RadeonSI Adds Zeroing vRAM Workaround To Help Rocket League Players

[Guest]

It's not required by the APIs. The driver has the send commands to the GPU to do this. That adds latency and reduces performance.

Well, I see this kind of "corruption" in almost all applications when they create a new window, menu or even a popup when using AMD GPU drivers. I think the tiny number of clear commands sent to the GPU isn't that big a problem vs this visual artifacts that appear everywhere.

I use Intel iGPUs and even an NVIDIA GPU several years ago, and never faced this problem - they're definitely clearing the memory and it doesn't seem to affect their performance.

[OneTimeShot]

Hmm... those "uncleared memory textures" look like data from other screens and programs running on the system (the desktop, etc).

Thats *really bad* and a security issue. What if that was a screenshot of a root terminal?

[Guest]

If you want something zero'd, you do it manually per spec. The driver zeros everything now, against spec, and every call. It's not a huge performance hit really, but it's spec to do it yourself, and the game breaks spec and this is all shitty to even have to look into.

Mesa shouldn't have to fix this. They also don't want to, as it hurts performance and can break stuff later. But they do to make sure it works better now, so it's whatever. Good it works, bad any of this is required.

Ah, so it's bascially applications not calling glClear even though they're supposed to? Not the fault of the driver then, as I assumed. Although if the application can read the contents of old buffers, that's still a security problem that should be handled by the driver (even if it's not according to spec).

[curaga]

I remember seeing this years ago. Besides being cleared by Windows, OpenBSD also turned out to clear those.

[JonathanM]

While I agree that applications should stick to the specs and clear vram themselves, I consider being able to read application window fragments this way to be a severe security issue. In a desktop/workstation setting the driver should clear it, at least when the memory can contain data from other applications.

[BlueJayofEvil]

I see this happen all the time when opening new browser windows. I've seen parts of video frames and similar data from previously closed apps (usually media players or other browser windows) show in the garbled mix of corruption. It's surprising how images that should've been cleared out of my memory hours ago can be lingering in vRAM.

[aufkrawall]

I see this crippled output from previous applications whenever I start games with XWayland...

Shouldn't this be fixed too?

That's an xwayland issue no one seems to care enough about to fix it.

[cybertraveler]

From reading this conversation my Official CyberTraveler Summary is!...

1) The dev ultimately responsible for this bug probably didn't see it occur on his dev machine because his dev machine happened to use a driver that happened to zero new buffers.

2) From a security standpoint, it would make more sense if the spec zero'd buffers by default.

3) From a performance standpoint, it would make more sense to keep the specs as-is (do not zero by default).

4) Practically speaking, the secure-by-default approach should be taken (from point 2) and performance critical apps should be able to disable this zero-buffers-by-default behaviour. Though, this point is debatable.

5) Meanwhile, back in the land of the real... Rocket League / it's engine is still broken and Mesa is now uglier for having to work around a bug in the app/engine.


Next week I will be presenting my Official CyberTraveler Summary of systemd vs non-systemd init systems. Before making a start on that summary list I am awaiting a response from a mathematician I hired to check that there are enough positive numbers in the universe capable of enumerating this list. (It's been a month and he still hasn't got back to me.)

[davidbepo]

It's not required by the APIs. The driver has the send commands to the GPU to do this. That adds latency and reduces performance.

okay, thanks for the answer

[GrayShade]

Every OS will zero memory allocated by the applications in order to prevent them from reading old data from other programs. VRAM should be handled the same way. There's no spec saying that memory should not be cleared.