http://gcc.gnu.org/bugzilla/show_bug.cgi?id=46590
--- Comment #10 from Richard Guenther <rguenth at gcc dot gnu.org> 2010-11-22 13:29:56 UTC --- At -O1 we have tree SSA rewrite : 23.32 (13%) usr 0.06 ( 3%) sys 22.80 (13%) wall 6392 kB ( 3%) ggc tree SSA incremental : 24.05 (14%) usr 0.07 ( 4%) sys 25.27 (14%) wall 3533 kB ( 2%) ggc tree FRE : 45.17 (25%) usr 0.51 (29%) sys 45.83 (26%) wall 2048 kB ( 1%) ggc dominance frontiers : 29.37 (17%) usr 0.01 ( 1%) sys 29.30 (16%) wall 0 kB ( 0%) ggc dominance computation : 19.49 (11%) usr 0.02 ( 1%) sys 19.29 (11%) wall 0 kB ( 0%) ggc loop invariant motion : 12.25 ( 7%) usr 0.01 ( 1%) sys 12.21 ( 7%) wall 648 kB ( 0%) ggc TOTAL : 177.31 1.74 179.57 196657 kB the testcase is simply large. And yes, the FRE/PRE alias stmt walks are not limited (compared to the DCE/DSE ones). Maybe it's time to add this to cover this kind of artificial testcases.