[tree-ssa] Inlining vs gimple vs compound expressions

[law@redhat.com]

In message <1046376409.1390.362.camel@p4>, Andrew MacLeod writes:
 >However, this begs a different question. We should be able to do
 >whatever we want without even knowing about the existince of a CE node.
 >Of course, all you need is insertion routines. Actually  in this case, I
 >think all you need are the linkig routines?  You would be dealing with a
 >tree-statement_iterator, n'est pas?  So if you have routines to
 >link_before and link_after, then you dont care about anything else.  
 >
 >Is that correct?
 >
 >so things like:
 >
 >tree_stmt_iterator tsi_link_before      PARAMS ((tree_stmt_iterator, tree));
 >tree_stmt_iterator tsi_link_after       PARAMS ((tree_stmt_iterator, tree));
 >tree_stmt_iterator tsi_delink           PARAMS ((tree_stmt_iterator));
 >tree_stmt_iterator tsi_new_stmt_list    PARAMS ((tree, tree_stmt_anchor *));
 >tree_stmt_iterator tsi_stmt_list_start  PARAMS ((tree_stmt_anchor));
 >
 >
 >for instance.
 >
 >Does that give you all the functionality you need? or do you need
 >something else?
Well, I've got the inliner using the tree statement iterators (it was
a rather easy conversion).  So, yes, I'm pretty sure those routines would
be everything I need.

If you've got some code you want to throw my way to poke at, feel free.

jeff