DRIConf Is Still A Mess & Leaves A Lot To Be Desired

[haagch]

Wait, driconf is developed in a svn repository on sourceforge and the last commit still was in 2007? And the last release was still 2006?

At least the configuration format is stable...

[haagch]

But then I would have to learn how to use the old gtk2 python bindings...

Though the xml implementation would probably be there and better than what I hacked together with ElementTree. Hm...

[curaga]

Driconf is one of the "nobody owns it, send patches and magic will happen" projects. If you want that to happen, send patches

[haagch]

This is an old topic but the question remains: Is driconf even actively developed?

I was missing e.g. the functionality that was introduced here:

http://lists.freedesktop.org/archives/mesa-dev/2014-May/060131.html

where with dri3 offloading you can choose a binary name and set up on which gpu it should be rendered.

Every time finding the ID_PATH_TAG of your GPU, and editing an XML file didn't look very convenient to me, so I made a small tool with pyqt.

Source: https://github.com/ChristophHaag/gpuchooser
(It's not terribly good yet, so back up your ~/.drirc before using it, if you have any modifications there)

But will driconf upstream getting something like this?

[Hamish Wilson]

Yes, it use lm_sensors. Today I modified some code responsible for gathering temp info from sensors, so maybe now it will work.

Cheers, it now works like a charm, with the right privileges.

[Vim_User]

It's all dynamic - the GUI generates the buttons by asking the driver what options it has. If you want to see the XML yourself, type "xdriinfo options 0", it even includes translations in a few languages.

So technically it doesn't matter the code is from 2006.

Oh, I have done that already, as I was curious and had a look at the code. I get the astonishing amount of 11 options here on my HD6870, except setting vblank none of them is useful (who the hell needs options like ""A post-processing filter to remove the blue channel").
That seems to be far from exposing all available options.

[marazmista]

It looks nice and is conveniently in the AUR, but it does not fully seem to probe my card properly:



EDIT: It needs root privileges - but still can't see my temperature, which is strange since it uses lm_sensors.

The clocks and volts data are in debugfs so it need to be mounted:

mount -t debugfs none /sys/kernel/debug

and for fstab:

debugfs  /sys/kernel/debug  debugfs  defaults  0  0

Yes, it use lm_sensors. Today I modified some code responsible for gathering temp info from sensors, so maybe now it will work.
I can provide support only for Radeons (I have SI in PC and R730 in laptop).

For the DRI, there is only these options?

http://dri.freedesktop.org/wiki/ConfigurationOptions/

[curaga]

So, do you think we should wait another 7 years until someone implements the available options?

It's all dynamic - the GUI generates the buttons by asking the driver what options it has. If you want to see the XML yourself, type "xdriinfo options 0", it even includes translations in a few languages.

So technically it doesn't matter the code is from 2006.

[schmidtbag]

Just to be clear, guys, DRIConf has new options/settings added to it all the time. The program just reads in xml files to generate the UI. I'm sure it's that process that hasn't changed in forever, no doubt because the devs feel like it performs well enough for their needs and they are focused on actually making the hardware work rather than beautifying the GUI.

True, but it doesn't mean it isn't still lacking in features, such as the ones mentioned in the article. Also, there could still be better ways of approaching some of the options.

[grigi]

I think what we need now is a protocol, sitting behind a dbus service (somewhat like what XRandR does). This way you can separate the job of writing clients and services, meaning developers can work on what is good for them. e.g. desktop guys can worry about getting it looking good, whereas backend/driver guys can worry about it working.