haskell-language-server/bench

Benchmarks

This folder contains a Shake script to simplify the performance analysis of HLS. It drives the ghcide-bench benchmark suite over a set of commits and experiments. To run it, use cabal bench. To configure it, edit bench/config.yaml. By default it compares HEAD with "origin/master"

Examples and experiments

The benchmark suites runs a set of experiments (hover, completion, edit, etc.) over all the defined examples (currently Cabal and lsp-types). Examples are defined in bench/config.yaml whereas experiments are coded in ghcide-bench/src/Experiments.hs.

Phony targets

The Shake script supports a number of phony targets that allow running a subset of the benchmarks:

• all : runs all the examples, unprofiled

• profiled-all : runs all the examples with heap profiling, assuming profilingInterval is defined

• Cabal-3.0.0.0 : runs the Cabal example, unprofiled

• profiled-Cabal-3.0.0.0 : runs the Cabal example, with heap profiling

• all-binaries : build all the HLS binaries for each of the versions under analysis

• etc

--help lists all the phony targets. Invoke it with:

cabal bench --benchmark-options="--help"

Targets:
  - bench-results/binaries/*/commitid
  - bench-results/binaries/HEAD/ghcide
  - bench-results/binaries/HEAD/ghc.path
  - bench-results/binaries/*/ghcide
  - bench-results/binaries/*/ghc.path
  - bench-results/binaries/*/*.warmup
  - bench-results/*/*/*/*.csv
  - bench-results/*/*/*/*.gcStats.log
  - bench-results/*/*/*/*.output.log
  - bench-results/*/*/*/*.eventlog
  - bench-results/*/*/*/*.hp
  - bench-results/*/*/*/results.csv
  - bench-results/*/*/results.csv
  - bench-results/*/results.csv
  - bench-results/*/*/*/*.svg
  - bench-results/*/*/*/*.diff.svg
  - bench-results/*/*/*.svg
  - bench-results/*/*/*/*.heap.svg
  - Cabal-3.0.0.0
  - lsp-types-1.0.0.1
  - all
  - profiled-Cabal-3.0.0.0
  - profiled-lsp-types-1.0.0.1
  - profiled-all