[lto] PATCH: new CALL_EXPR constructors

Paolo Bonzini paolo.bonzini@lu.unisi.ch
Mon Aug 7 13:57:00 GMT 2006

> I've looked at the patch, and I think it's OK.  In theory, I agree with
> you that it might have been better if Sandra submitted the static-change
> change separately, but, it's already done, and I don't see much value in
> going back to break it out now.  And, making global passes over the
> compiler to change things is pretty painful, so I can see why Sandra
> only wanted to make one pass and not two.
It's only apparently two, because the "remove static-chain" change could 
be sneaked in as obvious...
>> It's better not to introduce new TREE_CHAIN usages.  You may
>> put a FIXME comment saying that we may want to remove it. 
> I'm not sure that a FIXME here would be useful.  The only alternative to
> TREE_CHAIN, at the moment, would be a VEC, and that would use more
> memory.
Sandra is using it only to keep track of things allocated together, and 
an obstack can be good for that.  I could volunteer to do it though.
> Sandra, this patch is OK.  I'm looking forward to the next bit, where
> you actually get to change the representation!
Me too.  Please do not misunderstand my criticism, I am trying to be 
constructive but I can see that sometimes I failed.


More information about the Gcc-patches mailing list