Hello, There is also scope for reducing CPU load through algorithm modifications. For example the post filter uses a bunch of FFTs but could be implemented more efficiently in the time domain. It would be effectively the same signal processing, just implemented differently. Cheers,
David On Wed, 2013-10-09 at 10:22 -0700, Bruce Perens wrote: > Hi Frank, > > LibAV and FFTW both have support for NEON extensions. It would be > interesting to benchmark their FFT against what we're using now. That > could be a quick validation regarding whether NEON is helpful. Both > the softmodem and the codec can use accelleration. > > One concern of ours is the low end. We are looking at STM32F4 and > Sitara for embedding in radios. I think we're limited to conventional > floating point on those. > > Thanks > > Bruce > ------------------------------------------------------------------------------ > October Webinars: Code for Performance > Free Intel webinars can help you accelerate application performance. > Explore tips for MPI, OpenMP, advanced profiling, and more. Get the most from > the latest Intel processors and coprocessors. See abstracts and register > > http://pubads.g.doubleclick.net/gampad/clk?id=60134071&iu=/4140/ostg.clktrk > _______________________________________________ Freetel-codec2 mailing list > [email protected] > https://lists.sourceforge.net/lists/listinfo/freetel-codec2 ------------------------------------------------------------------------------ October Webinars: Code for Performance Free Intel webinars can help you accelerate application performance. Explore tips for MPI, OpenMP, advanced profiling, and more. Get the most from the latest Intel processors and coprocessors. See abstracts and register > http://pubads.g.doubleclick.net/gampad/clk?id=60134071&iu=/4140/ostg.clktrk _______________________________________________ Freetel-codec2 mailing list [email protected] https://lists.sourceforge.net/lists/listinfo/freetel-codec2
