Hi list,
I would just like to chime in with some observations I have made in the
embedded industry lately,
where it seems like any die shrink to 28nm for ARM usually goes hand in hand
with a move to the Cortex series of cores. The Cortex cores usually have SIMD
(NEON) acceleration as well as (at a minimum) single precision floating point,
which obliviates the need for rewriting codec2 to fixed point for these
architechures.
An interesting datapoint for this trend is the Arduino Tre
http://www.theverge.com/2013/10/5/4805844/arduino-tre-from-texas-instruments-promises-100x-power
which seems to blur the lines between Arduino micro-controllers performance,
and that of the Raspberry Pi like computers.
Personally I have experience in adapting algorithms to fixed point, but in
light of the trends I just mentioned, my feelings is that if I wanted to
contribute to the codec2 projects, ensuring that the code runs efficiently on
NEON would be time better spent compared to taking on the fixed point problem.
Cheers,
Frank / LA9OSA
------------------------------------------------------------------------------
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
