[dataflow] Don't rescan everything after CSE

Paolo Bonzini paolo.bonzini@lu.unisi.ch
Tue Mar 6 17:16:00 GMT 2007


> If this really works, it is ok to commit.  I would like see it
> bootstrapped on more platforms.  My skepticism is because Steven and I
> both tried to do this and we were never able to track down all of the
> places where insns were being modified.
> 
> Do you have access to a ppc or an ia-64?  You can get an ia-64 account
> from willy.

I have run some PPC compilations and it seems at the very least to not crash (I haven't run assembly comparisons yet as I don't have a baseline compiler quickly available).

I'm pretty sure that, if there are any problems, I can solve them quick; so if you agree, I can just put it on the branch so that you can try it on your platforms, and I will revert it tomorrow morning if you report any problem that I cannot quickly fix.

I'm pretty sure that, since you tried, some simplifications in CSE went on (some after fwprop was committed, and Andreas Krebbel introduced validate_canon_reg).

(I don't know who's willy).

Paolo



More information about the Gcc-patches mailing list