usenet.2001-05-09 at bagley.org
Wed May 9 04:21:00 EDT 2001
Gerrit Muller <gerrit.muller at philips.com> writes:
> One of the items you are obviously struggling with is the balance
> between the different criteria: do you program to optimize
> time-to-result, maintenance, performance or memory usage? This question
> pops up for every individual test.
For the most part, solutions on my site should be written first to be
fast, second to be memory economical, and probably none were really
written to be conservative in "lines of code", since I just added that
metric very recently. I would probably reject a solution if I felt it
was a flagrant memory waster, since you can't really get away with
that in real life. I'll see what I can do about improving the test
descriptions to reflect these criteria more clearly.
> A solution might be to do something comparable with SPEC: have a
> BASE test, which is programmed straightforward and also run an
> OPTIMIZED test, which is optimized for production environments
> (fast and low memory usage, at the cost of more development time
> and perhaps larger loc).
I'll try to add more examples of "straightforward" code, especially
for the cases where an optimized solution uses noticeably more locs,
or isn't clear.
> Another suggestion is to look for one "larger" benchmark, doing some
> real-life work. From loc point of view that might be a more "realistic"
I think that would be ideal. I would certainly need some help on
that one though. I don't think I could do justice to even a small
application written in 20+ languages.
I think that like all benchmarks, mine should be taken with a grain of
salt too, but if nothing else, it continues to be a good learning
Thanks for your input.
More information about the Python-list