Why can't I run a fglrx and a mesa/radeon X server at the same time?

[Detructor]

Hi,

why can I run two (or more) Xservers with one graphic driver (on :7 :8 ...) but can't run a fglrx based Xserver on :7 and a radeon based server on :8?

[bridgman]

The fglrx and radeon driver stacks use different GL libraries, and absent an ICD-type mechanism they can't co-exist.

Not sure if both kernel drivers can be restricted to only run on one of the two devices - I think they can but not 100% sure. Anyways, the GL lib conflict is probably what's stopping you.

[nanonyme]

The fglrx and radeon driver stacks use different GL libraries, and absent an ICD-type mechanism they can't co-exist.

Not sure if both kernel drivers can be restricted to only run on one of the two devices - I think they can but not 100% sure. Anyways, the GL lib conflict is probably what's stopping you.

Curiosity, cat. Will you be able to do that on amdgpu stack with open and closed variants?

[Detructor]

The fglrx and radeon driver stacks use different GL libraries, and absent an ICD-type mechanism they can't co-exist.

Not sure if both kernel drivers can be restricted to only run on one of the two devices - I think they can but not 100% sure. Anyways, the GL lib conflict is probably what's stopping you.

Hey, thank's for the fast answer, but what has the International Classification of Diseases to do with this?

[bridgman]

You asked what the problem was -- AFAIK the purpose of the International Classification of Diseases is to analyse symptoms and determine a root cause. Seems fitting to me.

There's also Installable Client Driver, but that's boring.

[agd5f]

More importantly, they each have different kernel drivers that expect to have exclusive access to the hardware.

[bridgman]

My interpretation of OP's question was whether they could run fglrx on one card and radeon on another card. Not 100% sure that is what they meant though.

[Detructor]

My interpretation of OP's question was whether they could run fglrx on one card and radeon on another card. Not 100% sure that is what they meant though.

that explains why I can't make sense of your answer. To clarify: My question was: why can't I run two Xservers with a different GPU driver on each Xservers? Only one GPU is involved here.

[bridgman]

Ahh, OK. If only one GPU then you can only have one kernel driver, and so only one driver stack (fglrx or radeon).

[Detructor]

Ahh, OK. If only one GPU then you can only have one kernel driver, and so only one driver stack (fglrx or radeon).

then I don't understand why I need to specify which driver I want to load (radeon or fglrx). If only one driver can exist on a system, shouldn't the XServer just query the kernel for which driver is currently loaded?