On 08/30/2013 11:07 AM, Robert O'Callahan wrote: > Being able to sample uniformly whether or not the thread is in a syscall is > important sometimes, so let's not lose that.
I am not proposing to lose that facility. Am only proposing (really, merely parroting Taras' suggestion) that it would be nice to have the option of some kind of adaptive backoff for threads blocked in syscalls. > To sample only non-blocked states can't we count CPU clocks in user space, > instead of the real-time clock, or something like that? I imagine there are multiple possible heuristics, and CPU clockery might be one of them. That said, AFAICS the core problem is that once a thread blocks and we drop the sampling rate, we have no obvious (direct, notification-based) way to know when it has resumed, and hence are in danger of losing high-frequency samples in gap between when it really resumed and when we (later) notice that it has. J _______________________________________________ dev-platform mailing list dev-platform@lists.mozilla.org https://lists.mozilla.org/listinfo/dev-platform
