AMD Catalyst 9.8 Delivers New Kernel Support

[energyman]

use opengl instead of xv. Nicer picture, better colours and the cpu load is neglible. So what is the problem?

[RealNC]

I know that, but there is really no other better option?

Nope. However, gl video works quite OK, even for HD content. And it's tear-free except when it's windowed while using compositing (full screen while using compositing is OK.)

I can compile lastest open source radeon and drm driver, but it does not have 3D support.
I can't use radeon with 3D visual effect of compiz or Kwin.

Xv on the radeon driver should work OK. I think.

[d2kx]

Okay, but don't get your hopes too high :/ although tropics improved quite a lot, but that because I've got errors in rendering :/ whole benchmark was missing water and screen was flickering before I've turned aa on so something is wrong with this drivers.

The improved performance is for real, the OpenGL code has been optimized. But Michael's statement in the news, that "there is a fix for water rendering in the Unigine Tropics benchmark" does not apply to this release of the AMD Catalyst release but to... another.

[phtpht]

??? wtf?

emmm... bkl is in linux since it can do smp. So at least 2.0. <Preemptive kernel? of course! for ages.

well bkl can be preempted or not. the same goes for rcu, which you did not reply to.

and as for general preemption i recall fglrx DID NOT like it and DID spew shitty messages into dmesg (and maybe even worse things).

and by the way was the module_license problem resolved?

[energyman]

Symbol: PREEMPT_VOLUNTARY [=y] │
│ Prompt: Voluntary Kernel Preemption (Desktop) │
│ Defined at kernel/Kconfig.preempt:19 │
│ Depends on: <choice> │
│ Location: │
│ -> Processor type and features │
│ -> Preemption Model (<choice> [=y])

and preempt:rcu:
│ This option reduces the latency of the kernel by making certain │
│ RCU sections preemptible. Normally RCU code is non-preemptible, if │
│ this option is selected then read-only RCU sections become │
│ preemptible. This helps latency, but may expose bugs due to │
│ now-naive assumptions about each RCU read-side critical section │
│ remaining on a given CPU through its execution. │
│

so I stay with no. Because rcu bugs can result in data corrpution. I like the stuff on my harddisks.

[haplo602]

opengl version

hmm, would be good to include opengl/glsl version in the driver news, so I have all on one page and don't need to dig around thanks

[Melcar]

Anyone else getting massive graphical distortions on the Tropics demo?

[RealNC]

Almost two days now and still "not found" for the release notes.

Can it be that nobody in AMD even gives a damn?

[phtpht]

Symbol: PREEMPT_VOLUNTARY [=y] │
│ Prompt: Voluntary Kernel Preemption (Desktop) │
│ Defined at kernel/Kconfig.preempt:19 │
│ Depends on: <choice> │
│ Location: │
│ -> Processor type and features │
│ -> Preemption Model (<choice> [=y])

and preempt:rcu:
│ This option reduces the latency of the kernel by making certain │
│ RCU sections preemptible. Normally RCU code is non-preemptible, if │
│ this option is selected then read-only RCU sections become │
│ preemptible. This helps latency, but may expose bugs due to │
│ now-naive assumptions about each RCU read-side critical section │
│ remaining on a given CPU through its execution. │
│

so I stay with no. Because rcu bugs can result in data corrpution. I like the stuff on my harddisks.

So you more or less confirm what I wrote earlier.

[phtpht]

Almost two days now and still "not found" for the release notes.

Can it be that nobody in AMD even gives a damn?

Let's just assume their team is terribly undermanned.