The search for a good rate 1/16 K=19 code continues without
success.

I improved the code search program.  It examines 2 million
polynomials per CPU second.  After some hundreds of CPU
hours it finds about 80,000 potential codes and these go for
further testing.  Some of these look quite good, 96 of them
have minimum distance equal to the Heller bound.

But strangely none perform better than 8K19A when using long
lists (eg 500k list size).   Several do better than 8K19A but
only up to list length of 10 or so.

Perhaps there is some subtle mistake in the program?  So I
tested the search program on rate 1/8 K=19 and it soon produced
several new good codes, one of which is consistently better
than 8K19A at all list lengths.

So something strange is going on.  I didn't have this difficulty
with the stronger codes.   The more I learn about this subject,
the less I seem to know.  Something I don't understand is
happening here.

--
Paul Nicholson
--