Native compilation article

[Tom Tromey]

Anthony> I suppose "gcj -O2" is a reasonable guess at picking right
Anthony> compiler options, although "-O3" is supposed to be the
Anthony> general `maximum optimization' flag.

Sometimes -O2 is faster, sometimes -O3 is faster.  It depends on the
application.  Also explicitly asking gcc to unroll loops sometimes
helps.  I've been meaning to try the IBM `primes' code with loop
unrolling to see how that does.

Tom