mirror of
https://github.com/protocolbuffers/protobuf-go.git
synced 2025-02-06 00:40:02 +00:00
39bbf13930
No need to call them multiple times, no matter how cheap they are. Mild
improvements, this CL is 01-cse-hasextension:
$ perflock -socket=@perflock -shared -cores 12 go test -tags=protolegacy -test.bench=BenchmarkExtension -test.benchmem -test.run=^# -test.count 10 internal/benchmarks/micro/micro_test.go
goarch: amd64
cpu: AMD Ryzen Threadripper PRO 3995WX 64-Cores
│ 00-cse-messageinfo │ 01-cse-hasextension │
│ sec/op │ sec/op vs base │
Extension/Has/None-12 103.30n ± 3% 96.73n ± 1% -6.36% (p=0.000 n=10)
Extension/Has/Set-12 113.0n ± 3% 107.1n ± 1% -5.22% (p=0.000 n=10)
Extension/Get/None-12 182.7n ± 2% 176.3n ± 2% -3.48% (p=0.000 n=10)
Extension/Get/Set-12 140.1n ± 2% 138.0n ± 1% -1.46% (p=0.024 n=10)
Extension/Set-12 218.6n ± 2% 219.5n ± 1% ~ (p=0.172 n=10)
geomean 145.6n 140.8n -3.25%
Change-Id: Ide1a0c0fe4e562ed24f88dc829249fca0f052d48
Reviewed-on: https://go-review.googlesource.com/c/protobuf/+/576315
LUCI-TryBot-Result: Go LUCI <golang-scoped@luci-project-accounts.iam.gserviceaccount.com>
Auto-Submit: Nicolas Hillegeer <aktau@google.com>
Reviewed-by: Michael Stapelberg <stapelberg@google.com>