On Fri, May 06, 2005 at 10:18:15PM +0200, Björn Haase wrote: > When implementing "lowering" of SImode and HImode expressions to QImode > sequences by splitters after reload, quite a number of new optimization > opportunities show up that presently are not realized.
No, the real problem is representational. We should be exposing this earlier. I'll post something about this shortly... r~
