> > It seems HAVE_HW_BREAKPOINT depends on PERF_EVENTS, I will double > > check this on Monday. If it is, would it be ok to have the overhead in > > default build? > > Are there benchmarking results quantifying the overhead of enabling > CONFIG_PERF_EVENTS=y?
No. I googled, but couldn't find any overhead measurement of enabling PERF_EVENTS. There are some about the overhead when events are enabled though. I tried to trace perf_event a bit, it seems software events cause some overhead, for example, in do_page_fault(): perf_sw_event(PERF_COUNT_SW_PAGE_FAULTS, 1, regs, addr); checks whether the event is enabled or not. For the other types, seems they create overhead only when they are started. So probably the overhead of enabling PERF_EVENTS is small. Ting _______________________________________________ dev-b2g mailing list [email protected] https://lists.mozilla.org/listinfo/dev-b2g
