bearophile wrote:
Andrei Alexandrescu:
I'm not sure on what machine you test, but on my (crappy old)
laptop, your Appender example completes in 4.95 seconds. Are you
sure you are measuring the right thing?
My purpose is to have a good standard library and a good language.
Errors of mine are always possible. Benchmarks are tricky things.
I have redone the tests, the CPU is unloaded, no computation is going
on, the system is WinXP with 2 GB RAM, 2 GHz Core 2.
New timings give no less than 22.6 seconds, repeated. As running time
I use a timing command similar to the "time" of Linux, that gives the
total running time of the program (final deallocations too).
I think Appender can be improved some.
I can't improve what I can't measure. On my system, liberally loaded
with a variety of programs, weaker than yours (1.8GHz/512MB laptop) I
get 4.8s with -O and 5.3s without. The major difference is the OS
(Ubuntu on my laptop), but I am puzzled as Appender doesn't assume a lot
about the OS. I time with Unix time. I copied the code straight from
your post. Please advise.
Andrei
