Uh oh!
There was an error while loading. Please reload this page.
ci: Fix benchmarks - #591
Merged
Merged
Conversation
Bencher's reports LIST endpoint stopped inlining each report's `results` (it now carries only a `counts` summary), which fetch-main read as "report found, zero rows" — every !benchmark base fetch on every backend then exited 3 and silently fell back to a FULL base-side rerun (rebuild the base binaries + re-measure), ~5 min of redundant work per entry despite the rows sitting on bencher. Fetch each SHA-matched report by uuid (the single-report GET still returns full results) before row parsing; an inlined results array is still used directly if the API populates it again. Validated against live bencher at main@7da4132: the aiur-sharded-env/ISLB fetch that drove the fallback in run 32749751301 now returns its row (5 measures, empty missing list, exit 0), and a compile fetch at the same SHA exercises the multi-report-per-testbed aggregation.
The five new plots synced with the slug fallback ("aiur-sharded-env-
check: check-time"); name them in the house style — Aiur Env Check
Time / Throughput / Peak RAM Usage, Aiur Env Constants, Aiur Env
Shards. Verified via `ix bench plots --dry-run`.`. Benchmarks/**` replaces the hand-listed ". Benchmarks/Compile": the recursive form reaches all seven packages under Benchmarks/ (Catalog, CatalogReal, CatalogSpine, Compile, CompileFC, and Catalog's nested RelocFixture workspaces) while skipping dotted directories, so .lake dependency checkouts stay untouched. CompileFC was previously left pinned to an old toolchain on purpose; it now updates with the rest.
samuelburnhamforce-pushed
the
sb/bench-ci-fixes
branch
from
August 24, 2026 20:18
aef8bee to
360c9d6Comparesamuelburnham
commented
Aug 24, 2026
MemberAuthor
!benchmark ooc |
samuelburnham
commented
Aug 24, 2026
MemberAuthor
!benchmark aiur-sharded-env |
Contributor
|
| env/constant | check-time (main) | check-time (PR) | Δ% | throughput (const/s) (main) | throughput (const/s) (PR) | Δ% | peak-ram (main) | peak-ram (PR) | Δ% |
|---|---|---|---|---|---|---|---|---|---|
InitStd | 7.252 s | 8.004 s | +10.4% (1.10× slower) | 13.62K | 12.34K | -9.4% (1.10× slower) | 2.33 GiB | 2.31 GiB | -0.8% |
Array.extract_append | 517.0 ms | 517.3 ms | +0.1% | 3.12K | 3.12K | -0.1% | 1.20 GiB | 1.33 GiB | +10.3% (1.10× larger) |
ByteArray.utf8DecodeChar?_utf8EncodeChar_append | 279.4 ms | 279.5 ms | +0.0% | 9.95K | 9.95K | -0.0% | 1.28 GiB | 1.47 GiB | +14.7% (1.15× larger) |
Char.ofOrdinal_le_of_le | 273.9 ms | 274.3 ms | +0.1% | 10.13K | 10.12K | -0.1% | 1.72 GiB | 1.71 GiB | -0.9% |
Std.HashMap | 266.7 ms | 267.2 ms | +0.2% | 7.77K | 7.76K | -0.2% | 1.95 GiB | 1.87 GiB | -4.0% 🟢 |
_private.Init.Data.Range.Polymorphic.SInt.0.Int64.instRxcHasSize_eq | 264.8 ms | 265.3 ms | +0.2% | 7.07K | 7.06K | -0.2% | 1.42 GiB | 1.44 GiB | +0.9% |
String.append | 253.2 ms | 253.3 ms | +0.0% | 1.30K | 1.30K | -0.0% | 688.84 MiB | 707.29 MiB | +2.7% |
Nat.add_comm | 252.5 ms | 252.3 ms | -0.1% | 182.160 | 182.330 | +0.1% | 503.75 MiB | 526.25 MiB | +4.5% |
per-constant drill-down — Σ check 204.47s → 221.51s (+8.3%), Σ cost +0.0%; 0 cost mover(s), 96 time-only mover(s)
Each entry is one constant's own check within the whole-env run (deps ingress lazily; each dep is checked in its own entry; entries sum to the env) — not the full-closure scope of the headline --consts measurements.
Time-only movers (cost flat — scheduling/locality noise)
| constant | main | PR | Δtime | Δcost (Zisk) | driver |
|---|---|---|---|---|---|
BitVec.getLsbD_concat_zero | 1.0ms | 17ms | +1679.1% (17.79× slower) | +0.0% | — |
List.reverse_perm._f | 0.8ms | 12ms | +1406.7% (15.07× slower) | +0.0% | — |
Std.Sat.AIG.mkGate | 1.6ms | 15ms | +858.1% (9.58× slower) | +0.0% | — |
BitVec.setWidth_neg_of_le | 6.0ms | 37ms | +506.6% (6.07× slower) | +0.0% | — |
Std.DTreeMap.Internal.Impl.Const.insertMany_empty_list_singleton | 4.5ms | 19ms | +315.7% (4.16× slower) | +0.0% | — |
Contributor
|
| constant | check-time (main) | check-time (PR) | Δ% | throughput (const/s) (main) | throughput (const/s) (PR) | Δ% | peak-ram (main) | peak-ram (PR) | Δ% | constants (main) | constants (PR) | Δ% | shards (main) | shards (PR) | Δ% |
|---|---|---|---|---|---|---|---|---|---|---|---|---|---|---|---|
ISLB | 1m 39.2s | 1m 33.7s | -5.5% (1.06× faster) 🟢 | 1.85K | 1.95K | +5.8% (1.06× faster) 🟢 | 95.34 GiB | 94.25 GiB | -1.1% | 183,062 | 183,062 | +0.0% | 151 | 151 | +0.0% |
samuelburnham
enabled auto-merge
August 24, 2026 20:35
arthurpaulino
approved these changes
Aug 24, 2026
Uh oh!
There was an error while loading. Please reload this page.
samuelburnham added a commit
that referenced
this pull request
Aug 28, 2026
CompileFC builds against formal-conjectures pinned to a commit hash, which lean-update reports and leaves alone, and mathlib reaches it as an inherited dependency at v4.27.0. Bumping only its lean-toolchain — which #591 opted it into by globbing Benchmarks/** — therefore pairs a v4.33.1 toolchain with a v4.27.0 mathlib and cannot build. The v4.33.1 PR branch had exactly that bump as its sole remaining change. The action's lake_package_directory accepts literal paths, /* and /**, with no exclusion syntax, so the package list is enumerated instead. /** matches descendants only, hence Benchmarks/Compile listed alongside Benchmarks/Compile/**. The resulting set is unchanged apart from CompileFC: root, Catalog's two relocation fixtures, Compile, Compile/TruthMines, TruthMines.
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 freeto join this conversation on GitHub.
Already have an account?
Sign in to comment
2 participants
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.
!benchmarkcomment caching issue caused by bencher API drift, now restores properly from main whenfreshis not passed as an argumentBenchmarks/**subdirs