On 2013-04-17 9:58 AM, Paul Adenot wrote:
I can't think of anything on the architectural side that would prevent us from optimizing Web Audio using a DSP or SIMD, because it boils down to implementing pure signal processing functions using SIMD intrinsics or assembly.
I have not looked into this at all, but using hardware support for effects such as convolution would only be an option if the hardware processing uses an algorithm with compatible results to the software implementation (but this is of course not an issue for SIMD-ifying our processing code.)
Simply implementing the specialized functions in a separate file and calling the right function based on the chip capabilities (like we do for SIMD) should be enough. iirc, we tried to group all the signal processing functions in the same file, or could do so rather easily.
Yes, see <http://mxr.mozilla.org/mozilla-central/source/content/media/AudioNodeEngine.cpp>.
Cheers, Ehsan _______________________________________________ dev-media mailing list [email protected] https://lists.mozilla.org/listinfo/dev-media
