Re: Freerunner and Wayland
On Thu, 7 Jul 2011 18:05:02 +0100 Neil Jerram wrote: > Just wondering... is it at all feasible, in the nearish future, for > Wayland to run on the Freerunner? I mean directly on KMS, not as an X > client. > > Would there be any advantage to that, compared to the current X usage? > I'm imagining it might perform better, but I don't really know. It's feasible, but not easy. Wayland is essentially a thin wrapper around the low-level DRM and KMS stuff allowing clients to submit hardware command sequences directly rather than going via X's acceleration pathways. A lot of the performance difficulties with the X pathway (not just on our hardware) seem to be because the server can't possibly know enough about what the client wants to accelerate it effectively. Fast and smooth graphics on the Freerunner should be perfectly possible, but would rely on exactly this kind of "clairvoyance" from the X server. So, getting Wayland to run on its own shouldn't be too difficult (famous last words..), but writing programs which can actually make use of it is significantly more difficult. Tom -- Thomas White ___ Openmoko community mailing list community@lists.openmoko.org http://lists.openmoko.org/mailman/listinfo/community
Re: Freerunner and Wayland
Hi Tom, Thanks for your answer... On 13 July 2011 10:01, Thomas White wrote: > acceleration pathways. A lot of the performance difficulties with the > X pathway (not just on our hardware) seem to be because the server > can't possibly know enough about what the client wants to accelerate it > effectively. Fast and smooth graphics on the Freerunner should be > perfectly possible, but would rely on exactly this kind of > "clairvoyance" from the X server. > > So, getting Wayland to run on its own shouldn't be too difficult > (famous last words..), but writing programs which can actually make use > of it is significantly more difficult. I have read that some toolkits, like Gtk+ and Cairo, have (or are in the process of having) support for Wayland as their backend directly (i.e. not via X). Also that it's possible to write clients using a GL API directly, and that the library providing that API would use Wayland directly. I guessed from that that the toolkit or GL implementation might be in a better position to have exactly that kind of clairvoyance - i.e. to know what kind of acceleration would be useful, and to ask the hardware driver for that. Hence, I thought, there might be some performance benefit in the acceleration-potential being in the toolkit or library, instead of in X; and also perhaps in just cutting out one of the layers. Also, I presume, I could write a new client today using e.g. the Cairo API, and that should Just Work. Is any of that correct? (Having said that, I don't recall reading yet of any Wayland support in the E toolkit, and certainly that would be a specific problem for SHR usage. But maybe Wayland is still worth experimenting with in a non-SHR setup.) Thanks again, Neil ___ Openmoko community mailing list community@lists.openmoko.org http://lists.openmoko.org/mailman/listinfo/community
Re: Freerunner and Wayland
On Wednesday 13 July 2011, Neil Jerram wrote: > Hi Tom, > > Thanks for your answer... > > On 13 July 2011 10:01, Thomas White wrote: > > acceleration pathways. A lot of the performance difficulties with the > > X pathway (not just on our hardware) seem to be because the server > > can't possibly know enough about what the client wants to accelerate it > > effectively. Fast and smooth graphics on the Freerunner should be > > perfectly possible, but would rely on exactly this kind of > > "clairvoyance" from the X server. > > > > So, getting Wayland to run on its own shouldn't be too difficult > > (famous last words..), but writing programs which can actually make use > > of it is significantly more difficult. > > I have read that some toolkits, like Gtk+ and Cairo, have (or are in > the process of having) support for Wayland as their backend directly > (i.e. not via X). Also that it's possible to write clients using a GL > API directly, and that the library providing that API would use > Wayland directly. Add Qt to that list. Meego may be using it, and kwin is being ported as the first stage in letting kde run on Wayland. Both of these may be a little heavy for the Glamo's GL capabilitites, but it shows writing applications for Wayland is getting much easier. ___ Openmoko community mailing list community@lists.openmoko.org http://lists.openmoko.org/mailman/listinfo/community