Re: voting on GOTO
Jeremy Cowgar wrote:
> c.k.lester wrote:
> > Jim Brown wrote:
> > > 9,999,999 function calls is about 2 seconds. 99,999,999 gotos is <1
> > > second.
> > While that kicks total butt, I have to wonder who calls a function almost
> > 10 million times? Just curious. I'm sure Kat with her processing of
> > gigabytes
> > of text probably makes use of some lines of code many many times... Kat?
> Have you done benchmarking before? The greater number of iterations you test,
> the more accurate your result is. The important thing is goto is substantially
> faster.
Yes, I've done benchmarking before, and I understand that more iterations gives
you a finer measurement. But there are theoretical considerations and there
are real-world considerations.
You said yourself you'd choose a Euphoria program of 25 minutes vs. an
assembly language of 10 seconds... so 2 seconds over the course of almost
10 million iterations doesn't seem so bad.
What I'm saying is, if it's going to take some astronomical number of
iterations to show a slight improvement, then it's probably not worth
adding to the core, because nobody's going to call a function that many
times.
I agree that for speed freaks, an improvement at 10 million iterations of
any amount is worth keeping, and generally I classify myself as a speedfreak.
But this one isn't compelling to me, especially when combined with its
unstructured nature. :/
|
Not Categorized, Please Help
|
|
