[RFC/RFA?] New estimation of function body size for inlining

[Jan Hubicka]

> On Mon, 30 Jun 2003, Jan Hubicka wrote:
> 
> > this patch implements new function to estimate function size.
> > Instead of counting number of statements that can be fooled by very
> > complex statements in one direction and very many lexical blocks and
> > similar noops in the other direction I now do simple recursive walk
> > without duplicates and attempt to count everything that looks like
> > operation with few special cases dealing with common NOOPs and common
> > expensive expressions.
> 
> Hopefully this fixes PR 10971 where -std=c99 affected the old statement
> counting because of additional scopes in C99....

Yes, even when not perfect it seems to fix this PRs, as well as the one
about inlining empty functions and some others I run across.
Sad that I don't have time for too much investigation right now, but I
will try to send some incremental improvements next week.

Honza
> 
> -- 
> Joseph S. Myers
> jsm28@cam.ac.uk