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: [4.2 only] RFA: PR 33848: reload rematerialisation of labels

> Cannot modulo bugs.  I've already explained why I think my patch solves
> Andrew's bug too.  It sounds like you think I'm reverting to the situation
> before Andrew's patch, but I'm not really. 

No, I think that (because of bugs) you could be reverting to the situation 
before Alexandre's patch.

> So Andrew's patch and mine are dealing with the same case: where the
> LABEL_REF we're replacing is not the JUMP_LABEL.  Andrew's PR was an
> example of (a) and mine is an example of (b).  Andrew's patch fixed
> (a) by making the LABEL_REF the JUMP_LABEL too, thus transforming a
> computed jump into an indirect jump with a known target.  My patch
> fixes (a) by stopping us from adding the note in that case, so that
> the jump in Andrew's PR remains a computed jump. 

I agree with this.

> As I said before, transforming a computed jump into an indirect jump to a
> known target, as Andrew did, is a nice CFG optimisation, but not one we
> should do here. 

OK, but I fear that you could be throwing the baby (a needed note) with the 
bathwater (the new JUMP_LABEL).

Sorry, I don't seem to be able to make up my mind about this affair or clearly 
formulate what I'm afraid of.  The thing is, it's reload and I'll never feel 
comfortable with patching it on a release branch.  So I'd suggest to run this 
by some other maintainer (RTH?), setting aside the current thread.

Eric Botcazou

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