fix(deps): upgrade mimalloc#13944
Draft
hardfist wants to merge 1 commit into
Draft
Conversation
github-actions
Bot
added
team
Contributor
|
📝 Benchmark detail: Open
Base persistent cache hit rate:
Threshold exceeded: ["arco-pro_development-mode_hmr + rss memory","arco-pro_production-mode_persistent-hot + rss memory","arco-pro_production-mode_source-map + rss memory","threejs_production-mode_10x_persistent-cold + rss memory"] |
Contributor
📦 Binary Size-limit
❌ Size increased by 28.04KB from 61.92MB to 61.95MB (⬆️0.04%) |
Contributor
Rsdoctor Bundle Diff AnalysisFound 6 projects in monorepo, 0 projects with changes. 📊 Quick Summary
Generated by Rsdoctor GitHub Action |
hardfist
force-pushed
the
codex/disable-mimalloc-thp
branch
from
61a1346 to
3dcb147
Compare
hardfist
changed the title
hardfist
force-pushed
the
codex/disable-mimalloc-thp
branch
from
3dcb147 to
9a8616c
Compare
hardfist
changed the title
Contributor
|
⏳ Triggered benchmark: Open |
hardfist
force-pushed
the
codex/disable-mimalloc-thp
branch
from
9a8616c to
bdde25c
Compare
Contributor
|
📝 Benchmark detail: Open
Base persistent cache hit rate:
Threshold exceeded: ["10000_big_production-mode_disable-minimize + exec","10000_development-mode + exec","10000_development-mode_noop-loader + exec","10000_production-mode + exec","10000_production-mode_persistent-cold + exec","10000_production-mode_persistent-hot + exec","10000_production-mode_source-map + exec","arco-pro_development-mode + exec","arco-pro_production-mode + exec","arco-pro_production-mode_generate-package-json-webpack-plugin + exec","arco-pro_production-mode_persistent-cold + exec","arco-pro_production-mode_persistent-hot + exec","arco-pro_production-mode_source-map + exec","arco-pro_production-mode_traverse-chunk-modules + exec","large-dyn-imports_development-mode + exec","large-dyn-imports_production-mode + exec","threejs_development-mode_10x + exec","threejs_production-mode_10x + exec","threejs_production-mode_10x_persistent-cold + exec","threejs_production-mode_10x_persistent-hot + exec","threejs_production-mode_10x_source-map + exec"] |
hardfist
force-pushed
the
codex/disable-mimalloc-thp
branch
3 times, most recently
from
22481ce to
c37553f
Compare
hardfist
changed the title
hardfist
changed the title
hardfist
force-pushed
the
codex/disable-mimalloc-thp
branch
2 times, most recently
from
8262baf to
0de2710
Compare
hardfist
changed the title
hardfist
force-pushed
the
codex/disable-mimalloc-thp
branch
from
0de2710 to
247fd34
Compare
hardfist
force-pushed
the
codex/disable-mimalloc-thp
branch
from
247fd34 to
2820dbd
Compare
hardfist
changed the title
Contributor
|
📝 Benchmark detail: Open
Base persistent cache hit rate:
Threshold exceeded: ["10000_big_production-mode_disable-minimize + exec","arco-pro_development-mode_hmr + rss memory","arco-pro_production-mode_persistent-hot + rss memory","threejs_production-mode_10x_persistent-cold + rss memory"] |
This file contains hidden or bidirectional Unicode text that may be interpreted or compiled differently than what appears below. To review, open the file in an editor that reveals hidden Unicode characters.
Learn more about bidirectional Unicode characters
Sign up for free
to join this conversation on GitHub.
Already have an account?
Sign in to comment
1 participant
Add this suggestion to a batch that can be applied as a single commit.This suggestion is invalid because no changes were made to the code.Suggestions cannot be applied while the pull request is closed.Suggestions cannot be applied while viewing a subset of changes.Only one suggestion per line can be applied in a batch.Add this suggestion to a batch that can be applied as a single commit.Applying suggestions on deleted lines is not supported.You must change the existing code in this line in order to create a valid suggestion.Outdated suggestions cannot be applied.This suggestion has been applied or marked resolved.Suggestions cannot be applied from pending reviews.Suggestions cannot be applied on multi-line comments.Suggestions cannot be applied while the pull request is queued to merge.Suggestion cannot be applied right now. Please check back later.
Summary
related to #13942
Upgrade
mimallocfrom0.1.48to0.1.50.This also removes the stale
v3feature usage fromrspack_allocatorandrspack_benchmark.mimalloc 0.1.50no longer exposes that feature; it already vendors mimalloc v3.This PR intentionally does not change mimalloc runtime defaults such as
allow_thp,arena_purge_mult,purge_delay, ordisallow_arena_alloc.Binding RSS Investigation
The observed regression is from the binding ecosystem benchmark, not the Rust benchmark. The benchmark samples the Node.js process RSS via
process.memoryUsage().rssafter the native.nodebinding runs a build.The RSS increase is not caused by a
v3.3.1arena purge behavior change relative tov3.3.0. The relevant arena purge defaults and main purge path are effectively the same betweenv3.3.0andv3.3.1:purge_delay = 1000msarena_purge_mult = 1purge_delay * arena_purge_multThe root cause is the larger allocator policy change from mimalloc
v3.1.5tov3.3.x, combined with the binding benchmark's sampling point:mimalloc 0.1.48vendors mimallocv3.1.5;mimalloc 0.1.50vendors mimallocv3.3.1.v3.3.xenables mimalloc's large page size class by default (MI_ENABLE_LARGE_PAGES=1;v3.1.5defaulted to0).v3.3.xadds Linux THP-aware purge behavior. When THP is available andallow_thpis enabled,_mi_os_minimal_purge_size()returns the large OS page size, usually2MiB, so arena purge avoids fragmenting THP.In short: the benchmark is observing retained arena memory from the newer mimalloc policy. This is an allocator tradeoff for less frequent decommit/recommit and less THP fragmentation, not a live-heap leak in Rspack.
Evidence
Same-machine artifact reruns for
threejs_production-mode_10x_persistent-cold:0.1.48/v3.1.52.83s744MiB0.1.50/v3.3.12.86s870MiBMIMALLOC_PURGE_DELAY=100.1.50/v3.3.12.64s759MiBMIMALLOC_PURGE_DELAY=200.1.50/v3.3.12.67s758MiBMIMALLOC_SHOW_STATS=1also points in the same direction:v3.1.5: usually one arena, about1GiBreserved, current committed around570-590MiB, purged around360-380MiB.v3.3.1: usually two arenas, about2GiBreserved, current committed around1.0GiB, purged around440-515MiB.Custom mimalloc instrumentation showed that the second arena is not caused by one huge allocation. It can be triggered by normal page allocations such as:
Options that reduce RSS were tested but are not used in this PR:
MIMALLOC_DISALLOW_ARENA_ALLOC=1lowers RSS but changes allocation strategy broadly and can slow binding builds.arena_purge_mult=0lowers RSS by making arena purge immediate, but it changes runtime allocator behavior and may increase decommit/recommit overhead.allow_thp=0or compile-timeMI_ENABLE_LARGE_PAGES=0only explains part of the RSS difference and changes allocator/runtime tradeoffs.arena_max_object_sizeto16M/32Mdid not materially recover RSS for this workload.Validation
cargo update -p mimalloc --precise 0.1.50cargo tree -p rspack_allocator --target x86_64-unknown-linux-gnu -i mimalloc -e featurescargo tree -p rspack_benchmark --target x86_64-unknown-linux-gnu -i mimalloc -e featurescargo check -p rspack_allocator --target x86_64-unknown-linux-gnucargo check -p rspack_benchmark --target x86_64-unknown-linux-gnucargo fmt --all --checkpnpm dlx @taplo/cli@0.7.0 format --check Cargo.toml crates/rspack_allocator/Cargo.toml xtask/benchmark/Cargo.tomlgit diff --checkRelated links
9199d54b: enabled mimalloc large page size class by default again.#1246: previous arena purge scheduling issue; the relevant fixes are already in the v3.3 line.Checklist