compiler performance

[Felix Lee]

Robert Dewar <dewar@gnat.com>:
> I looked at the benchmark code. It is naive and useless.
> Worse than I expected actually. The writer has no concept
> of the notion of liveness, and is also apparently
> blissfully unaware of loop optimization.

the interesting thing to me is the author has an A.B. in compsci
from Harvard and a Ph.D. in philosophy from Berkeley.  his
dissertation is an examination of the question of machine
consciousness.  he's not completely naive about computation and
compilation, but he's missing all these concepts about
optimization and efficiency that I tend to assume that most
programmers know.  which obviously isn't true.

it seems to me that unless programmers have a reasonable idea of
"what a compiler can and can't do for me", then they're not going
to be able to write reasonable programs in general.  in a sense,
this is a PR issue.  perhaps the compiler should be better at
explaining what it's doing, for average programmers.
--