Part 8 · 1 chapters · ~8 min

Performance and Profiling

pprof for CPU, heap, goroutines, mutexes and blocking, benchmarks with allocation counts, reducing allocations (escape analysis, sync.Pool, preallocation), strings and bytes, JSON encoding costs and faster alternatives, profile-guided optimisation, and tracing with go tool trace.

Profiles and allocations

code

import _ "net/http/pprof"     // exposes /debug/pprof on the default mux (serve it on an internal port only)
go func() { log.Println(http.ListenAndServe("127.0.0.1:6060", nil)) }()

go tool pprof -http=:8081 http://127.0.0.1:6060/debug/pprof/profile?seconds=30   # CPU flame graph
go tool pprof http://127.0.0.1:6060/debug/pprof/heap                            # in-use memory
curl -s http://127.0.0.1:6060/debug/pprof/goroutine?debug=2 | head              # every goroutine's stack
go test -bench=. -benchmem -cpuprofile cpu.out ./ledger                          # profile a benchmark
go build -pgo=default.pgo                                                        # profile-guided optimisation (1.21+)

common cost	fix
many small allocations in hot loops	preallocate slices (`make([]T, 0, n)`), reuse buffers, sync.Pool
string concatenation in loops	`strings.Builder`
encoding/json reflection cost	fewer fields, json/v2 or code-generated encoders for hot paths
lock contention	mutex profile; shard or remove the lock
too many goroutines blocked	goroutine profile; bounded worker pools