BioShock Infinite Now Available For Linux, Demands Binary Graphics Drivers

[zanny]

Torchlight 2 works fine on radeonsi, dunno what prob with it zanny have

Alright, well I have a 290 on Mesa 10.5, X 1.17, kernel 3.18.6. Games like Civ 5, TF2, Portal 2, Metro LL, and Witcher 2 all work.

Here are the dumps if anyone cares.

Only really obvious thing is Assertion Failed: m_nNumRegisteredWriters == 0, which is common to every crash, but that isn't very informative since Googling it turns nothing up and its an assertion fail in the Steam client.

I already have libstdc++ removed from the Steam Runtime, hence how the other titles work.

Of course, I'm on Arch, using an r9 290, so I expect to be in the wild west here, but it doesn't change how reporting these issues is the worst kind of maze.

[gens]

If anyone is dissatisfied about the driver support, they should stop complaining and create better drivers. Somehow, the RMS extremists still are going to make this Nvidia's fault.
Software should always be made to match the specs, if the drivers don't follow the specs then the drivers should be fixed, not the other software.

from what i understand, creating good drivers without the datasheets is kinda hard
2'nd part i agree

[nanonyme]

from what i understand, creating good drivers without the datasheets is kinda hard
2'nd part i agree

It's hard even with datasheets. AMD has experience of how OSS drivers don't write themselves just if docs exist. You actually need paid developers to get predictable development speed

[sarmad]

Looking at required specs mentioned on the Steam store page, it looks like eON is adding some big overhead:
* Minimum RAM: 2GB on Windows vs 4GB on Linux
* Recommended RAM: 4GB on Windows vs 8GB on Linux
* Minimum VRAM: 512MB on Windows vs 1GB on Linux
* Recommended VRAM: 1GB on Windows vs 2GB on Linux

That said, I think eON is still a good solution for porting older games. Coming Steam Machines will have enough power to cover for the overhead and it's better to spend the resources porting more of those games than perfecting fewer games.

[tomtomme]

Phoronix: BioShock Infinite Now Available For Linux, Demands Binary Graphics Drivers

BioShock Infinite is now available to Linux gamers... assuming you're okay with using the binary AMD and NVIDIA Linux graphics drivers...

http://www.phoronix.com/scan.php?pag...-Linux-Release

The game Ravens Cry appears to have a benchmark mode, but I do not know if it is available on linux

they mention the benchmark functionality in the patch 1.02 notes here:

http://store.steampowered.com/news/?appids=281130

[You-]

OpenGL 4.2 tick box

I read a developer post somewhere that some of the OpenGL 4.2 stuff had not been completed yet because it would just be ticking boxes and not features actually being used, so it was lower priority.

Now that they are being used, maybe thre would be more impetus for providing support.

[Drago]

I read a developer post somewhere that some of the OpenGL 4.2 stuff had not been completed yet because it would just be ticking boxes and not features actually being used, so it was lower priority.

Now that they are being used, maybe thre would be more impetus for providing support.

I am afraid that they are not being used in Bioshock. The game probably just peeks the main OGL version number, and doesn't care about the extensions it actually needs, and can perfectly run on current Mesa drivers. Sad.

[MadCatX]

Alright, well I have a 290 on Mesa 10.5, X 1.17, kernel 3.18.6. Games like Civ 5, TF2, Portal 2, Metro LL, and Witcher 2 all work.

Here are the dumps if anyone cares.

Only really obvious thing is Assertion Failed: m_nNumRegisteredWriters == 0, which is common to every crash, but that isn't very informative since Googling it turns nothing up and its an assertion fail in the Steam client.

I already have libstdc++ removed from the Steam Runtime, hence how the other titles work.

Of course, I'm on Arch, using an r9 290, so I expect to be in the wild west here, but it doesn't change how reporting these issues is the worst kind of maze.

How exactly did you determine that this issue has anything to do with the GPU drivers? This seems to be a bug in the Steam runtime. If you go through what Google returns, you'll find that it isn't specific to Linux. Here (http://pastebin.com/YY012jA1) is one of the examples where it crashes on a Windows machine.

[zanny]

How exactly did you determine that this issue has anything to do with the GPU drivers? This seems to be a bug in the Steam runtime. If you go through what Google returns, you'll find that it isn't specific to Linux. Here (http://pastebin.com/YY012jA1) is one of the examples where it crashes on a Windows machine.

Some context on that would be nice - I don't think the assert failing is what actually crashes all these games. I think that just usually happens but is not logged except in crash logs. At that, I've tried replacing the steam runtime with the unofficial native runtime repo on Arch, and all that did was make Dota2 crash the GPU driver and freeze the system =P.

I think it is something else, but there is very little way to debug this stuff - if I can't get the game to start gallium tools won't help, and none of these titles run at all when invoked via their executables without being called by steam to the best of my knowledge and attempts, nothing shows up in my dmesg, journalctl, or xorg logs, so I'm firmly stuck hoping the steam crash dumps are informative but they really don't seem to be in this case.

Its not like its free software so I can just valgrind or gdb the thing with symbols exported. Its a black box and I'm stumped on how else to approach this / these bugs.

[Nobu]

Don't have Torchlight so I can't test that one, but most of the games I have run fine if I run steam with STEAM_RUNTIME=0. That's with either Radeon HD 7770 or R9-270X, on Arch x64.