- From: Marcus Geelnard <mage@opera.com>
- Date: Mon, 06 Aug 2012 20:48:13 +0000
- To: public-audio@w3.org
Citerar Chris Wilson <cwilso@google.com>: > On Mon, Aug 6, 2012 at 11:31 AM, Jussi Kalliokoski < > jussi.kalliokoski@gmail.com> wrote: > > 2) Can you elaborate on how the performance can be "at least" > equaled? I don't see how you expect to get significantly better > performance than > natively-implemented code doing DSP operations on a separate high-priority > thread. Actually, that's not how I read it - not sure what Jussi meant - but anyway, I can imagine some cases where you'd actually get better performance with a custom node, if: - You have a native implementation of the DSP API (i.e. the custom code should be about as fast as the corresponding native nodes when doing the same tasks). - You can bunch together a number of native nodes into a single custom node (i.e. less routing/mixing overhead). - You can optimize/eliminate some operations, since you have more freedom to deal with data/operations as you wish, without being confined to the routing interfaces of audio nodes. Now, this is purely speculation, since I have no concrete examples of such cases. I just know from experience that low-level usually offers more optimization opportunities than high-level (e.g. assembler vs. C++), so I'm just pointing at the possibilities here. In general, and especially for common operations, my bet would be that native nodes will be faster than custom nodes. /Marcus
Received on Monday, 6 August 2012 20:48:42 UTC