[webkit-dev] parallel layout research

[Maciej Stachowiak]

I think parallel layout is an interesting topic to explore, in light  
of where CPUs are going. We probably can't use code in Cilk++  
directly, but there might be useful ideas. It also seems to me from  
their paper that they are only handling some of the more basic aspects  
of CSS layout so far. Definitely interesting research for us to follow.

One thing to keep in mind - parallelizing layout or other CSS  
operations may not be the most effective solution to any given  
problem. If there are Google properties that are gated on layout  
performance, then by all means file bugs and this is the sort of thing  
we can likely investigate and greatly optimize without having to  
completely rearchitect layout. We have had a lot of success with  
improving layout performance given test cases that demonstrate slowness.

Regards,
Maciej

On Jul 31, 2009, at 11:59 AM, Alex Russell wrote:

> After a discussion this morning about the potential for parallel CSS
> layout with Erik Aarvidson, he pointed out a group at Berkeley doing
> research in this area:
>
>    http://www.eecs.berkeley.edu/~lmeyerov/projects/pbrowser/
>
> The bits that jumped out to me were:
>
>   * the source is available [0]
>   * they've implemented an abstract language for describing the CSS
> spec for parameterizing layout. It reminds me a lot of the work Badros
> did circa 2000 [1]. They suggest that this model does layout of the
> CSS 2.1 spec in time similar to the hand-tuned sequential engine in
> Gecko [2]
>   * using Cilk++ to handle low-level paralleism, they've seen
> significant speedups beyond that for multi-core [2]
>
> It's exciting to me to see this kind of parallelism being extracted
> from CSS layout operations, which have shown up in profiling of Google
> apps as a serious bottleneck. I hesitate to ask about the WebKit
> community's interest in this since I assume that if patches show up
> and can prove their value, they'll be accepted, but I would like to
> know where people are either supportive or skeptical about such an
> approach.
>
> Regards
>
> [0]: http://code.google.com/p/pbrowser/source/browse/#svn/trunk
> [1]: http://www.badros.com/greg/papers/gjbadros-dissertation.pdf
> [2]: http://www.eecs.berkeley.edu/~lmeyerov/projects/pbrowser/retreatwinter2009/playout.pdf
> , slide #19
> _______________________________________________
> webkit-dev mailing list
> webkit-dev at lists.webkit.org
> http://lists.webkit.org/mailman/listinfo.cgi/webkit-dev