What are the benefits over the built in pprof?

It's not at all a replacement. If you've used the built in pprof it's very detailed. All trees if you will.

Nitro provides the forest. Using this along with pprof helps narrow down where to focus your attention.

I've used it a bit and found top and -cum very useful. Nitro seems to combine memory and CPU(?) but besides that I don't understand the benefit.

Edit: The author expanded here: https://news.ycombinator.com/item?id=6022679

Nitro seems quite nice.

Dave Cheney also just released a profiling wrapper, would you mind elaborating on the differences?

Dave's and pprof are both similar in that they run the application and poll randomly which routine the execution is currently in.

The output of pprof and Dave's (which uses pprof) would look like:

       2  13.3%  13.3%        2  13.3% go/scanner.(*Scanner).next
       2  13.3%  26.7%        2  13.3% path.Clean
       1   6.7%  33.3%        3  20.0% go/scanner.(*Scanner).Scan
       1   6.7%  40.0%        1   6.7% main.hasPathPrefix
       1   6.7%  46.7%        3  20.0% main.mountedFS.translate
       1   6.7%  53.3%        1   6.7% path.Dir
       1   6.7%  60.0%        1   6.7% path/filepath.(*lazybuf).append
       1   6.7%  66.7%        1   6.7% runtime.findfunc
       1   6.7%  73.3%        2  13.3% runtime.makeslice
       1   6.7%  80.0%        2  13.3% runtime.mallocgc

They are telling you what percent of the time the polling was in that routine.

Nitro is very different. You put breaks at major sections in your application where you want to measure and label those breaks.

For instance if you wrote an application to take markdown files and convert them to html you may split that up into 3 sections and see the following output.

    import markdown files:
      4.664481ms (5.887625ms)         0.43 MB     4583 Allocs
    convert to html:
      65.196788ms (71.107809ms)      17.13 MB     70151 Allocs
    write html files:
      1.823434ms (72.960713ms)        0.12 MB     3720 Allocs

Hopefully the difference is clear. If you wanted to break it up further you can.

Perhaps you don't find it usable, that's fine. It's not for you.

Personally I find this high level really valuable. When I make changes... say tweaking a process or adding a new feature, it's great to see the difference between execution times, memory use & allocations. Using pprof it is rather hard to get that kind of information.

Thanks!