benchproc: new package for high-level benchmark processing

This provides a transform/filter/projection processing model for
working with benchmark results, and mini-languages for constructing
filters and projections. This model is the result of quite a bit of
experimentation, and at this point I've built three tools on top of
this package, so I consider it reasonably well validated.

This also introduces a benchproc/syntax package, akin to the
regexp/syntax package. The syntax package documents the syntax and
provides low-level parsers for filter and projection expressions. If
we ever port the web UI over to this, this will make it easy to
transform these expressions into database filters.

This requires updating the language version in go.mod to 1.13 because
benchproc uses binary literals and signed shifts.

Change-Id: If9da95ec1e583594644c123f3bc3fc377a0e97e5
Trust: Austin Clements <>
Run-TryBot: Austin Clements <>
Reviewed-by: Michael Knyszek <>
TryBot-Result: Gopher Robot <>
24 files changed
tree: aa28852f4440701fcd32b85025a76eee707fdb62
  1. analysis/
  2. benchfmt/
  3. benchmath/
  4. benchproc/
  5. benchstat/
  6. benchunit/
  7. cmd/
  8. internal/
  9. storage/
  13. go.mod
  14. go.sum

Go performance measurement, storage, and analysis tools

Go Reference

This subrepository holds the source for various packages and tools related to performance measurement, storage, and analysis.

cmd/benchstat contains a command-line tool that computes and compares statistics about benchmarks.

cmd/benchsave contains a command-line tool for publishing benchmark results.

storage contains the benchmark result storage system.

analysis contains the benchmark result analysis system.

Both storage and analysis can be run locally; the following commands will run the complete stack on your machine with an in-memory datastore.

go install
go install
localperfdata -addr=:8081 -view_url_base=http://localhost:8080/search?q=upload: &
localperf -addr=:8080 -storage=http://localhost:8081

The storage system is designed to have a standardized API, and we encourage additional analysis tools to be written against the API. A client can be found in the storage package.


The easiest way to install is to run go install You can also manually git clone the repository and run go install

Report Issues / Send Patches

This repository uses Gerrit for code changes. To learn how to submit changes to this repository, see

The main issue tracker for the perf repository is located at Prefix your issue with “x/perf:” in the subject line, so it is easy to find.