On Sat, 12 Feb 2005 06:33:42 -0700 (MST), Roger Sayle <[EMAIL PROTECTED]> wrote: > > On Sat, 12 Feb 2005, Nathan Sidwell wrote: > > I question if it is better to fold early. As I've said before, I think > > the optimizations that fold performs should be turned into a proper SSA > > optimization phase% -- that can be repeatedly applied as necessary. > > As for a proper tree-ssa optimization pass, I believe Andrew Pinski's > proposed tree-combiner pass which attempts to merge/combine consecutive > tree statements and check whether the result can be folded, would fit > your description. > > However, the utility of early fold to the GCC compiler is much greater > than simply compile-time evaluating expressions with constant operands. > One of the reasons that fold is so fast, is that it can rely on the fact > that all of a trees operands have already been folded. In fact, much > of the middle-end makes or can make use of the assumptions that constant > operands to binary operators appear second, that a NOP_EXPRs only appear > where needed, that NEGATE_EXPR never wraps another NEGATE_EXPR and that > "x+x" and "2*x" are represented the same way.
And, for example, one problem I'm facing continuously is the lack of a canonical way of representing array accesses - both the C and C++ frontend emit different initial trees for that. And even the (maybe nonexistant) semantic differences of the middle-end ARRAY_REF vs. INDIRECT_REF/ADDR_EXPR is not clear to me. Richard.
