Here are some things:

• The execution tracer, used to explain long pauses due to lock contention etc., is discussed a few places:
  • Francesc Campoy, of Google, talked about it at https://medium.com/@francesc/using-the-go-execution-tracer-to-speed-up-fractal-rendering-c06bb3760507. (Will say those example program’s problems are way more obvious than the ones you’d usually look for in real life.)
  • Filippo Valsorda, formerly of CloudFlare, gave a talk at https://www.youtube.com/watch?v=nsM_m4hZ-bA
• Not profiling, but https://segment.com/blog/allocation-efficiency-in-high-performance-go-services/ talks about improving some GC stuff with real examples.
• The testing tools’ benchmark facility is handy (can test both CPU speed and allocs), and discussed in http://go-talks.appspot.com/github.com/davecheney/presentations/writing-high-performance-go.slide#16 – looks like @hackintoshrao’s post covers it too, from the title
• Docs for runtime, expvar, and net/http/pprof point to some of the APIs you have. The tricks to expose data from a live server are neat.

These are, again, perf but not profiling, but I really want to write (and have started on) a super beginner-focused post about memory use. I wish someone would write a balanced intro to concurrency–from StackOverflow, a couple common mistakes are dividing work into very tiny tasks (with all the overhead that causes) when NumCPU workers and large chunks would do better, and using channels when you really do just want a lock or whatever kind of shared datastructure.

I’ve used torch for flamegraphs. It’s nice that the new web UI is going to include them out of the box.