Re: sub-optimal code for packed boolean arrays -- bug or inherent limitation

[tmoran@acm.org]

> That's not a proper timing. 3.15p is using a very old backend (2.x if my
> memory is good). Since then there have been two major versions of the
> GCC backend, both bringing better optimizations. It would be better to
> do the measurement with GCC/FSF 4.2 or GNAT GPL 2007.
  Gnat 3.15p was handy.  Please post your results on timing the different
codes with a different Ada compiler.
  Since my machine does have two cores, I tried it with a second task
running the benchmark subroutine, ie two copies of the benchmark were
running simultaneously.  Elapsed time till they were both done was only
20% longer than running a single copy.  So it might be worth looking into
a multi-tasking design of the chess program.