🚀 go-pugleaf

I have a broad view. I used to work for Lawrence Livermore National
Laboratories, which is a nuclear weapons design lab. All we cared
about is the performance of programs that ran for weeks on Crays.

I wrote the book ``Performance and Evaluation of Lisp Systems'' in
which I pointed out that it is improper to assign to any one benchmark
the overall `performance' of a Lisp system. If you assume a reasonably
competent programmer (I use this as a shorthand for designer/coder),
it is most likely that things like dispatch don't happen as the only
operation in the only inner loop that matters. More likely to affect
performance is the design of the algorithms and the data structures.

In my shop we did a several person-decade project in C++ during which
I witnessed a factor of 3 productivity decline over the Lisp group. By
my reckoning, using C++ instead of Lisp cost me $4,000,000 (US), and
the end result was no faster (as judged by comparing it with a similar
Lisp-written product) than it would have been in Lisp.


			-rpg-