On Thu, Nov 7, 2013 at 2:44 PM, Joseph S. Myers <jos...@codesourcery.com> wrote: > On Thu, 7 Nov 2013, Richard Biener wrote: > >> Well, I'm betting that you'll re-invent sth like 'tree' just don't >> call it 'tree' ;) >> You need to transparently refer to constants, SSA names and decls >> (at least) as GIMPLE statement operands. You probably will make >> a "gimple statement operand" base class. Well - that's a 'tree' ;) > > Uses of constants and decls in expressions are why it's not obvious that > those get a static type different from expressions (or at least, they > probably do need a common base class). But my model of how "tree" should > ideally be split up has expressions (maybe including decls) as a > completely separate static type from types, and identifiers as yet another > static type, none of those inheriting from a common base class "tree".
Sure. Also you make containers no longer trees (TREE_CHAIN anyone, TREE_VEC, TREE_LIST ...). Should be the first step I think (already started some time ago with using VECs for some pieces). Richard. > -- > Joseph S. Myers > jos...@codesourcery.com
