Files
pnpm/.github/workflows/benchmark.yml
Andrew Stegmaier 11a7fddb74 perf(npm-resolver): cache disk-loaded metadata in memory on offline resolution (#12760)
When resolving with --offline or --prefer-offline, pickPackage loaded a
package's metadata from the on-disk cache mirror but never stored it in
the in-memory metaCache, so every subsequent resolution of the same
package (once per dependent that references it) re-read and re-parsed
the full packument from disk. Store the parsed metadata in the
in-memory cache on those disk-return paths, so each package's metadata
is parsed at most once per command. On a large monorepo this collapses
23,746 metadata loads to 5,124 (66.8 GB of JSON parsed to 8.8 GB) and
takes pnpm dedupe --offline from ~140s to ~75s with a byte-identical
lockfile. The cache is the existing bounded LRU; online and network
paths are untouched.

Promotion alone would have made the top-of-function cache hit terminal:
a disk-promoted packument that predates a later-requested range would
turn a previously-recoverable network fallback into
ERR_PNPM_NO_MATCHING_VERSION. Track registry-unverified disk promotions
(a WeakSet keyed by object identity in TypeScript; a registry_verified
flag carried inside the cache value in pacquet, so meta and state are
read atomically) and let a cache hit whose pick fails on such an entry
fall through to the regular network-validating flow unless offline. The
revalidating fetch stores a verified entry, so each package revalidates
at most once. The pre-existing exact-version fast paths on both stacks
had the same latent failure mode and are covered by the same marking.
Pacquet's PackageMetaCache::set_unverified is a required trait method
so custom cache implementations cannot silently inherit the terminal
behavior, and the TypeScript PackageMetaCache documents the
object-identity contract its WeakSet relies on.

Also registry-qualify pacquet's verifier shared-meta-cache reads
(read_shared_meta queried bare name keys while the resolver stores
registry-qualified ones, so the fast path could never hit), mirroring
the TypeScript readSharedMeta lookup order.

Add an isolated-linker.fresh-resolve.hot-cache.offline benchmark
scenario (install --offline --lockfile-only against a warm mirror, with
an untimed online pre-warm pass and per-iteration node_modules +
lockfile wipes so the up-to-date short-circuit cannot skip the measured
resolution) to guard offline resolution on all Bencher testbeds. Fixing
it to measure real work also revived the pnpm benchmark pipeline,
silently dead since the pnpm11/ move: bench.sh resolved REPO_ROOT to
pnpm11/ where no Cargo.toml exists, the orchestrator probed the
pre-move pnpm/dist bundle path, and continue-on-error masked every
failure as a green run. The scenario measures this change directly:
pnpm@HEAD resolves the fixture offline in 2.41s vs 3.17s for pnpm@main
(1.31x), with metadata-mirror reads dropping from 2625 to 1297.

---------

Co-authored-by: Zoltan Kochan <z@kochan.io>
2026-07-04 17:50:02 +02:00

212 lines
7.0 KiB
YAML

name: Benchmarks
on:
workflow_dispatch:
inputs:
pr_number:
description: 'PR number to benchmark (works with fork PRs too)'
required: false
default: ''
type: string
runs:
description: 'Number of benchmark runs per scenario'
required: false
default: '10'
type: string
warmup:
description: 'Number of warmup runs before timing'
required: false
default: '1'
type: string
push:
# Build Bencher's continuous baseline for the `pnpm` testbed.
# Every merge to main that touches the TypeScript stack re-runs the
# bench so PR comparisons have an up-to-date reference; cancel-in-progress
# stays off below so we never throw away a partial run.
branches: [main]
paths:
- 'pnpm11/**'
- 'package.json'
- 'pnpm-workspace.yaml'
- 'pnpm-lock.yaml'
- '.github/workflows/benchmark.yml'
permissions:
contents: read
pull-requests: write
checks: write
# Don't cancel-in-progress — killing a bench mid-run wastes a long
# CI job and produces no usable data.
concurrency:
group: benchmark-${{ inputs.pr_number || github.ref }}
cancel-in-progress: false
jobs:
benchmark:
name: Run Benchmarks
runs-on: ubuntu-latest
timeout-minutes: 180
steps:
- name: Checkout
uses: actions/checkout@9c091bb21b7c1c1d1991bb908d89e4e9dddfe3e0 # v7.0.0
with:
fetch-depth: 0
persist-credentials: false
- name: Checkout PR head
if: inputs.pr_number != ''
env:
GH_TOKEN: ${{ secrets.GITHUB_TOKEN }}
PR_NUMBER: ${{ inputs.pr_number }}
run: |
echo "Fetching PR #$PR_NUMBER head..."
git fetch origin "refs/pull/${PR_NUMBER}/head:refs/remotes/origin/pr-${PR_NUMBER}"
git checkout "origin/pr-${PR_NUMBER}"
echo "Checked out PR #$PR_NUMBER at $(git rev-parse --short HEAD)"
- name: Install pnpm and Node
uses: pnpm/setup@77cf06832101b3ac8c65caaf76a21643936d07a4
with:
runtime: node@26.4.0
- name: Install Rust Toolchain
uses: ./.github/actions/rustup
with:
shared-key: pnpm-benchmark
- name: Install hyperfine
uses: ./.github/actions/binstall
with:
packages: hyperfine@1.18.0
- name: Run benchmarks
id: bench
run: |
set -o pipefail
./pnpm11/benchmarks/bench.sh 2>&1 | tee bench-output.txt
BENCH_DIR=$(grep "Temp directory kept at:" bench-output.txt | sed 's/Temp directory kept at: //')
echo "bench_dir=$BENCH_DIR" >> "$GITHUB_OUTPUT"
env:
RUNS: ${{ inputs.runs }}
WARMUP: ${{ inputs.warmup }}
- name: Install Bencher CLI
if: steps.bench.outputs.bench_dir != ''
uses: bencherdev/bencher@ec56bb69a7f34096002ca3384d10e2f6676b063e # v0.6.7
- name: Upload results to Bencher
if: steps.bench.outputs.bench_dir != ''
env:
BENCHER_API_TOKEN: ${{ secrets.BENCHER_API_TOKEN }}
BENCH_DIR: ${{ steps.bench.outputs.bench_dir }}
EVENT_NAME: ${{ github.event_name }}
INPUT_PR_NUMBER: ${{ inputs.pr_number }}
REF_NAME: ${{ github.ref_name }}
GITHUB_TOKEN: ${{ secrets.GITHUB_TOKEN }}
run: |
if [ -z "${BENCHER_API_TOKEN:-}" ]; then
echo "::notice::BENCHER_API_TOKEN not set, skipping Bencher upload"
exit 0
fi
if [ ! -f "$BENCH_DIR/bencher-results.json" ]; then
echo "::warning::bencher-results.json not found, skipping upload"
exit 0
fi
# `bencher run --file` takes hyperfine JSON via the
# shell_hyperfine adapter. Branch policy:
# - push to main → record into the `main` branch (baseline)
# - workflow_dispatch with pr_number → record into `pr/<n>`,
# forked from main at the latest baseline
# - workflow_dispatch without pr_number → record into the
# ref's branch name (e.g. a feature branch), forked from main
args=(
--project pnpm
--testbed pnpm
--adapter shell_hyperfine
--file "$BENCH_DIR/bencher-results.json"
--github-actions "$GITHUB_TOKEN"
)
# `--start-point-clone-thresholds` so the forked branch inherits
# the threshold configured on main; `--err` so the workflow fails
# when a sample breaches the upper boundary. Main pushes skip
# both — by then the regression has already landed.
if [ "$EVENT_NAME" = "push" ] || [ "$REF_NAME" = "main" ]; then
args+=(--branch main)
elif [ -n "$INPUT_PR_NUMBER" ]; then
args+=(
--branch "pr/$INPUT_PR_NUMBER"
--start-point main
--start-point-reset
--start-point-clone-thresholds
--err
)
else
args+=(
--branch "$REF_NAME"
--start-point main
--start-point-reset
--start-point-clone-thresholds
--err
)
fi
bencher run "${args[@]}"
- name: Comment on PR
if: steps.bench.outputs.bench_dir != ''
env:
GH_TOKEN: ${{ secrets.GITHUB_TOKEN }}
BENCH_DIR: ${{ steps.bench.outputs.bench_dir }}
INPUT_PR_NUMBER: ${{ inputs.pr_number }}
REF_NAME: ${{ github.ref_name }}
RUN_ID: ${{ github.run_id }}
SERVER_URL: ${{ github.server_url }}
REPO: ${{ github.repository }}
RUNS: ${{ inputs.runs }}
ACTOR: ${{ github.actor }}
run: |
RESULTS_FILE="$BENCH_DIR/results.md"
if [ ! -f "$RESULTS_FILE" ]; then
echo "::warning::Results file not found at $RESULTS_FILE"
exit 0
fi
echo "--- Benchmark Results ---"
cat "$RESULTS_FILE"
echo "-------------------------"
if [ -n "$INPUT_PR_NUMBER" ]; then
PR_NUMBER="$INPUT_PR_NUMBER"
else
PR_NUMBER=$(gh pr list --head "$REF_NAME" --json number --jq '.[0].number' 2>/dev/null || echo "")
fi
if [ -z "$PR_NUMBER" ]; then
echo "::notice::No open PR found for branch $REF_NAME. Results printed above."
exit 0
fi
MARKER="<!-- pnpm-benchmark-results -->"
{
echo "$MARKER"
cat "$RESULTS_FILE"
echo ""
echo "_Run [${RUN_ID}](${SERVER_URL}/${REPO}/actions/runs/${RUN_ID}) · ${RUNS} runs per scenario · triggered by @${ACTOR}_"
} > /tmp/comment-body.md
COMMENT_ID=$(gh api "repos/${REPO}/issues/${PR_NUMBER}/comments" \
--jq "[.[] | select(.body | startswith(\"$MARKER\"))] | .[0].id // empty" 2>/dev/null || echo "")
if [ -n "$COMMENT_ID" ]; then
echo "Updating existing benchmark comment $COMMENT_ID on PR #$PR_NUMBER"
gh api "repos/${REPO}/issues/comments/${COMMENT_ID}" \
-X PATCH \
-F "body=@/tmp/comment-body.md"
else
echo "Creating new benchmark comment on PR #$PR_NUMBER"
gh pr comment "$PR_NUMBER" --body-file /tmp/comment-body.md
fi