Re: PPC optimisations

[Ralph Schmidt <laire at basis dot owl dot de>]

You wrote:
> It still is not clear to me whether you are referring to the
> gcc-2.95 releases on which Franz Sirl's PowerPC Linux RPMs are based or to
> the gcc development sources.  When you refer to snapshots, do you mean the
> GCC snapshots or Franz Sirl's snapshots?  It was my understanding that
> Franz Sirl's snapshots were tracking the gcc-2.95 release branch.
>

I meant Frank's snapshot.

> To where have you sent "bug reports" about missed optimization
> opportunities?  To the Franz Sirl and the PowerPC Linux mailinglists or to
> the GCC mailinglists or both?
>

I sent the bugreport to the bug mailinglist. Franz reacted and reported that
the bug was already fixed in his snapshot.
The bug report also contained 2 points about missing optimisation
opportunities in the bug example.
Maybe some people missed it...

> When you discuss missed PowerPC optimizations when compared to
> x86, are you referring to architecture-specific optimizations or generic
> compiler transformations which have more impact on RISC architectures than
> CISC architectures?
>

Mostly peephole optimisations...so..architecture-specific.

> Cygnus currently is implementing a number of optimizations in the
> PowerPC target of GCC under contract from IBM, such as scheduled prologue
> and epilogue, better use of condition registers, and others which will
> continue to be implemented over the coming year.
>

That sounds indeed great.

> As Clint said, it is very difficult to respond to your statements
> or try to improve GCC when everything is described in the abstract.
>

I know...i want to go into more detail when i have franz snapshot really
working on my host so i don't report already fixed optimisation misses.


Regards

---
Ralph Schmidt,laire@popmail.owl.de(private),NextMail welcome