Re: some interesting timing statistics

[Zack Weinberg <zack at bitmover dot com>]

Bernd Schmidt wrote:
> > I have a source file which demonstrates pretty nicely how much slower
> > the compiler's gotten since 2.7.x.  It is 12583 lines long, with 286
> > functions.  On x86:
> 
> One thing to take into account in comparisons like this: were both compilers
> compiled with the same optimization flags?

Probably not.  I doubt it would make that much difference though.

> > I'm surprised -O2 slows down the parser, of all things, compared to
> > -Os.  A hefty chunk of the performance loss comes from there.
> 
> That could be because all the time not accounted for otherwise is counted
> against the parser.  Most likely it's a simple accounting error.

I didn't know that.  Might be worth looking into...

zw