Running newperf on /echo returns unsupported for stall_cycle_backend. How can I enable this counter? Intel CPUs (Ivy Bridge) fail, while AMD shows values.
$ newperf stat /echo
<unsupported> stall_cycle_backend
Running newperf on /echo returns unsupported for stall_cycle_backend. How can I enable this counter? Intel CPUs (Ivy Bridge) fail, while AMD shows values.
$ newperf stat /echo
<unsupported> stall_cycle_backend
hey, im not 100% sure but ive seen that intel sometimes misses some perf events on older kernels. might worth trying a kernel update or look at intel docs for more tweaks
In my experience, the issue stems from the inherent limitations of some Intel microarchitectures in supporting certain perf events. On Ivy Bridge processors, certain counters, including stall_cycle_backend, might be either undocumented or unsupported due to how the hardware implements performance monitoring. This contrasts with AMD systems where the same counters are fully implemented. Checking for microcode updates and verifying kernel support have offered limited improvements. Although alternative performance events may be used, the device limitation typically necessitates alternative tools or different event selections for accurate performance analysis.
hey, i noticed similar struggls on older systems. maybe its a kernel setting not enabled for ivy bridge? maybe your microcode/drivers need updating. has anyone found a workaround?
hey, im wonderin if this is more a kernel config issue than a strict intel cpu limit. maybe some intel quirks disable certain counters. has anyone tinkered with different kernel settings and seen a change? would love to know any experimnts.