[Date Prev][Date Next] [Thread Prev][Thread Next] [Date Index] [Thread Index]

Re: gcc 3.2 not faster



On Wed, Jan 08, 2003 at 06:10:42PM +0100, Russell Coker wrote:
> On Wed, 8 Jan 2003 17:00, Michael Stone wrote:
> > On Wed, Jan 08, 2003 at 03:38:04PM +0100, Russell Coker wrote:
> > >I thought that gcc 3.2 was supposed to be faster, however I have just done
> > >some benchmarks to show the opposite:
> > >
> > >GCC 2.95:
> > >Version 1.93b          write   read putcNT getcNT   putc   getc  putcU 
> > > getcU lyta                     437    559   9052   9478   1694   1734 
> > > 24757  48029
> > >
> > >GCC 3.2:
> > >Version 1.93b          write   read putcNT getcNT   putc   getc  putcU 
> > > getcU 441    568   7955   8573   1617   1698  18731  28544
> >
> > It's helpful when posting benchmarks to give some indication of what the
> > numbers represent. Is it operations per time? Or elapsed seconds? Or
> > bogomips?
> 
> Thousands of operations per second.
> 
> The "default" for reporting any benchmark results is that bigger numbers are 
> better.

That's hardly true.  Most of the benchmarks I work with report times. 
This is why it's important to specify.

> The relevant fact here is not the absolute numbers, but the fact that GCC 3.2 
> produces code that is slower.

But the important question is why.

> I wonder what will happen when libc6 is compiled with GCC 3.2...  For 
> getc/putc operations what happens in the libc6 is more complex than what 
> happens in the application.  If the same performance hit occurs when 
> compiling libc6 then things will really suck, and I'll probably get CC'd on 
> some more amusing flame-wars.

You're missing two important details:
  - glibc has been building with GCC 3.2 since well before the
    transition; two or three months now at least.
  - You can't build glibc 2.3.1 with any earlier version of GCC anyway.

-- 
Daniel Jacobowitz
MontaVista Software                         Debian GNU/Linux Developer



Reply to: