KHR_no_error Support Merged In Mesa For Potentially Helping CPU Usage

[phoronix]

Phoronix: KHR_no_error Support Merged In Mesa For Potentially Helping CPU Usage

Timothy Arceri at Valve has recently been working on OpenGL KHR_no_error support while now that initial code has been merged into Mesa 17.2-devel...

http://www.phoronix.com/scan.php?pag...o-Error-Merged

[davidbepo]

will test it when i can

[davidbepo]

my little test shows no performance improvements and the game worked OK
note: this test is on only one game and one situation so there could be improvements in other games

[VikingGe]

What did you test, and was it CPU-bound at all?

I wonder if this would benefit Hitman in its current state, Hitman is one of those games that are somewhat playable on my old 1090T on Linux, but I wouldn't mind a few more FPS.

[Azpegath]

It doesn't really matter if the game is "error free" or not, it just stops validating state while running (assuming that the developer has already tested it before releasing), just like with Vulkan and its validation layers.
Previously the game would probably have crashed when the state machine (or a single call) failed to validate, now it might instead show something incorrect on the screen (or any other undefined behavior).

[nhaehnle]

It's actually the other way around: Previously, the game would have shown something incorrect - perhaps some objects would be missing because the corresponding draw calls were dropped because of errors.

With no_errors enabled, these errors wouldn't be detected, and the driver would happily submit a draw call that has some invalid state in it. That might lead to crashes.

[donkeykong]

I can see the pcsx2 emulator will try to benefit from this (see the link). Don't know if it'll make a big difference, but as emulators tend to be very cpu intensive, it could give a small boost.