[Bug tree-optimization/79489] Strange static branch prediction for n != 0

[marxin at gcc dot gnu.org]

https://gcc.gnu.org/bugzilla/show_bug.cgi?id=79489

--- Comment #4 from Martin Liška <marxin at gcc dot gnu.org> ---
Created attachment 40746
  --> https://gcc.gnu.org/bugzilla/attachment.cgi?id=40746&action=edit
Patch for nonzero on non-boolean types

With the patch, I get following numbers for CPU2006 SPEC:

HEURISTICS                               BRANCHES  (REL)  BR. HITRATE          
 HITRATE       COVERAGE COVERAGE  (REL)
nonzero (on trees)                           9230  20.3%       49.43%   54.40%
/  83.05%    51413332496   51.41G   4.7%

As mentioned by Honza, it's quite poor (54.4%). We can measure speed
improvement in next stage1.