[Bug tree-optimization/93243] misoptimization: minor changes of the code leads change up to +/- 30% performance on x86_64, -Os faster than -Ofast/O2/O3

[leo at yuriev dot ru]

https://gcc.gnu.org/bugzilla/show_bug.cgi?id=93243

Leo Yuriev <leo at yuriev dot ru> changed:

           What    |Removed                     |Added
----------------------------------------------------------------------------
           Keywords|missed-optimization         |
             Target|x86_64                      |
          Component|target                      |tree-optimization

--- Comment #1 from Leo Yuriev <leo at yuriev dot ru> ---
Minor additions:
  - benchmarks results above made with -Ofast.
  - one of gcc 9.x -Os result is below.

./heapsort-bench, cc 9.2.1 20191008
pass 1, small:
  0.805970 seconds, baseline
  0.799507 seconds, case-1, 99.2% of baseline
  0.784975 seconds, case-2, 97.4% of baseline
  0.789067 seconds, case-1+2, 97.9% of baseline
pass 1, large:
  2.859491 seconds, baseline
  2.779411 seconds, case-1, 97.2% of baseline
  2.782386 seconds, case-2, 97.3% of baseline
  2.743033 seconds, case-1+2, 95.9% of baseline

I.e. gcc 9.2.x with -Os faster then -Ofast.