------- Additional Comments From steven at gcc dot gnu dot org 2005-09-19 11:07 ------- Even the bad timings for postreload-gcse are almost entirely to blame on RTL alias analysis: CPU: P4 / Xeon with 2 hyper-threads, speed 3194.18 MHz (estimated) Counted GLOBAL_POWER_EVENTS events (time during which processor is not stopped) with a unit mask of 0x01 (mandatory) count 1000000 samples % symbol name 4503 10.9264 rtx_equal_for_memref_p 3080 7.4736 canon_rtx 2298 5.5760 memrefs_conflict_p 2219 5.3844 find_base_term 2057 4.9913 ix86_find_base_term 1877 4.5545 nonoverlapping_component_refs_p 1474 3.5766 for_each_rtx_1 1412 3.4262 nonoverlapping_memrefs_p 1386 3.3631 addr_side_effect_eval 1279 3.1035 mems_in_disjoint_alias_sets_p 764 1.8538 get_addr 710 1.7228 exp_equiv_p 682 1.6549 invalidate 604 1.4656 check_dependence 587 1.4243 is_gimple_stmt 579 1.4049 cselib_invalidate_mem 553 1.3418 canon_true_dependence 532 1.2909 rtx_equal_p 406 0.9851 cse_insn 398 0.9657 decl_for_component_ref 359 0.8711 ldst_entry 341 0.8274 verify_expr 304 0.7376 walk_tree 272 0.6600 alias_sets_conflict_p 272 0.6600 base_alias_check
-- http://gcc.gnu.org/bugzilla/show_bug.cgi?id=23490
