https://gcc.gnu.org/bugzilla/show_bug.cgi?id=109691
--- Comment #2 from Andrew Pinski <pinskia at gcc dot gnu.org> --- Still working on this but I ran into an issue where we now remove a dead load but that load has an exception edge (due to -fnon-call-exceptions) and we don't cleanup the eh edge and get an ICE due to that. FAIL: g++.dg/pr55263.C -std=gnu++14 (internal compiler error: verify_flow_info failed) FAIL: g++.dg/tree-ssa/pr36766.C -std=gnu++14 (internal compiler error: verify_flow_info failed) Also still have to update a few other testcases. FAIL: gcc.dg/pr81192.c scan-tree-dump-times pre "(?n)find_duplicates: <bb .*> duplicate of <bb .*>" 1 FAIL: c-c++-common/goacc/kernels-alias-8.c scan-tree-dump-times ealias "clique 1 base 1" 2 kernels-alias-8.c is because we also dce a load I have to double check how to handle that one but -fno-tree-ccp might work ... gcc.dg/pr81192.c might need a similar thing now too but I am not 100% sure. pr81192.c should really be a Gimple testcase but gimple testcases were not around when the testcase was added.
