This is the mail archive of the mailing list for the GCC project.

Index Nav: [Date Index] [Subject Index] [Author Index] [Thread Index]
Message Nav: [Date Prev] [Date Next] [Thread Prev] [Thread Next]
Other format: [Raw text]

Re: [PATCH]: Code sinking patch

On Wed, 2004-12-22 at 09:10 -0800, Mark Mitchell wrote:
> Giovanni Bajo wrote:
> > Strongly agreed. 3% is really a good improvement which we should not
> True.  That's why I didn't just say "no, wait until 4.1, go away". :-)
Not only is it a 3% improvement, it's an improvement gained by
generating better code as opposed to source level changes.  That's
my favorite kind of improvement as it is likely there's other programs
which will see improvements.

> I was asked by someone whose judgement I trust to consider the patch, so 
> I did.  After reviewing it, I was pleased with what I saw, but I think 
> that the default position has to be to reject this patch, as it's 
> clearly outside the guidelines we've set down.  What I'd be looking for 
> in order to make an exception would be (a) optimizer experts who say 
> that Dan's code is really good, after going over it carefully, and (b) 
> nearly unanimous sentiment from everyone else that it's a good idea to 
> put this in at this point, even though clearly this would be unfair, in 
> that other people didn't get to put their stuff in, etc.  I'm not going 
> to be *that* inconsistent without full approval of all concerned.
> I don't think we've got either side of that equation at this point, and 
> there have in fact been some objections to (b).
I'd actually tend to lean towards including it based strictly on its
results, particularly if others have looked at the code and think it's
reasonably clean.  I haven't looked at the code.


Index Nav: [Date Index] [Subject Index] [Author Index] [Thread Index]
Message Nav: [Date Prev] [Date Next] [Thread Prev] [Thread Next]