Metrics (Intel) - phoronix/n-queensThis is a test of the OpenMP version of a test that solves the N-queens problem. The board problem size is 18.
sh - pid 26949 On_CPU 1.000 On_Core 7.998 IPC 1.096 Retire 0.502 (50.2%) FrontEnd 0.035 (3.5%) Spec 0.395 (39.5%) Backend 0.068 (6.8%) Elapsed 34.94 Procs 10 Maxrss 10K Minflt 261 Majflt 0 Inblock 0 Oublock 16 Msgsnd 0 Msgrcv 0 Nsignals 0 Nvcsw 55 (3.5%) Nivcsw 1512 Utime 279.424005 Stime 0.011999 Start 762166.05 Finish 762200.99
The program is On_CPU for 100%. Interesting to see such a high amount of speculative misses.
Metrics (AMD) - phoronix/n-queenssh - pid 20321 On_CPU 0.999 On_Core 15.982 IPC 1.165 FrontCyc 0.011 (1.1%) BackCyc 0.097 (9.7%) Elapsed 17.99 Procs 18 Maxrss 10K Minflt 282 Majflt 0 Inblock 0 Oublock 16 Msgsnd 0 Msgrcv 0 Nsignals 0 Nvcsw 98 (0.4%) Nivcsw 27853 Utime 287.521327 Stime 0.000633 Start 10663.92 Finish 10681.91
Process Tree - phoronix/n-queens
Process Tree
The process tree is simple
26949) sh
26950) n-queens
26951) qn24b_openmp
26952) qn24b_openmp
26953) qn24b_openmp
26954) qn24b_openmp
26955) qn24b_openmp
26956) qn24b_openmp
26957) qn24b_openmp
26958) qn24b_openmp
About this graph
Overall shows all cores scheduled to run n-queens.
An IPC of ~1.1 overall.
About this graph
Bad speculation is the largest issue and not much problem with stalls.
retire 0.502 ms_uops 0.000 speculation 0.394 branch_misses 99.98% machine_clears 0.02% frontend 0.036 idq_uops_delivered_0 0.002 idq_uops_delivered_1 0.008 idq_uops_delivered_2 0.022 idq_uops_delivered_3 0.039 backend 0.068 resource_stalls.sb 0.000 stalls_ldm_pending 0.116
The bad speculation appears to be branch misses.
Next steps: Understand branch misses in n-queens.