lean4-htt/tests/bench
Sebastian Graf cae4decead
test: speed up bench/mvcgen/sym ctest entry (#13498)
This PR drops `LEAN_NUM_THREADS=1` from the `run_test.sh` of
`bench/mvcgen/sym`. The single-threaded restriction was originally there
to get reproducible benchmark timings, but `run_test.sh` runs as a test
rather than a benchmark, and we do not care about timing reproducibility
for tests. Allowing the default thread count cuts the wall time of what
was the slowest ctest entry from ~30s to ~20s.
2026-04-22 13:41:07 +00:00
..
build chore: don't fail on running build bench on built stage3 (#13467) 2026-04-18 22:07:21 +00:00
mergeSort feat: Array.mergeSort (#12385) 2026-03-06 13:18:13 +00:00
mvcgen test: speed up bench/mvcgen/sym ctest entry (#13498) 2026-04-22 13:41:07 +00:00
qsort chore: relative lean-toolchains (#12652) 2026-02-25 10:23:35 +00:00
size chore: improve how test suite interacts with stages (#12913) 2026-03-16 15:20:03 +00:00
sym chore: minor tweaks to Sym.simp test and benchmark (#13468) 2026-04-18 21:11:30 +00:00
.gitignore chore: switch to new test/bench suite (#12590) 2026-02-25 13:51:53 +00:00
accumulate_profile.py
arith_eval.ml
binarytrees.ghc-6.hs
binarytrees.ocaml-2.ml
binarytrees.st.hs
binarytrees.st.mlton-2.sml
binarytrees.st.sml
binarytrees.st.swift
binarytrees.swift
binarytrees5.ml
binarytrees5_multicore.ml
compile.sh
const_fold.hs
const_fold.ml
const_fold.sml
const_fold.swift
cross.yaml
dag_hassorry_issue.lean chore: deprecate levelZero and levelOne (#12720) 2026-03-04 01:03:08 +00:00
dag_hassorry_issue.lean.args
dag_hassorry_issue.lean.out.expected chore: improve how test suite interacts with stages (#12913) 2026-03-16 15:20:03 +00:00
delayed_assign.lean
deriv.hs
deriv.ml
deriv.sml
deriv.swift
flake.lock
flake.nix
full-stdlib.exec.yaml
ghc-gc.py
lean-gc.py
Makefile
mlkit-gc.py
ocaml-gc.py
perf.py
qsort.hs
qsort.ml
qsort.sml
qsort.swift
rbmap.hs
rbmap.ml
rbmap.sml
rbmap.swift
rbmap2.lean
rbmap3.lean
rbmap500k.lean
rbmap_checkpoint.hs
rbmap_checkpoint.ml
rbmap_checkpoint.sml
rbmap_checkpoint.swift
rbmap_checkpoint2.lean
rbmap_checkpoint2.sml
rbmap_checkpoint_cpp_lean3.cpp
rbmap_checkpoint_cpp_std.cpp
rbmap_cpp_lean3.cpp
rbmap_cpp_std.cpp
README.md
report.py
run.sh
speedcenter.yaml
states35.lean
test_single.sh
unionfind_clean.lean

Lean Benchmark Suites

This folder contains multiple small Lean programs for benchmarking used by two separate benchmark suites based on the temci benchmarking tool:

  • The light-weight "Speedcenter" suite benchmarks the current build of Lean. It can be used for quick comparisons on the cmdline and powers the Lean Speedcenter website.
  • The heavy-weight "Cross" suite benchmarks multiple Lean configurations and other functional compilers against each other and generates CSV and HTML reports from that. It was created for the paper "Counting Immutable Beans - Reference Counting Optimized for Purely Functional Programming" (IFL19).

Speedcenter Suite

Requirements:

  • A local Lean build in ../../build/release. Build at least the bin target.
  • temci. Using Nix, open a nix-shell in the project root directory to add a compatible version to your PATH. Alternatively, try pip3 install git+https://github.com/parttimenerd/temci.git.

To execute the suite and save the results in base.yaml, run (in this folder)

temci exec --config speedcenter.yaml --out base.yaml

Other interesting exec flags:

  • use --runs N to modify the default number of 10 runs per benchmark
  • use --included_blocks fast to excluded slow benchmarks like the stdlib benchmark. You can replace fast with any benchmark name or label in speedcenter.exec.yaml.

If you have multiple saved result files, you can compare them with

temci report --config speedcenter.yaml report1.yaml report2.yaml ...

Cross Suite

We recommend using Nix for building/obtaining all Lean variants and used compilers in a reproducible way. After installing Nix, running the benchmarks is as easy as

nix develop
make

This will record 50 runs for each benchmark configuration (this can be changed with runs in cross.yaml), generate results in report_lean.csv and report_cross.csv, and print them to stdout in a tabulated format. It will also generate HTML reports in report/ comparing the time-based benchmarks.

In order to reduce noise in the benchmarking data, you may instead want to try calling make inside a temci shell:

temci short shell --sudo --preset usable --cpuset_active make

Using root powers, this will temporarily configure your machine similarly to the LLVM benchmarking recommendations and move all your other processes to a single CPU core.