On Sat, 01 Sep 2012 18:16:52 -0700, Ben Widawsky <ben at bwidawsk.net> wrote: > On 2012-09-01 12:22, Chris Wilson wrote: > > On Sat, 01 Sep 2012 11:35:13 -0700, Ben Widawsky <ben at bwidawsk.net> > > wrote: > >> I have no problem with Daniel's patch. It's just a matter of cutting > >> through some scheduler BS of "when the GPU wants to change > >> frequency" > >> vs. "when we actually change the GPU frequency." I think *both* are > >> interesting. > > > > We already trace the interrupt, which has been invaluable in > > debugging. And there is no other source currently... :) > > -Chris > > No we do not trace it (/me is looking on cgit, so perhaps a bit error > prone). Sure we trace GT interrupts, but not pm interrupts (again, > unless I am missing something via code browsing on the web). Oh we definitely do otherwise it would have been a bit difficult to have noticed that we woke up every 10ms to service a PM event that we then ignore... > Furthermore, even if we have added a generic interrupt trace event and > I've missed it: I think it's still nice to have a power/performance > specific one for users like powerTOP, as opposed to driver developers. As complete BS as GPU op/s? If you can find a tracepoint that is actionable from within powertop, go for it. -Chris -- Chris Wilson, Intel Open Source Technology Centre
