------- Comment #24 from steven at gcc dot gnu dot org 2007-11-05 22:29 ------- Implementing this kind of load PRE for RTL again is more trouble than it is worth. It would require tracking of addresses in MEM rtx'en, hashing the MEM and the MEM address independently but solving the dataflow equations with some kind of fake dependency (e.g. to avoid hoisting the a load but not the address of the load), and validating that an address replacement is correct.
-- steven at gcc dot gnu dot org changed: What |Removed |Added ---------------------------------------------------------------------------- CC| |steven at gcc dot gnu dot | |org Status|NEW |RESOLVED Resolution| |WONTFIX http://gcc.gnu.org/bugzilla/show_bug.cgi?id=23488