EXA and MigrationHeuristic

Maarten Maathuis madman2003 at gmail.com
Fri May 30 03:53:42 PDT 2008


On Fri, May 30, 2008 at 9:45 AM, Michel Dänzer
<michel at tungstengraphics.com> wrote:
> On Thu, 2008-05-29 at 14:01 +0200, Juliusz Chroboczek wrote:
>> > "greedy" is faster than "always" in some cases because it's quite
>> > conservative about migrating pixmaps offscreen. Any single low level
>> > operation can only be accelerated if all pixmaps involved are offscreen,
>> > so this results in fewer operations being accelerated, thereby avoiding
>> > migration overhead for operations that can't be accelerated for other
>> > reasons.
>>
>> Let me change the emphasis.  Main-memory pixmaps are moderately fast
>> for all operations.  Off-screen pixmaps are very fast for some
>> operations, and glacially slow for others.
>
> No, that's not how it is or what I said. Anyway, I'll leave the flaming
> fun to you guys in my western neighbour country. ;)
>
> That said, while I think changing the default would clearly be a bad
> idea for the Git master branch, for stable branches and distros it's
> ultimately up to the respective maintainers of course. However, I think
> the decision should be based on more than some specific use cases that
> don't represent what EXA is designed to do well.
>
> Also, from the Ubuntu log snippet somebody pasted, it looks like their
> change might even prevent using the other heuristics at all, which would
> be very bad. Even if that's not the case, it might have been nice to
> discuss it with upstream beforehand.
>
>
> --
> Earthling Michel Dänzer           |          http://tungstengraphics.com
> Libre software enthusiast         |          Debian, X and DRI developer
>
> _______________________________________________
> xorg mailing list
> xorg at lists.freedesktop.org
> http://lists.freedesktop.org/mailman/listinfo/xorg


Maybe it's an idea to change the migration default for drivers that
only provide blit and and solid fill, in any other case always is the
right default (imo).

Maarten.



More information about the xorg mailing list