https://gcc.gnu.org/bugzilla/show_bug.cgi?id=80838

--- Comment #3 from Markus Trippelsdorf <trippels at gcc dot gnu.org> ---
bootstrap-lto/PGO:

 Performance counter stats for 'g++ -Ofast -w tramp3d-v4.cpp' (10 runs):        
      16381.906087      task-clock (msec)         #    0.998 CPUs utilized     
      ( +-  0.32% )                                                            
                                1,408      context-switches          #    0.086
K/sec                    ( +-  0.31% )                                          
                 6      cpu-migrations            #    0.000 K/sec             
      ( +-  3.45% )                                                             
           270,871      page-faults               #    0.017 M/sec              
    63,434,476,091      cycles                    #    3.872 GHz               
      ( +-  0.27% )                                                             
    13,204,293,203      stalled-cycles-frontend   #   20.82% frontend cycles
idle     ( +-  0.25% )                                                          
    13,046,634,059      stalled-cycles-backend    #   20.57% backend cycles
idle      ( +-  0.03% )                                                         
    76,669,663,652      instructions              #    1.21  insn per cycle     
                                                  #    0.17  stalled cycles per
insn  ( +-  0.03% )                                                             
    16,930,988,799      branches                  # 1033.518 M/sec             
      ( +-  0.03% )                                                             
       619,469,535      branch-misses             #    3.66% of all branches   
      ( +-  0.16% )                                                             

      16.408429260 seconds time elapsed                                        
 ( +-  0.31% )

pure PGO:

 Performance counter stats for 'g++ -Ofast -w tramp3d-v4.cpp' (10 runs):

      15688.921067      task-clock (msec)         #    0.998 CPUs utilized     
      ( +-  0.11% )
             1,345      context-switches          #    0.086 K/sec             
      ( +-  0.10% )
                 6      cpu-migrations            #    0.000 K/sec             
      ( +-  6.72% )
           269,717      page-faults               #    0.017 M/sec              
    60,747,706,165      cycles                    #    3.872 GHz               
      ( +-  0.08% )
    13,442,559,819      stalled-cycles-frontend   #   22.13% frontend cycles
idle     ( +-  0.14% )
    12,919,375,998      stalled-cycles-backend    #   21.27% backend cycles
idle      ( +-  0.02% )  (83.31%)
    73,128,792,903      instructions              #    1.20  insn per cycle     
                                                  #    0.18  stalled cycles per
insn  ( +-  0.02% )
    16,607,093,842      branches                  # 1058.524 M/sec             
      ( +-  0.02% )
       617,220,915      branch-misses             #    3.72% of all branches   
      ( +-  0.17% )

      15.718059194 seconds time elapsed                                        
 ( +-  0.11% )

Reply via email to