mainline "exploding"

[Jan Hubicka]

> On Sun, 19 Oct 2003, Jan Hubicka wrote:
> > Hmm, on the other hand the function size estimates is much larger on tree-ssa
> > branch because of new nodes introduced by gimplification process.  I
> > will write proper generic code size estimation soonish so we can see
> > better comparisons then.
> 
> Curious... is this merely for comparison with mainline?  My naive

Yes, this is relative to mainline.

> expectation is that we could get better code size estimates from GIMPLE
> than GENERIC.

I believe so, but we need to run trivial optimization passes first.
genericizing produces some redundant mess that gets accounted (the
redundant mess is of course implicitly in GENERIC and language-specific
trees representation too so this is not real a regression) and makes the
results less accurate.

I hope to have size esimate patch today.  Once -funit-at-a-time gets
re-enabled we can start poking around with reordering the inliner and
optimization passes.

Honza
> 
> Jeff
>