Re: [tree-ssa] RFC: Making control flow more explicit

[Daniel Berlin <dberlin at dberlin dot org>]

> >     177.mesa          1400   173       808
> >     179.art           2600   428       607
> >     183.equake        1300   142       917
> >     188.ammp          2200   554       397
> >
> >
> > With his changes:
> > 177.mesa          1400   171       820
> > 179.art           2600   419       620
> > 183.equake        1300   138       944
> > 188.ammp          2200   540       407
>
> Yummie that's impressive.  Hope you can stamp out the bugs, Zdenek :)
>
>
> Gr.
> Steven
>
> PS.  Hmm now I think of it, how are those numbers possible?  Didn't
> these patches break the loop optimizer???

Look at my forward substitution and LICM patch (i've actually updated 
it to hoist loads too).
You'll note i made changes to the loop optimizer to not canonicalize 
loop headers.
With that change made, you can bootstrap with using the loop optimizer 
in tree mode.
It also never broke the RTL loop optimizer.