Re: [wwwdocs] gcc-4.1/changes.html: Mention optimizer improvements.

[Mark Mitchell <mark at codesourcery dot com>]

Paolo Carlini wrote:
> Giovanni Bajo wrote:
>
>
> > At the very least there should be a little explanation following it; a
> > better way would be to show the transformation performed by this pass in
> > pseudo C. Even better if the code was compilable C and the effect was
> > demonstrable by looking at the generated assembly source (but I understand
> > that this might not be possible because some of these optimizations could be
> > already performed by RTL optimizers).
>
> Seconded. I should also add that, judging by some PRs he filed, Kazu
> seems the right person for this kind of work.

I, too, agree.

Users are better able to make sense out of a statement like "GCC's 
performance on the often-used SPEC 2000 benchmark has improved by as 
much as X% on many platforms, including IA32" than they are "GCC now has 
a value-range propogation pass".  If these passes don't improve the code 
that comes out of the compiler, or make the compiler faster, most users 
aren't going to think that's a very interesting improvement -- even 
though they might be very valuable in that they allow us to make 
internal improvements that we want to make.

If we do want to use the technical compiler-speak, then I think that 
showing code examples ia a great idea.  That's something programmers 
will understand.

--
Mark Mitchell
CodeSourcery, LLC
mark@codesourcery.com
(916) 791-8304