RADV Starts Off Another Exciting Week Of Development

[ResponseWriter]

I think you have to compile the kernel by yourself to get GC1 working, right? I also think HD79xx still don't work so well?

Run this command:

zgrep AMDGPU /proc/config.gz

If you see CONFIG_DRM_AMDGPU_SI=y then your kernel has support for your card without recompiling (likely only kernel 4.9 and later).

Can't speak for SI but AMDGPU works well for me on a 290 (CIK). I wonder if CIK ought to be on by default by now, unless some issue I don't know about.

The work going towards this driver is much appreciated but won't it somewhat go to waste once the official driver from AMD comes out?
... or is it about having a good placeholder until that happens?

There are at least two OpenGL implementations (Clover, AMDGPU-Pro) and you can choose to install one or the other. Don't see why there can't be two Vulkan implementations, except maintenance burden.

[nomadewolf]

the official driver in the official linux kernel is the RadeonSI+AMDGPU driver.

the closed source driver IS NOT OFFICIAL.

So, by official, you mean, in the Linux Kernel. Is that it?

[bridgman]

the official driver in the official linux kernel is the RadeonSI+AMDGPU driver.

the closed source driver IS NOT OFFICIAL.

Hah... that's funny to hear after years of "the open source driver is not official because I don't download it from amd.com and because some non-AMD developers contributed to it".

[Herem]

bridgman are you able to share anything new about the progress on opening up the AMD Vulkan driver?

[shmerl]

bridgman are you able to share anything new about the progress on opening up the AMD Vulkan driver?

And if not details, may be some rough ETA for completing this tedious process?

[droste]

[...]
I wonder if CIK ought to be on by default by now, unless some issue I don't know about.
[...]

HDMI audio is not fully working until DAL/DC lands. That's at least what AMD devs were saying.

[bridgman]

HDMI audio is not fully working until DAL/DC lands. That's at least what AMD devs were saying.

... or until we (or someone) port the audio code from radeon to amdgpu (for SI).

[pal666]

At this rate, AMD open-sourcing their Vulkan driver is going to be near-pointless.

pointlessness works both ways

[bridgman]

best example for this they write to you about a AMDGPU-pro bug and you write back they should write a AMDGPU/Radeon bug-report because the time-pipe process of bug-fixing in the closed driver is "2-4 month" at minimum. and people are really confused by this ... the open driver do have bug-fix pipe timeline of maybe 1-2 days. I had such bug's and bugrepots and bugfixes in the past.

If the problem is in common code between open and hybrid drivers (which initially seemed to be the case assuming you are talking about discussion with Amarildo) then why not use the open source development process to fix it ?

Please don't use quotation marks ("2-4 month") to suggest I said something that I did not. As always, feel free to quote what I actually say though, as long as it is not taken out of context.

For example, quoting me saying "yes" but implying it was my response to a different question would not be acceptable

[indepe]

... or until we (or someone) port the audio code from radeon to amdgpu (for SI).

Generally it seems that all the code anyone is asking for, already exists in some place(s)....