Re: [tree-ssa] Dominator opts fixes and enhancements

[Andrew MacLeod <amacleod at redhat dot com>]

On Mon, 2003-08-25 at 08:54, Diego Novillo wrote:
> 

> When new symbols get exposed, it may happen that the symbol is partially
> renamed into SSA (i.e., it's renamed in some statements and not in
> others).  In particular, it may happen that the first reference to the
> symbol is a USE or VUSE operand.  In those cases, the initial SSA name
> for that symbol has been lost after the original rename pass.  Now, if
> the SSA renamer needs to create a default definition for a symbol, we
> keep it in the variable's annotation.  IIRC, Dan and Andrew had asked me
> about this a while ago.  Is this still useful to you, folks?

It has its uses, yes, but IIRC it was more useful to have a list of
them, as opposed to being able to ask if this partiticular one is
suppose to have a default definition. That being said, its not terribly
important to me at the moment :-)  So if a variable had a real
definition, and that got deleted or changed, the var would have no
definition, and the annotation would *not* have a default defintion
listed right? So we can tell that it is in error that it doesn't have a
definition... 

Also, this new symbol expose of partial ranges isnt going to result in
potential virtual/real use mixes is it?

Andrew