FlameGraphs for Code Optimization with Golang and SpeedScope.

Profiling your Go programs for cpu and memory is easy and built into the go tooling (for free). I used to follow Fransesc Campoy’s Go Tooling in Action video to demonstrate how to do profiling, one part of which is getting some amazing flame charts to pin-point where your application is using resources most. Follow his video tutorial to learn how you can improve the performance of your Go programs.

In a nutshell, X-axis is time and Y-axis is the call stack. Finding the functions you wrote that has long spreads on the X-axis could help you identify areas you could rewrite for performance.

Go tooling provides an easy way to generate the profiles, but visualizing them is a common task across languages and therefore the latter part requires one to install some graphing tools — GraphViz, go-torch, and pprof. Honestly, it’s easy, but it could be easier.

Today i came across an online app that allows you to do the same: SpeedScope.app.

Here’s a quick way to try this.

Step 1: create a new directory and cd into it.

mkdir mydir && cd mydir

Step 2: Copy the code from the below into a file named, say, mycode_test.go

package mypkg
import (
 "log"
 "regexp"
 "strings"
 "testing"
)
var (
 s = "The quick brown fox"
 r = "^The.*$"
 pat = "The"
)
func BeginsWithRegex(s, r string) bool {
 ok, err := regexp.MatchString(r, s)
 if err != nil {
  log.Println(err)
 }
 return ok
}
func BeginsWithString(s, pat string) bool {
 return strings.HasPrefix(s, pat)
}
func Benchmark_BeginsWithRegex(b *testing.B) {
    for n := 0; n < b.N; n++ {
      BeginsWithRegex(s, r)
    }
}
func Benchmark_BeginsWithString(b *testing.B) {
    for n := 0; n < b.N; n++ {
      BeginsWithString(s, pat)
    }
}

Step 3: Run the following commands on the command line which will benchmark the two functions and create the corresponding cpu profile files.

go test -bench=BeginsWithString -cpuprofile=cpuprof_string.out
go test -bench=BeginsWithRegex -cpuprofile=cpuprof_regex.out

Step 4: Open https://www.speedscope.app/ and open the .out files which will show you the flame graph.

Try it out, it’s super easy.