CPU Profiling on Mac OS Sierra

[bob.ziu...@gmail.com]

Hi,

I am attempting to profile a Go application but keep getting empty pprof data with go 1.7.3 on Mac Sierra (package installer version from https://golang.org/dl).  The profiling output exists, but the top10 output is meaningless.  Profiling the same code on ubuntu 16.04 with go 1.7.3 returns meaningful output.

I've tried profiling several different applications and I get the same meaningless profiling data for each on Mac OS Sierra.  The simplest shareable example I've tried is that of GitHub user eapache: https://gist.github.com/eapache/580b7dbd8800fe9a3234#file-test-go.  The output of that test is below:

Output on Mac OS Sierra:

$ go tool pprof profile.out

Entering interactive mode (type "help" for commands)

(pprof) top10

31.22s of 31.22s total (  100%)

      flat  flat%   sum%        cum   cum%

    31.22s   100%   100%     31.22s   100%

Output on Ubuntu 16.04:

$ go tool pprof profile.out 

Entering interactive mode (type "help" for commands)

(pprof) top10

21320ms of 34690ms total (61.46%)

Dropped 119 nodes (cum <= 173.45ms)

Showing top 10 nodes out of 97 (cum >= 660ms)

      flat  flat%   sum%        cum   cum%

    8910ms 25.68% 25.68%     8910ms 25.68%  runtime.memclr

    4030ms 11.62% 37.30%     4030ms 11.62%  runtime.memmove

    1860ms  5.36% 42.66%     4220ms 12.16%  runtime.sweepone

    1670ms  4.81% 47.48%     1670ms  4.81%  runtime/internal/atomic.Xchg

    1080ms  3.11% 50.59%     1080ms  3.11%  runtime/internal/atomic.Xadd

     880ms  2.54% 53.13%    20890ms 60.22%  runtime.mallocgc

     790ms  2.28% 55.41%     1740ms  5.02%  runtime.writebarrierptr_nostore1

     730ms  2.10% 57.51%    27300ms 78.70%  runtime.concatstrings

     710ms  2.05% 59.56%     1770ms  5.10%  runtime.(*mheap).allocSpanLocked

     660ms  1.90% 61.46%      660ms  1.90%  runtime.procyield

Is there anyone else running Mac OS Sierra who can check/confirm correct profiling functionality on Go 1.7.3?

Thanks,

Bob

[rhys.h...@gmail.com]

Try specifying the path to the binary as the second to last argument (see "go tool pprof -h").

On Linux as of Go 1.7, CPU profiles include information on what executables are mapped into memory. This allows the pprof tool to locate the binary. Raw CPU profiles taken on macOS don't include this information, so you need to specify the path to the binary by hand.

[Dave Cheney]

http://talks.godoc.org/github.com/davecheney/high-performance-go-workshop/high-performance-go-workshop.slide#36

See also think linked issue at the bottom of the slide.

[bob.ziu...@gmail.com]

That explains it.  Thank you very much for the info.  I've confirmed that does indeed resolve the issue.

Bob