Compilebench tries to age a filesystem by simulating some of the disk IO common in creating, compiling, patching, stating and reading kernel trees. It indirectly measures how well filesystems can maintain directory locality as the disk fills up and directories age. This current test is setup to use the makej mode with 10 initial directories
Here is the compile bench project page. Similar to sqlite, this seems to be primarily a test of filesystem performance and not of processor.
Metrics (Intel) - phoronix/compilebenchsh - pid 27636 On_CPU 0.018 On_Core 0.146 IPC 1.244 Retire 0.353 (35.3%) FrontEnd 0.332 (33.2%) Spec 0.024 (2.4%) Backend 0.290 (29.0%) Elapsed 133.95 Procs 13 Maxrss 31K Minflt 14667 Majflt 15 Inblock 4248536 Oublock 21841952 Msgsnd 0 Msgrcv 0 Nsignals 0 Nvcsw 76281 (98.5%) Nivcsw 1131 Utime 4.819486 Stime 14.796258 Start 341798.71 Finish 341932.66
Metrics demonstrate a very small amount of time On_CPU, a large amount of I/O blocks being written/read and a very high percentage of voluntary context switches. All signs this workload is I/O bound on the filesystem.
Metrics (AMD) - phoronix/compilebenchsh - pid 7313 On_CPU 0.008 On_Core 0.135 IPC 1.144 FrontCyc 0.189 (18.9%) BackCyc 0.140 (14.0%) Elapsed 191.06 Procs 13 Maxrss 34K Minflt 15229 Majflt 15 Inblock 4261664 Oublock 21841968 Msgsnd 0 Msgrcv 0 Nsignals 0 Nvcsw 75263 (97.2%) Nivcsw 2185 Utime 6.600898 Stime 19.272151 Start 346179.43 Finish 346370.49
AMD metrics are similar.