Re: [PATCH] Fix up reassoc (PR tree-optimization/58946)

2013-11-04 Thread Richard Biener
On Fri, 1 Nov 2013, Jakub Jelinek wrote: > Hi! > > My recent reassoc patch caused following regression (though, it only started > failing on this testcase with Andrew's ifcombine changes). > > The issue is that update_ops relies on walking the same stmts as get_ops > did, and uses has_single_use

[PATCH] Fix up reassoc (PR tree-optimization/58946)

2013-11-01 Thread Jakub Jelinek
Hi! My recent reassoc patch caused following regression (though, it only started failing on this testcase with Andrew's ifcombine changes). The issue is that update_ops relies on walking the same stmts as get_ops did, and uses has_single_uses (either directly or indirectly through is_reassociable