Profiling Go Program with Pprof and Trace packages

Profiling is a technique commonly used in the software development cycle to analyze the performance of a program, often to compare program differences or to identify procrastination problems and areas for improvement. . Profiling involves measuring and analyzing a variety of metrics such as memory usage, CPU usage, runtime, and other system-level metrics.

Profiling aims to identify resource 'consuming' parts of a program so that it can be optimized for better performance. Profiling also helps with debugging, memory management optimization, and concurrency tuning.

Profiling in Go

You have many tools for profiling in Go. Some other popular tools include the profiling tool pprof available in Go and popular third-party packages like Go Tool Trace and the Go-Torch package.

The pprof package is part of the runtime. The pprof package provides profiling logging functionality that runs in a format that the pprof visualization tool can interpret.

Here's how you can import the pprof package into a Go program:

import "pprof"

Go provides a number of commands and flags to work with the source code. Run the following tool command to access the profiling results in different formats.

go tool pprof

This command exports usage details to the pprof command .

Profiling Go Program with Pprof and Trace packages Picture 1

CPU Profiling in Go

CPU profiling measures the time a program spends while running functions. CPU profiling is useful in identifying the most CPU-hungry parts of code.

The pprof package provides functions to collect CPU profiles, start and stop CPU profiling, and a function to write profile data to a file.

Here's how to start and stop a CPU profile, writing data to a profiling file:

import (     "os"     "runtime/pprof" ) func main() {     f, err := os.Create("cpu_profile.prof")     if err != nil {         panic(err)     }     defer f.Close()     err = pprof.StartCPUProfile(f)     if err != nil {         panic(err)     }     defer pprof.StopCPUProfile()     // code to be profiled }

The main function creates a file and closes the file stream with a defer statement & the Close function of the file version. The StartCPUProfile function starts the CPU configuration and writes data to the file.  StopCPUProfile closes the configuration stream with a defer statement . After starting and stopping the CPU configuration, you can continue writing the code you want to analyze.

Here is the output of the pprof command with the profile file from this program:

Profiling Go Program with Pprof and Trace packages Picture 2

Running the pprof command with the file starts an interactive shell that allows you to explore profiling data. You can use commands like top and list to see which functions take the longest time to run.

Profiling memory in Go

Memory profiling is a technique used to identify leaks and excessive memory usage in code by measuring memory usage in code.

You can start creating a memory profile using the WriteHeapProfile function . This function takes the file version and writes the configuration data to the file.

import (     "os"     "runtime/pprof" ) func main() {     f, err := os.Create("mem_profile.prof")     if err != nil {         panic(err)     }     defer f.Close()     err = pprof.WriteHeapProfile(f)     if err != nil {         panic(err)     }     // code to be profiled }

The main function creates a profiling file. The WriteHeapProfile function takes the file version as an argument and returns a type of write error after writing the file.

Block profiling with Go

Block profiling measures how long a program waits for synchronization such as converters and channels. Block profiling is useful in identifying pieces of code that cause blocking problems.

The Lookup function returns a profile with the name of a specific string. Lookup 's WriteGo function writes a pprof- formatted snapshot of that configuration to the file.

Here's how you can implement block profiling for a Go program:

import (     "os"     "runtime/pprof" ) func main() {     f, err := os.Create("block_profile.prof")     if err != nil {         panic(err)     }     defer f.Close()     err = pprof.Lookup("block").WriteTo(f, 0)     if err != nil {         panic(err)     }     // code to be profiled }

This function creates a file that stores block profile data, finds blocks using the Lookup function , and writes block profile data to that file.

Trace Profiling with Go

Trace Profiling is a technique for measuring program execution, including goroutine scheduling and system calls. Trace Profiling is useful for identifying areas of 'congestion' and, at the same time, understanding the interactions between different program parts.

The trace package provides functions for trace profiling. This package is also part of the runtime package .

import (     "os"     "runtime/trace" ) func main() {     f, err := os.Create("trace.out")     if err != nil {         panic(err)     }     defer f.Close()     err = trace.Start(f)     if err != nil {         panic(err)     }     defer trace.Stop()     // code to be profiled }

This program creates a trace file to store trace data, starts a tracer using the Start function that takes the file version and returns an error type, and delays the tracer using the Stop function .

Above is how to profiling a Go program . Hope the article is useful to you.

Kareem Winters

Update 13 March 2023