CRIHAN  - Mars 2001
Outils d’optimisation - Page 23
3- D. perfex : analyse compilation en -O2/-O3
•Statistics                                                                      L1            L2           MEM
•                                                                            n =   20      n =  300      n = 1000
•=================================================================================================================
•Graduated instructions/cycle............................................     2.185076      1.710898      0.385768
•Graduated floating point instructions/cycle.............................     0.383449      0.353419      0.046896
•Graduated loads & stores/cycle..........................................     0.804269      0.712300      0.127843
•Graduated loads & stores/floating point instruction.....................     2.097459      2.015457      2.726113
•L1 Cache Line Reuse.....................................................106617.081506      6.744774      1.093389
•L2 Cache Line Reuse.....................................................   119.146341   8468.581597     26.642523
•L1 Data Cache Hit Rate..................................................     0.999991      0.870881      0.522306
•L2 Data Cache Hit Rate..................................................     0.991677      0.999882      0.963824
•Time accessing memory/Total time........................................     0.804341      1.551875      1.633047
•L1--L2 bandwidth used (MB/s, average per process).......................     0.079952    581.518247    381.962787
•Memory bandwidth used (MB/s, average per process).......................     0.078700      0.417081     56.276275
•MFLOPS (average per process)............................................    74.772563     68.916666      9.144644
•
•Graduated instructions/cycle............................................     1.835561      1.824490      1.692755
•Graduated floating point instructions/cycle.............................     0.804350      0.906906      0.810151
•Graduated loads & stores/cycle..........................................     0.559774      0.493832      0.441724
•Graduated loads & stores/floating point instruction.....................     0.695933      0.544524      0.545236
•L1 Cache Line Reuse..................................................... 54360.530320      7.210080      6.383013
•L2 Cache Line Reuse.....................................................    14.237530  22305.063745     64.097218
•L1 Data Cache Hit Rate..................................................     0.999982      0.878199      0.864554
•L2 Data Cache Hit Rate..................................................     0.934373      0.999955      0.984638
•Time accessing memory/Total time........................................     0.559917      1.036340      1.052615
•L1--L2 bandwidth used (MB/s, average per process).......................     0.111141    396.560745    397.235215
•Memory bandwidth used (MB/s, average per process).......................     0.059111      0.117584     35.873723
•MFLOPS (average per process)............................................   156.848197    176.846720    157.979467