README - Pyroscope — ContextQMD

An example for setting up profile-guided optimization (PGO) using Pyroscope

Profile-Guided Optimization (PGO) is a Go compiler feature that uses runtime profiling data to optimize code. Now fully integrated in Go 1.21, PGO is a powerful tool to boost application performance.

PGO enhances performance primarily through two mechanisms:

Inlining hot (frequently executed) methods
Devirtualization of interface calls

Take a look at https://go.dev/doc/pgo and https://go.dev/blog/pgo for more information on PGO.

Using PGO in the Rideshare example application

Here are the steps needed to use PGO in the Rideshare example application.

Run the application that we will enable PGO for.
shell
```
docker-compose up --build --detach
```

Run the prepared benchmark

shell

go test -bench=BenchmarkApp -count=10 main_test.go

Extract a profile in pprof format with profilecli (see the Profile CLI documentation for further reference)
shell
```
profilecli query go-pgo \
    --query='{service_name="ride-sharing-app"}' \
    --from="now-1h" \
    --to="now" \
    --output=pprof=./default.pgo
```
This command will create a default.pgo (pprof) file in the current folder (/examples/golang-pgo/rideshare/).
Rebuild the Rideshare application. This will pick up the newly created PGO file automatically.
shell
```
docker-compose down
docker-compose up --build --detach
```

(Optional) Verify that the application was built with PGO

shell

docker exec -it golang-pgo-rideshare-go-1 /bin/bash
go version -m main
exit

You should see this line in the output:

shell

...
build	-pgo=/go/src/app/default.pgo
...

Run the benchmark again

shell

go test -bench=BenchmarkApp -count=10 main_test.go

(Optional) Use a tool such as benchstat to compare the two benchmarks.