A New Patch For Radeon DRM Power Savings

[phoronix]

Phoronix: A New Patch For Radeon DRM Power Savings

While OpenGL acceleration and GPU-assisted video playback are often most viewed as the areas that are severely lacking for the open-source Linux graphics drivers in comparison to what the binary-only ATI/NVIDIA drivers offer, another area that has not yet caught up to speed with the binary competition is power management. For years (going back to 2005) AMD has implemented PowerPlay support in their fglrx driver for dynamically clocking the GPU and memory clocks along with adjusting the voltages accordingly, based upon the user's input and then later generations of PowerPlay are more dynamic in nature...

http://www.phoronix.com/vr.php?view=NzcyNw

[sundown]

Oh please, pelase, pelase... power management!

[mcgreg]

generally it sounds very interesting to me since I have a laptop with radeon 4650.
even if the features aren't complete yet, is there any possibility of damaging the hardware? some graphical glitches on the screen wouldn't be the problem for the beginning. This actually would be my only concern at the moment.

Gregory

[Louise]

It is interesting reading the code. It is not complicated in anyway, but having the knowledge and understanding and being able to write the code seams to be challenge.

I imagine it requires an overall understanding of how the drivers works.

Respect.

[Zajec]

generally it sounds very interesting to me since I have a laptop with radeon 4650.
even if the features aren't complete yet, is there any possibility of damaging the hardware? some graphical glitches on the screen wouldn't be the problem for the beginning. This actually would be my only concern at the moment.

I believe the worst thing that may happen is lock up, for example if 50% of your default engine clock is too low value. As listed in TODO we need to read AtomBIOS data table for real values and eventually count something ourself as well.

[Zajec]

It is interesting reading the code. It is not complicated in anyway, but having the knowledge and understanding and being able to write the code seams to be challenge.

I imagine it requires an overall understanding of how the drivers works.

That's really perfect summary. Patch itself is quite simple but I had many problems to get it working. First I had to learn about timers (there are many kinds of timers in kernel, some with limitations). Experienced about 50 locks up when fighting with that one. Then I tried to use ring info to determine GPU load. This didn't work so I had to switch to fence (again, learning what fence is...).

But hopefully I have now more idea about driver stuff and will able to do rest faster

[Louise]

That's really perfect summary. Patch itself is quite simple but I had many problems to get it working. First I had to learn about timers (there are many kinds of timers in kernel, some with limitations). Experienced about 50 locks up when fighting with that one. Then I tried to use ring info to determine GPU load. This didn't work so I had to switch to fence (again, learning what fence is...).

It is scary to think about how much time that goes into coding that the programmer doesn't get credit for

But hopefully I have now more idea about driver stuff and will able to do rest faster

I read the reply on the mailing list. What will you next step be?

Figure out the API to ask if the GPU is idle?
Only downclock if vblank?

[Chewi]

Does this stuff work with KMS?

[madman2k]

Does this stuff work with KMS?

actually it only works with KMS

[Chewi]

actually it only works with KMS

Great. That just leaves Zaphod mode, which wasn't working the last time I checked. Then I can finally switch!