Radeon Linux Graphics Stack, RadeonSI Shaders Update From FOSDEM 2018

[microcode]

Glad to see AMD acknowledging the existence and popularity of RADV.

[LeJimster]

Funny thing is "immediate support for new GPUs" is kinda worthless if it performs like junk. The last time I checked RADV was performing much better on Vega than AMDVLK.

I think the slide is trying to say," hey devs target our open source driver not RADV, we got the tools and will offer the best support going forward."

It'll be interesting to see the competition going forward. =)

[Marc Driftmeyer]

Funny thing is "immediate support for new GPUs" is kinda worthless if it performs like junk. The last time I checked RADV was performing much better on Vega than AMDVLK.

I think the slide is trying to say," hey devs target our open source driver not RADV, we got the tools and will offer the best support going forward."

It'll be interesting to see the competition going forward. =)

Last I saw RADV doesn't run on Windows and Vulkan Vega on Windows flies. Wanna wager where the bulk of their sales revenues streams exists? It's not RedHat and other Linux distros.

[droste]

OpenCL 4.5 conforma- Dang it!!!

I won't appear as public typo reporter anymore due to complaints.

There's always someone complaining. No matter what you do

[LeJimster]

Last I saw RADV doesn't run on Windows and Vulkan Vega on Windows flies. Wanna wager where the bulk of their sales revenues streams exists? It's not RedHat and other Linux distros.

No doubt. But I was making comparisons between AMDVLK and RADV. Also these slides clearly are focused on Linux, not Windows... Even if that's where the money lies right now.

I'm not sure what the performance delta is between RADV and the closed source Vulkan driver, but I had an excellent experience completing Doom with RADV @ 1440p Maxed on my Vega 56. Amazing really considering the bulk of the work is done by 2 guys.

[pal666]

No. The question is, why not extract AMDOGL/radeonsi driver from Mesa and make it cross-platform like AMDVLK?

because they never extracted radv to make amdvlk

[pal666]

All of the red points for RADV are useless data points.

all your posts are useless sposts

AMD contributions - well, AMD could have contributed since long ago and chose not to.

amd windows team can't contribute to radv, which btw didn't exist when they started amdvlk, and which is slower than amdvlk, but you chose to post bullshit

[pal666]

Glad to see AMD acknowledging the existence and popularity of RADV.

presentation author is mesa developer. if you think there are mesa devs not acknowledging existence of radv, stop abusing drugs

[pal666]

Amazing really considering the bulk of the work is done by 2 guys.

amazing amount of idiots who believe this. most of work on radv was done by amd team(you know, radv does not work without kernel driver or shader compiler), non-amd work consisted of merging intel anv with amd radeonsi. and community contributors consist not only of 2 guys, but of several others, including valve team

[xxmitsu]

No doubt. But I was making comparisons between AMDVLK and RADV. Also these slides clearly are focused on Linux, not Windows... Even if that's where the money lies right now.

I'm not sure what the performance delta is between RADV and the closed source Vulkan driver, but I had an excellent experience completing Doom with RADV @ 1440p Maxed on my Vega 56. Amazing really considering the bulk of the work is done by 2 guys.

It's not just two guys. Let's not forget the contributions backed by Valve.

https://cgit.freedesktop.org/mesa/me...hor&q=Pitoiset

contributors consist not only of 2 guys, but of several others, including valve team

exactly