Caring less

Posted Jul 1, 2013 6:51 UTC (Mon) by dlang (guest, #313)
In reply to: Caring less by mjg59
Parent article: Riddell: Kubuntu Won't be Switching to Mir or XMir

since XMir is the X protocol on top of Mir, why would drivers that work with Mir not work with XMir?

Caring less

Posted Jul 1, 2013 15:05 UTC (Mon) by mjg59 (subscriber, #23239) [Link] (17 responses)

An X client connects to XMir. It sends an XDrawRectangle command. How does XMir turn that into a hardware-specific draw command?

Caring less

Posted Jul 1, 2013 21:04 UTC (Mon) by raven667 (subscriber, #5198) [Link] (15 responses)

I don't know but I would guess EGL or its not hardware accelerated. I would guess that it's not much different than XWayland, XAqua, XWin, etc. any system where X.org isn't directly driving the video hardware is going to have these same constraints.

Caring less

Posted Jul 1, 2013 21:32 UTC (Mon) by mjg59 (subscriber, #23239) [Link] (14 responses)

Which provides no advantage over using X.org. XWayland loads modified X.org drivers to provide hardware acceleration, and my understanding is that XMir behaves in the same way. I've seen no evidence that XMir is able to use Android drivers in any meaningful way.

(The fairly obvious subtext here is that dlang should stop making authoritative statements about things he doesn't understand)

Caring less

Posted Jul 1, 2013 21:40 UTC (Mon) by dlang (guest, #313) [Link] (10 responses)

I posted links to the official Ubuntu site and to one of the developers FAQ sites.

Both of them explicitly say that Mir is able to use unmodified Android drivers to provide fully accelerated graphics.

If you are claiming that Mir windows on a Mir system will be accelerated, but XMir cannot use those drivers at all, that would seem to the statement that requires proof. It would seem obvious that if some windows on the screen can use Android drivers, that all the other windows on the screen would be using the same drivers.

Caring less

Posted Jul 1, 2013 22:03 UTC (Mon) by mjg59 (subscriber, #23239) [Link] (9 responses)

Look at the XMir source code. See any references to Mir protocol calls? It creates a Mir window and then renders into it. Mir has *zero* idea what's going on. In that situation, how could Mir accelerate any of the operations?

Bluntly, it would seem obvious that you don't know what you're talking about. Other windows are accelerated because the toolkits rendering into them are able to make mir calls. X clients can't.

Caring less

Posted Jul 2, 2013 11:23 UTC (Tue) by nye (subscriber, #51576) [Link] (8 responses)

You said:

>XMir doesn't run on top of unmodified Android drivers.

Then you said:

>It creates a Mir window and then renders into it

>how could Mir accelerate any of the operations?
>Bluntly, it would seem obvious that you don't know what you're talking about

You've just moved the goalposts and then attacked someone for not understanding where you've hidden them.

Caring less

Posted Jul 2, 2013 13:47 UTC (Tue) by mjg59 (subscriber, #23239) [Link] (7 responses)

Yes, I was inaccurate. What I should have said is that XMir (as opposed to Mir) will not make use of unmodified Android drivers. As such there's no advantage in running XMir rather than Mir - dlang's original claim of "There is unfortunately a lot of hardware out there that has good Android drivers, but no good Xorg drivers. If you try to run on hat hardware, you will probably be better off with XMir than with Xorg" was wrong.

But you're right, the way I disagreed with it initially was (strictly) wrong and potentially misleading. I apologise for that.

Caring less

Posted Jul 2, 2013 15:51 UTC (Tue) by dlang (guest, #313) [Link] (6 responses)

if Mir runs on a device, why would XMir not run on that same device (since it is just a layer on top of Mir)

since there are devices that have Android drivers but no X.org drivers, this means that XMir will run on some devices that X.org will not.

Other than the "it's not a free enough driver, so you should not use it" argument, how is someone not better off with XMir than Xorg in such a case?

Caring less

Posted Jul 2, 2013 15:55 UTC (Tue) by mjg59 (subscriber, #23239) [Link] (5 responses)

XMir will run on that device. Unaccelerated. X.org will also run on that device. Unaccelerated. How does that result in the user being better off?

Caring less

Posted Jul 2, 2013 16:01 UTC (Tue) by dlang (guest, #313) [Link] (4 responses)

In my experience, Xorg does not run on all the embedded devices/tablets/etc that have Android drivers. In some cases it's been a matter of using the Android drivers or no display.

In addition, as others noted, even if Xorg is able to get some display on the device via the framebuffer driver, it can't always handle the modesetting and multiple outputs that the Android driver can.

Having a display beats not having a display.

Caring less

Posted Jul 2, 2013 16:03 UTC (Tue) by mjg59 (subscriber, #23239) [Link] (3 responses)

Do you see any evidence of modesetting interface code in XMir?

Caring less

Posted Jul 2, 2013 19:26 UTC (Tue) by dlang (guest, #313) [Link] (2 responses)

it doesn't need to be in XMir, it just needs to be in Mir.

since XMir 'just' puts windows on the Mir desktop, any modesetting that Mir does benefits the entire desktop, including any XMir windows.

Caring less

Posted Jul 2, 2013 19:44 UTC (Tue) by mjg59 (subscriber, #23239) [Link] (1 responses)

I'm running an X environment on XMir. I want to move the screen on my HDMI output from the left of my internal display to the right of my internal display. How does that happen?

Caring less

Posted Jul 3, 2013 8:25 UTC (Wed) by renox (guest, #23785) [Link]

You change a configuration file for Mir?

Caring less

Posted Jul 1, 2013 21:59 UTC (Mon) by dlang (guest, #313) [Link] (2 responses)

> Which provides no advantage over using X.org.

Well, if there are Android drivers for a device and there are not X.org drivers for a device, getting a display would seem like at least a minor advantage to me.

If the Mir stack can then take full advantage of acceleration (as indicated by the various links) then there's no reason to expect that XMir would not be able to make use of it.

Caring less

Posted Jul 1, 2013 22:06 UTC (Mon) by mjg59 (subscriber, #23239) [Link] (1 responses)

X.org will run on any hardware that provides a framebuffer interface. You'll have zero acceleration, but in the absence of an X.org driver, XMir will also be unaccelerated.

Caring less

Posted Jul 2, 2013 6:10 UTC (Tue) by raof (subscriber, #57409) [Link]

You do still get the advantage of whatever modesetting support the Android driver provides - the framebuffer interface isn't known for its sterling multi-head support.

Caring less

Posted Jul 1, 2013 21:20 UTC (Mon) by dlang (guest, #313) [Link]

being that I'm not an X developer, I would assume that it works very similar to how X works for anything else. It has both the unaccelerated capability and whatever acceleration the drivers provide, XMir chooses which version to use based on what's available, just like Xorg would with different cards, some of which provide acceleration while others don't.

In either case, you can have the situation where sometimes the back-end provides acceleration, in other cases it doesn't. the X server (be it Xorg or XMir) will receive the same requests from the application and will decide how to render it.