AMDVLK Driver Gets Fixed For Rise of the Tomb Raider Using Application Profiles

[Almindor]

This is exactly the reason why I say two drivers are bad. Game developers just don't want to deal with this kind of crap. The whole profiling for games in drivers idea is ludicrous.

[F.Ultra]

so this the game's fault and not of the driver or the standard ?

It's probably just that the geometry support in AMDVLK is in beta so they disabled it by default but then they found out that this particular game requires it in order to work.

[andrei_me]

In this case though the application is not doing anything wrong.
Correct?

That's my doubt too, we have DriConf to deal with faulty applications in Mesa, so I don't see any issue with application profile if they do something faulty or need a perf boost

[agaman]

That's my doubt too, we have DriConf to deal with faulty applications in Mesa, so I don't see any issue with application profile if they do something faulty or need a perf boost

What they did in AMDVLK is not something generic like DriConf where you can enable/disable features based on config for any program without changing the driver.
They have harcoded inside the pipeline compiler a check that changes the shader compiler behaviour when the program it's Rise of the Tomb Rider...

[marek]

Yes. All Mesa DriConf per-app options can be overridden by users and apps freely.

[ElectricPrism]

Let the debate begin!

Idealism vs Practicality

A fictional utopia vs the real world

[dungeon]

I don't see anything wrong with that, as their existing userbase is likely more familiar with just renaming executables in case of a problem. I don't ignore majority, ever. So, down to that average 99% Joes of steam, I would mention it even psichologically also, down to the end-user multitasking or multi/whatever leads to poor and broken brans The less info you have, better for you - the more knowledge you have is the best for you. Where you is not you but right that one average Joe.

It is probably all about - should we promote bugzillas or should we do our job right and stright to the average Joe end-user also with GUI on top of it with glorious clickete click invented and promoted so much after DOS? Since Windows 95 appeared, industry majorly agreed that this is the way to go.

These are still unresolved questions, which also does not have tendency to ever be resolved... just like "should this door be open or closed" or better "should this windows be open or closed" or the best is with major W "should this Windows be open or closed"... and answer there is "why not both, you know doors and windows feature double standards, could be like this or like that - is that an user freedom or not?" Freedom where again user is an average Joe and not you, Joe the user is the one by definition always stupider than you

Bottom line there seems to be - freedom for doors and windows Should penguins ever adopt to warmer climate?

[zoomblab]

Application specific workaround? That is ugly. It shouldn't be so.

[M@yeulC]

Question: isn't it a serious problem that a userspace application (the game, compiler toolchain, libdri and whatnot) is able to hang the GPU? Shouldn't that be recoverable, or protected by some application-specific context, enforced by a MMU?

[Soul_keeper]

So how long before this game works with a stable version of mesa/llvm/kernel without having to hack it ?