On Mon, Oct 21, 2019 at 4:38 AM Nidhi Agrawal <nidhi11...@gmail.com> wrote: > > Ok, Also is there any harm if we set block/mutex rate at the start of the > program ? Will it cause any memory allocations or any kind of performance > impact?
It will cause some additional memory allocations as the program records profiling information. But it won't be all that much memory. The other performance impact should be minimal. Ian > On Mon, Oct 21, 2019 at 2:47 PM Ian Lance Taylor <i...@golang.org> wrote: >> >> On Sun, Oct 20, 2019 at 8:57 PM Nidhi Agrawal <nidhi11...@gmail.com> wrote: >> > >> > Implementation in http/net/pprof library is leading me to conclude that >> > they are different. >> > Yes, In runtime/pprof block, mutex, cpu all are implemented in the same >> > way, but http/net/pprof is not using them in the same way. >> > In CPU profiling rate is set internally, while in block, mutex we are >> > supposed to set the rate. >> >> Yes, that is true. CPU profiling start with pprof.StartCPUProfile, >> while block and mutex profiling start with runtime.SetBlockProfilerate >> and runtime.SetMutexProfileFraction. But you don't have to set the >> block and mutex rate at the start of the program. You set it when you >> want to start profiling. >> >> You can also change the CPU profile rate with >> runtime.SetCPUProfileRate, but that is hard to use without >> pprof.StartCPUProfile. >> >> Ian >> >> >> > On Sun, Oct 20, 2019 at 1:27 PM Ian Lance Taylor <i...@golang.org> wrote: >> >> >> >> On Sun, Oct 20, 2019 at 12:51 AM Nidhi Agrawal <nidhi11...@gmail.com> >> >> wrote: >> >> > >> >> > I came to this conclusion because the pprof implemented the cpu with >> >> > the assumption that the client gives how much time to capture the >> >> > profiling data. but when it comes to the mutex and block, the pprof >> >> > didn't implement it to support this but asked the client to call the >> >> > profile rate before calling the pprof function. This made me question >> >> > why is there a difference in the profiling technique, is it expected >> >> > from us to set profile rate at the start of the program, if not then >> >> > why didn't they implement it similar to the cpu. >> >> >> >> I'm sorry, I don't really understand what you mean. >> >> >> >> There is really no difference among CPU, block, and mutex profiling >> >> with regard to how they are used. In all cases you start profiling, >> >> do some work, and then fetch the profile. They are all more or less >> >> the same. I don't understand what is leading you to conclude that >> >> they are different. >> >> >> >> Heap profiling, on the other hand, really is different, and it is >> >> documented differently in runtime/pprof. >> >> >> >> Ian >> >> >> >> >> >> > On Sun, Oct 20, 2019 at 1:10 PM Ian Lance Taylor <i...@golang.org> >> >> > wrote: >> >> >> >> >> >> On Sun, Oct 20, 2019 at 12:36 AM Nidhi Agrawal <nidhi11...@gmail.com> >> >> >> wrote: >> >> >> > >> >> >> > Because it is explicitly written in the documentation here >> >> >> > https://golang.org/pkg/net/http/pprof/ that we should set block >> >> >> > profile rate at the start of application, unlike CPU profiling where >> >> >> > the rate is being set internally before profiling starts. >> >> >> >> >> >> I'm likely missing something, but I'm looking at >> >> >> https://golang.org/pkg/net/http/pprof, and I don't see that. I see >> >> >> that it says that you have to call runtime.SetBlockProfileRate or >> >> >> runtime.SetMutexProfileFraction, but I don't see anything that says >> >> >> that you have to call it at the start of the application. >> >> >> >> >> >> Ian -- You received this message because you are subscribed to the Google Groups "golang-nuts" group. To unsubscribe from this group and stop receiving emails from it, send an email to golang-nuts+unsubscr...@googlegroups.com. To view this discussion on the web visit https://groups.google.com/d/msgid/golang-nuts/CAOyqgcUaR1kT98vbuEyYYofnSr-N%3D0aCYAz1d%2BNh%3DoW4MqVa_Q%40mail.gmail.com.
