Mesa 20.2 Released With RADV ACO By Default, Initial RDNA2 Graphics Support

[pal666]

I could try bisecting from master, but then how long am I gonna take?

several runs

[pal666]

And now I wonder how I can verify this when testing some games.

it looks like you are wasting your time on making your computer slower(wasting its time)

[smitty3268]

I could try bisecting from master, but then how long am I gonna take?

Probably less than 2 years.

Seeing as how nobody else seems to run into this, your problem is likely never going to get fixed if you can't even submit a bug report.

[tildearrow]

Probably less than 2 years.

Seeing as how nobody else seems to run into this, your problem is likely never going to get fixed if you can't even submit a bug report.

Because I know my bug report will be marked as duplicate of another bug report that gets no fix so far.

Somebody did reproduce this, and then they fixed it and claimed to have fixed it.
However, it still happened.

[Venemo]

...as long as you can tolerate your card hanging when trying to record the screen, or using Mesa 19.0.

It's 2020 and I still have not updated to Mesa 20 from the moment I hit that bug...

Sorry, I have no time to deal with that, OK?
I am not a graphics driver developer and bisecting is extremely difficult because Mesa devs use branches for every release.
I could try bisecting from master, but then how long am I gonna take?

Sounds like a very annoying issue. Please open a bug report where you tell us what hardware you use exactly, with what distro, and how exactly to reproduce this issue.

I have personally not experienced this but I think maybe you used some other program.

[R41N3R]

it looks like you are wasting your time on making your computer slower(wasting its time)

Well, you could have said this at some point as well for DXVK ;-) And now the Vulkan translation is performing extremly good. The games I'm planning to run with Zink would be older anyway, so the difference is probably only that I'm wasting a little more power to render 60FPS, but it would not make it slower.

[pal666]

Well, you could have said this at some point as well for DXVK ;-)

you are a bit confused here. to make analogy with dxvk you have to compare dx on vulkan versus native dx. and surely native dx will be faster, we just don't have luxury of having it available on linux(except gallium nine, so i hope valve will add support for it in proton)

And now the Vulkan translation is performing extremly good. The games I'm planning to run with Zink would be older anyway, so the difference is probably only that I'm wasting a little more power to render 60FPS, but it would not make it slower.

so we are in agreement that you are wasting both your and your computer resources for no benefit?

[R41N3R]

you are a bit confused here. to make analogy with dxvk you have to compare dx on vulkan versus native dx. and surely native dx will be faster, we just don't have luxury of having it available on linux(except gallium nine, so i hope valve will add support for it in proton)so we are in agreement that you are wasting both your and your computer resources for no benefit?

I agree, I have no issues wasting some compute power :-) My system is mostly idle anyway and I'm interested in testing some new technology.

If you believe that Gallium nine might be faster than DXVK at some point (as implied by native) then you should give Zink the same chance. It is just another Gallium driver with a Vulkan output. Even Gallium nine should be able to use Zink. The luxury of having radeonsi/iris is not available for every GPU, so it might be quite important in the future where only a good Vulkan driver needs to be developed and Zink provides the OpenGL driver almost for free.

[pal666]

If you believe that Gallium nine might be faster than DXVK at some point (as implied by native)

i'm pretty sure it was always faster than dxvk(d9vk part).

then you should give Zink the same chance.

sigh. you are ignoring my explanations. you can't become faster by stacking additional converters, the only road is down

It is just another Gallium driver with a Vulkan output. Even Gallium nine should be able to use Zink.

in this case dxvk will be faster, because dx->vulkan is strictly less work than dx->gallium->vulkan

The luxury of having radeonsi/iris is not available for every GPU, so it might be quite important in the future where only a good Vulkan driver needs to be developed and Zink provides the OpenGL driver almost for free.

surely for some imaginary future gpu with good vulkan driver and bad/non-existent opengl driver zink will be helpful. but how is that relevant to your tests with radeonsi? are you silently hoping for radeonsi's demise? what makes you think radv will still be alive then? in case it's not clear, radeonsi shares frontend with zink and backend with radv, so it's pretty hard for it to die while other are alive