[PATCH]: Interprocedural detection of readonly and non-addressable static variables

Diego Novillo dnovillo@redhat.com
Thu Jun 30 13:53:00 GMT 2005


On Wed, Jun 29, 2005 at 03:49:51PM -0400, Kenneth Zadeck wrote:

> I modified the comment to remove the word "static".  I need to rely on 
> the intersection because the set of static variables that it is 
> intersected against is a subset of all of the static variables.  The 
> intersected set are the variables that satisfy a bunch of constraints 
> such as not escaping, not have their address taken, having a type that 
> we are willing to promote and not having any silly attributes that few 
> people know about or understand.  Thus at best, the single intersection 
> (per call to this function) could be replace by a bitmap lookup for each 
> variable found.  This does not seem profitable.
>
I'm not following.  I was thinking

	if (is_global_var (var))
	  add to bitmap

That adds all the statics.  The intersection later on removes the
non-interesting statics.  What am I missing?


Thanks.  Diego.



More information about the Gcc-patches mailing list