I've run some data collection using perf stat set at an interval of 10ms. However, when looking at the output file, it seems like it's close to 10ms between runs but not the same. Is there somewhere in the perfwiki or perf-stat manpages that explains this?

My thought is if it uses interrupts it cannot achieve a perfect interval even with some offset, but I'm not sure. I would like to be able to cite some reason in the official documentation if possible.

I have attached my sample command and some sample output from the perf runs. The discrepancies start small but grows over time since I'm collecting data on long running binaries. $2 is the list of events passed in.

Command: sudo perf kvm --host --guest --guestkallsyms=/tmp/kallsysms --guestmodules=/tmp/modules stat -a -o "perf-stat.out" -e "$2" -I 10 sleep 1000.

Sample output 1:

#           time             counts unit events
     0.010105854         75,816,307      branch-loads                                                  (65.55%)
     0.010105854          6,665,278      branch-misses             #    9.73% of all branches          (65.54%)
     0.010105854         68,507,071      branches                                                      (95.03%)
     0.010105854          7,850,357      branch-load-misses                                            (99.97%)
     0.010105854          8,777,217      cache-references                                            
     0.010105854         56,341,006      ls_dispatch.store_dispatch                                     (74.06%)
     0.020916272        112,106,733      branch-loads                                                
     0.020916272         12,542,142      branch-misses             #   10.88% of all branches        
     0.020916272        115,317,475      branches                                                      (68.24%)
     0.020916272         12,615,815      branch-load-misses                                            (62.95%)
     0.020916272         15,555,283      cache-references                                              (69.91%)
     0.020916272         79,817,220      ls_dispatch.store_dispatch                                     (98.88%)
     0.031640789        113,604,284      branch-loads                                                  (62.59%)
     0.031640789         12,952,174      branch-misses             #   11.29% of all branches          (81.41%)
     0.031640789        114,720,636      branches                                                      (99.37%)
     0.031640789         12,952,273      branch-load-misses                                          
     0.031640789         15,583,838      cache-references                                              (92.92%)
     0.031640789         84,572,328      ls_dispatch.store_dispatch                                     (63.74%)

Sample output 2:

#           time             counts unit events
     0.010073045        127,931,342      ls_dispatch.ld_dispatch                                       (63.79%)
     0.010073045        191,430,277      L1-dcache-loads                                               (64.98%)
     0.010073045          5,523,387      L1-dcache-load-misses     #    2.89% of all L1-dcache accesses  (99.70%)
     0.010073045            805,856      L1-icache-load-misses                                       
     0.010073045          1,185,271       cache-misses                                               
     0.010073045        194,303,477       ls_dc_accesses                                               (71.63%)
     0.020798715        126,189,500      ls_dispatch.ld_dispatch                                     
     0.020798715        191,902,849      L1-dcache-loads                                               (98.73%)
     0.020798715          5,564,363      L1-dcache-load-misses     #    2.90% of all L1-dcache accesses  (62.98%)
     0.020798715            821,214      L1-icache-load-misses                                         (62.69%)
     0.020798715          1,228,591       cache-misses                                                 (75.58%)
     0.020798715        191,873,243       ls_dc_accesses                                             
     0.031456427        124,588,383      ls_dispatch.ld_dispatch                                       (62.47%)
     0.031456427        188,866,685      L1-dcache-loads                                               (88.05%)
     0.031456427          5,161,135      L1-dcache-load-misses     #    2.73% of all L1-dcache accesses
     0.031456427            802,225      L1-icache-load-misses                                       
     0.031456427          1,205,606       cache-misses                                                 (87.00%)
     0.031456427        186,961,816       ls_dc_accesses                                               (62.50%)