Re: Filter Templates

On Thu, Feb 24, 2011 at 6:15 PM, Chris Lilley <chris@w3.org> wrote:

> On Friday, February 25, 2011, 1:54:48 AM, Robert wrote:
>
> ROC> On Fri, Feb 25, 2011 at 12:06 PM, Rik Cabanier <cabanier@gmail.com>
> wrote:
> ROC>  Is it possible to implement hardware acceleration for SVG
> ROC> filters? I'm not an expert on the subject but it seems that some
> ROC> of the filters would be very hard to write as a shaders.
>
> There have been hardware-accelerated implementations of SVG filters. This
> was on an Texas Instruments chip, which was an Acorn ARM processor with
> integrated DSP, and was around 2004 or so.
>
> At the time, we were trying to work out a suitable subset of filters
> suitable for mid-range (at that time) mobile devices. The feedback from
> BitFlash, who were doing the hardware accelerated implementation, was that
> "all of them are suitable for mobile". I conclude that they had at that time
> accellerated all of the filters.
>

I asked our GPU people if they can take a look at this since I'm not an
expert on what's possible.


> >
> ROC> Anyway, it doesn't really make sense to go off and define an
> ROC> all-new standard because we think some SVG filters might not be
> ROC> acceleratable on some hardware. As Dean pointed out, that will be
> ROC> true for any adequately powerful filter API. We'll simply have to
> ROC> fall back to a software implementation or refuse to apply the filter.
>
> I agree.
>
> That seems reasonable. A browser might use something like PixelBender under
the hood to create the filter graph or render the filter in software.

Rik