[Bug tree-optimization/87465] [8/9 Regression] Loop removal regression

[rguenth at gcc dot gnu.org]

https://gcc.gnu.org/bugzilla/show_bug.cgi?id=87465

Jakub Jelinek <jakub at gcc dot gnu.org> changed:

           What    |Removed                     |Added
----------------------------------------------------------------------------
             Status|UNCONFIRMED                 |NEW
   Last reconfirmed|                            |2018-09-29
   Target Milestone|9.0                         |8.3
     Ever confirmed|0                           |1

Richard Biener <rguenth at gcc dot gnu.org> changed:

           What    |Removed                     |Added
----------------------------------------------------------------------------
           Keywords|                            |missed-optimization
             Status|NEW                         |ASSIGNED
           Assignee|unassigned at gcc dot gnu.org      |rguenth at gcc dot gnu.org

--- Comment #2 from Richard Biener <rguenth at gcc dot gnu.org> ---
Agreed.  Note we do have such brute-forcing but only quite limited with
loop_niter_by_eval ().

Note that with -funroll-loops (or -O3) we peel the innermost loop
but then

Not unrolling loop 1: number of branches on hot path in the unrolled sequence
reaches --param max-peel-branches limit.

But I have a simple fix for the testcase...

Index: gcc/tree-ssa-loop-ivcanon.c
===================================================================
--- gcc/tree-ssa-loop-ivcanon.c (revision 264734)
+++ gcc/tree-ssa-loop-ivcanon.c (working copy)
@@ -368,8 +368,8 @@ tree_estimate_loop_size (struct loop *lo
            size->non_call_stmts_on_hot_path++;
          if (((gimple_code (stmt) == GIMPLE_COND
                && (!constant_after_peeling (gimple_cond_lhs (stmt), stmt,
loop)
-                   || constant_after_peeling (gimple_cond_rhs (stmt), stmt,
-                                              loop)))
+                   || !constant_after_peeling (gimple_cond_rhs (stmt), stmt,
+                                               loop)))
               || (gimple_code (stmt) == GIMPLE_SWITCH
                   && !constant_after_peeling (gimple_switch_index (
                                                 as_a <gswitch *> (stmt)),