David Airlie Preps Latest Round Of RADV Radeon Vulkan Patches

[phoronix]

Phoronix: David Airlie Preps Latest Round Of RADV Radeon Vulkan Patches

David Airlie is prepping the latest round of work to the open-source RADV Radeon Vulkan driver...

http://www.phoronix.com/scan.php?page=news_item&px=RADV-December-2016

[LeJimster]

At this point I hope AMD are focusing fully on the DAL/DC rather than open sourcing their own Vulkan driver, while Dave and Bas focus on RADV which is shaping up nicely. It's going to be an exciting next 6-12 months for us Open source Radeon users if this all starts landing and maturing.

[boxie]

It's going to be an exciting next 6-12 months for us Open source Radeon users if this all starts landing and maturing.

It certainly is - and it has been quite an exciting last 18 months too

[agaman]

At this point I hope AMD are focusing fully on the DAL/DC rather than open sourcing their own Vulkan driver, while Dave and Bas focus on RADV which is shaping up nicely.

Yes. And open sourcing their OpenCL implementation it's important too. Their Vulkan implementation can wait a bit more, until all the other things are done. Meanwhile we have radv. (Maybe it can reach a point where radv performs better than the amdgpu-pro Vulkan, so we won't need it anymore... )

[ThrowAway3000]

Can AMD show the AMDGPU-PRO Vulkan code to RADV developers? Make them sign NDA if it's top-secret, the idea is to speed up RADV development without waiting until AMD lawyers can greenlight open sourcing Vulkan.

[agd5f]

At this point I hope AMD are focusing fully on the DAL/DC rather than open sourcing their own Vulkan driver, while Dave and Bas focus on RADV which is shaping up nicely. It's going to be an exciting next 6-12 months for us Open source Radeon users if this all starts landing and maturing.

It's not either or. There are separate teams working on display and vulkan. They are largely independent.

[Azpegath]

Have they fixed the issue with colliding symbols and LLVM? I couldn't run vkcube, vkinfo or any of the other vulkan demos because of some weird LLVM symbol collision.

See here: https://lists.freedesktop.org/archiv...er/130774.html

[darkbasic]

Yes. And open sourcing their OpenCL implementation it's important too. Their Vulkan implementation can wait a bit more, until all the other things are done. Meanwhile we have radv. (Maybe it can reach a point where radv performs better than the amdgpu-pro Vulkan, so we won't need it anymore... )

AMD doesn't listen on this point. They probably think nobody needs OpenCL.

[pal666]

At this point I hope AMD are focusing fully on the DAL/DC rather than open sourcing their own Vulkan driver, while Dave and Bas focus on RADV which is shaping up nicely

why do you hope for duplication of effort and non-working driver?

[pal666]

And open sourcing their OpenCL implementation it's important too. Their Vulkan implementation can wait a bit more, until all the other things are done. Meanwhile we have radv.

meanwhile you have clover, so why not the other way around?