How to parallelize IIR filter with CUDA?

Hi!

Is it possible to speed up IIR filter with CUDA? As IIR filter is defined recursively it’s quite challenging.


https://docs.scipy.org/doc/scipy/reference/generated/scipy.signal.lfilter.html#scipy.signal.lfilter

This is outside my area of expertise. Have you performed a literature search? A simple search of the literature appears to indicate that parallelized GPU implementations of IIR filters are possible, for certain specific use cases and/or subject to specific constraints.

I used the following search terms with Google Scholar: GPU IIR filter

GTC 2018:
http://on-demand-gtc.gputechconf.com/gtc-quicklink/ef9vOTF

GTC 2014:
http://on-demand.gputechconf.com/gtc/2014/presentations/S4382-flexible-iir-filtering-for-audio-processing.pdf