case when NVIDIA started shipping a full OpenGL driver. (I'm not sure when AMD/ATI started shipping theirs)

Re:Microsoft to acquire ZeniMax Media and Bethesda Softworks... by qayxc on Monday September 21, 2020 @12:41PM (#24545484)
> Initially (throughout XP and maybe some of Vista) OpenGL support on Windows was done by a OpenGL

Initially (Windows 95), OpenGL support was provided directly by the OS. Starting with Windows 98, Microsoft stopped updating the OGL version of their reference driver, so users were stuck with OGL 1.1 unless the graphics card driver shipped with a custom OpenGL implementation.

So whenever an application uses an OGL version higher than v1.1, it is provided by the graphics card driver and that has nothing to do with DirectX. There is no translation layer in that case (unless of course, that's what the driver does internally, but that's up to the manufacturer).

TL;DR Custom OGL drivers shipped with every graphics card that supported OGL in Windows since 1998.

Re:Microsoft to acquire ZeniMax Media and Bethesda Softworks... by smileybarry on Wednesday September 23, 2020 @10:57AM (#24567862)
That hasn't been the case, then, because I clearly remember OpenGL being translates to DirectX in XP/Vista days. Whether it was because anything >1.1 called that up or because the driver vendors chose translation over native, I don't know.

Original link seems to be dead but Slashdot references Vista layering OpenGL on top of Direct3D:

https://slashdot.org/story/05/08/06/177251/windows-vista-may...

Re:Microsoft to acquire ZeniMax Media and Bethesda Softworks... by qayxc on Thursday September 24, 2020 @02:45PM (#24582067)
That comment references something else entirely - namely the Aero rendering system of the OS. This didn't affect applications that didn't use the Aero glass scheme (it was an option during window creation), however, or apps running in full-screen mode.

The Aero glass scheme was hardware accelerated and only worked with Direct3D, thus any OGL context created for a window using this renderer would have to run via Direct3D.

This is a very special case and as noted earlier, easily circumvented by simply not using this feature in your app.