Update the zero test to not rely on hardware counters
Issue #34
new
Phil Mucci proposed modifying the zero (sanity) tests to use the following.
#ifdef Linux event[0] = PERF_COUNT_SW_TASK_CLOCK; event[1] = PERF_COUNT_SW_CPU_CLOCK; #endif
These work on every arch I’ve tested (ARM, PPC, x86, MIPS on kernels from 2.6.32 -> 4.40) that has a working perf substrate, regardless of whether the underlying PMU hardware is accessible or not, or whether or not we are running in a VM.
Comments (1)
@Vince Weaver do you have an update on this one? looks like there is nothing to be done here…