Radeon ROCm Updates Documentation Reinforcing Focus On Headless, Non-GUI Workloads

[qarium]

I'm not saying it's a good thing that they go back to closed source development but that seems to be happening if they unify the stack under AMDGPU-PRO and not open source its parts fully. If they go closed source tho they are done. Finished. There will be no reason to buy it since Nvidia is way better anyway. Most people including me switched to AMD from Nvidia just to support their open source efforts. But since I can't actually get anything done with their broken drivers I'm kinda starting to regret that decision.
Going open source should be a good thing, investing in it, engaging with community, moving things forward, having a clear roadmap, addressing users issues, being responsible. But they kinda open sourced it and divested from linux at the same time.

i don't know why you have thoughts like this and you do not read phoronix.com news...
in the phoronix.com news you can read that more and more workstation features and performance hacks come into mesa
to support the last of the old OpenGL/workstation market users and soon AMD will replace the closed source openGL stack in the AMDGPU-PRO with the MESA version. but there will still be a AMDGPU-PRO driver.

the reason for still have a AMDGPU-PRO driver is this: you can not support these Workstation needs with MESA/AMDGPU upstream. one reason is because the customer want certification that it works and it stay this way. because of thix they pick up the opensource driver do certification and put it into amdgpu-pro and let the pro customers use it in the same way they used the old closed source driver. yes this is fully Idiotocracy comlete stupid but for a company like AMD you do what your cusomer wants... or you have no "sale" of hardware.

"Most people including me switched to AMD from Nvidia just to support their open source efforts."

not all people are like this. i am in this group to. but be sure the PRO/workstation customers they want certification means proof that their tools work and if this is not the case they don't buy AMD hardware. they do not buy hardware only to support opensource.

"But since I can't actually get anything done with their broken drivers I'm kinda starting to regret that decision."

see the workstation customers are not stupid like that they want certification means proof that it does what they want or else they do not buy AMD hardware.

[qarium]

I'm not sure where the idea about going closed source came from - certainly not from us.

yes ... very strange... how can people think this...

[acowley]

The messaging about not supporting GUI apps only applies to the ROCm *stack* (see previous post for more specifics), not the individual ROCm *components*.

Perhaps the GitHub Issues with ROCm OpenCL not working for GUI apps should have been moved to the right component rather than closed.

[qarium]

a couple of closed source components (our workstation GL driver

can you tell us the status of replacing the closed source workstation GL driver with the MESA opensource driver in AMDGPU-PRO?

for sure there is improvement. but what do you think is the status of this ?

[bridgman]

Perhaps the GitHub Issues with ROCm OpenCL not working for GUI apps should have been moved to the right component rather than closed.

A couple of people have hinted at this but I haven't been able to find any specific examples yet. Have you seen this happen anywhere ?

In general this should not be happening, although it's possible that a few of our people have not yet understood that we are shipping ROCm all the way up to OpenCL with our graphics stack today and will be shipping more in future. If so, we can fix that pretty quickly.

EDIT - I did find some posts last night pointing to tickets which were closed citing this message. I don't think we have found them all yet but we do plan to re-open those tickets ASAP.

[bridgman]

can you tell us the status of replacing the closed source workstation GL driver with the MESA opensource driver in AMDGPU-PRO?

for sure there is improvement. but what do you think is the status of this ?

Same as before - we talked about it, that's all. People are telling me that Microsoft's port of Mesa to Windows will cover native Windows as well as LSW/WSL (with the implication that will address the problems) while my understanding had been that it was only for LSW/WSL.

[unis_torvalds]

If they can't get enough resources on ROCm to keep it working, then move development resources to the "shared" OpenCL implementation.

Except that ROCm does work. Just not for users like you and I. Hence the documentation update.

It's not like anyone is clamouring for an AMD-specific compute stack...

Except maybe 99% of their revenue-generating linux userbase: in datacentre and HPC.

[qarium]

we talked about it

why not put a RDNA and a CDNA chip on consumer graphic card?
this way people can use graphics and compute at the same time.
to do a chiplet design for RDNA is very hard but if you do hybrid RDNA+CDNA the chiplet design would be very easy.
many people here would not have these problems with using compute and graphics at the same time with a card like this.
so they can use MESA on RDNA and ROCm on the CDNA part.

[gururise]

Same as before - we talked about it, that's all. People are telling me that Microsoft's port of Mesa to Windows will cover native Windows as well as LSW/WSL (with the implication that will address the problems) while my understanding had been that it was only for LSW/WSL.

Will we see support for RDNA2 (and even RDNA1) on ROCM this year? All we want to do it run accelerated tensorflow with our new AMD 6000XT cards!

[smitty3268]

why not put a RDNA and a CDNA chip on consumer graphic card?

Huh? If they want graphics they wouldn't have removed it from CDNA in the first place.

Not to mention, there's currently a massive shortage of available chips. Doubling up the # required for a card seems like it would be a particularly poor decision right now.