[dataflow] Don't rescan everything after CSE

Kenneth Zadeck zadeck@naturalbridge.com
Tue Mar 6 16:46:00 GMT 2007


Paolo Bonzini wrote:
> This saves ~1% of compilation time.
>
> More ideas that I have, but did not implement are:
>
> 1) delay computation of LR/UR until where we did it in the trunk.  fwprop and GCSE do not need it, for example.
>
>   
It is possible that there is some mileage to be gained by add a
changeable flag to df_analyze that says that this pass does not need
lr/ur and so you can just let it rot even as you do the other problems
like building the chains.  However, remember that you need these to
build the dead/unused notes, so if that gets specified, you should not
turn off the lr/ur processing.



> 2) don't update REG_DEAD/REG_UNUSED notes for patches that don't need it.  Updating them could/should be a changeable flag.
>
> Bootstrapped/regtested i686-pc-linux-gnu.
>
> Paolo
>   



More information about the Gcc-patches mailing list