Hello all
I believe I have found a bug in digital_fll_band_edge_cc.cc: with the
same
input under different cpu loads I see different outputs. It seems to
have
something to do with how the block uses output history from the
output_items parameter. Modifying it to use its own internal buffer to
store output for the next iteration instead of using the output_items
parameter seems to fix the problem.
Also modifying digital_fll_band_edge_cc to use SSE instructions for its
dot
product instead of a for loop gives a nice performance improvement.
I have attached diffs of my version of the files.
Regards
Hendrik van Wyk