Skip to content

Initial Performance Pass#876

Open
KetanReddy wants to merge 5 commits into
mainfrom
perf/first-pass
Open

Initial Performance Pass#876
KetanReddy wants to merge 5 commits into
mainfrom
perf/first-pass

Conversation

@KetanReddy

@KetanReddy KetanReddy commented Jun 4, 2026 •
edited
Loading

Copy link
Copy Markdown
Member

Optimizes the hot paths in core/player — data get/set, binding parse/construction,
expression evaluation, string/template resolution, and the view-resolution tree walk.
These run on every data change and view render, so per-call costs compound across a flow.

All changes preserve existing behavior and conventions, keep the file layout unchanged,
and are written to stay cheap once transpiled to ES5 for the Hermes mobile runtime
(prefer indexed loops / Object.assign / direct property access over spread, for…of,
and destructuring, which become heavier helper calls on-device).

What changed

Data model

  • data/local-model — batch set with structural sharing. A multi-binding
    transaction now clones each touched container at most once (tracked in an owned
    WeakSet) instead of re-cloning shared ancestors per binding. Single-binding sets keep
    using timm's setIn directly. The COW helper lives in data/owned-set-in.ts and
    mirrors timm's setIn exactly (intermediate creation, no-op short-circuit,
    symbol-preserving clone, progressive reads), so it's a behavior-preserving drop-in.
  • controllers/data (DataController.set) — === fast path before the deep
    dequal, indexed loops instead of reduce/map/destructuring, Set built without an
    intermediate array.

Bindings

  • BindingParser.parse reuses parserOptions directly when no overrides are passed and
    merges updates in place; BindingInstance constructor uses an indexed loop.

Expressions

  • The per-node execution context in the evaluator is built lazily, so leaf nodes
    (literals, identifiers, model refs, member access) skip the allocation.

String / template resolution

  • traverseObject no longer calls timm setIn per key (which re-cloned the whole object
    each time, ≈O(n²)); it now clones once, lazily, on first change. Hoisted the expression
    regex to a module constant and use match.index.

View resolver (computeTree)

  • Per-child folding into resolved uses the same copy-on-write setIn (shared
    data/owned-set-in.ts), cloning shared ancestors once per node instead of once per
    child. Guarded to nodes with more than one child — narrow nodes keep plain setIn
    so the WeakSet bookkeeping never becomes overhead where there's nothing to share. The
    fresh-per-node owned set guarantees the hook value / reused previousResult.value is
    never mutated (first write always clones), preserving the "unchanged → same reference"
    invariant the cache relies on.
  • Dependency set handoff — child deps are already folded into the model's read-deps,
    so the complete set is just getDependencies(); dropped the redundant per-node copy +
    re-merge and hand the (local, discarded-after) set off directly.
  • Object.assign instead of spread for the per-node options and children rebuild (cheaper
    once transpiled, behavior identical).

Benchmarks

  • Added *.bench.ts coverage for the data model, data controller, string-resolver,
    schema, binding, and an end-to-end Player.start / data-set re-resolve.
  • Broadened inputs to characterize gains across sizes: a large realistic nested structure
    for traverseObject, larger and scattered-path batch sets, larger data-controller
    batches, and a wide-node resolver view (many sibling children) that exercises the
    resolver COW path the existing deep-but-narrow view could not.

Benchmark results

Vitest bench, ops/sec, measured with controlled before/after A/Bs on the same machine.
Only changes above the ~±5% run-to-run noise floor are listed; everything else was
neutral, and there were no regressions.

Benchmark Before After Change
LocalModel set batch (15) 30.8K 114.7K +272%
DataController set batch (15, all changed) 23.8K 55.3K +133%
resolveDataRefs object — flat (30 keys) 61.2K 117.5K +92%
view resolver — data change on wide node (50 children) 6.9K 9.7K +39%
resolveDataRefs object — nested (depth 3) 238.0K 336.5K +41%
BindingParser.parse — foo.pets.1.name 1.35M 1.64M +22%
view resolver — initial resolve (wide: 50 children) 4.0K 4.7K +18%
view resolver — resolving from cache 49.8K 56.9K +14%
resolveExpressionsInString — single 2.61M 2.98M +14%
expression eval — {{foo}} = 1 + 3 1.45M 1.54M +7%
BindingParser.parse — cache hit 4.94M 5.26M +6%
DataController set — no-op / changed scalar 2.97M 3.12M +5%

The high-frequency data/binding micro-ops (single get/set, BindingInstance
construction) showed only low-single-digit gains on V8, but the loop / no-spread /
Object.assign rewrites target the ES5/Hermes output where the avoided helper calls
cost more — so on-device gains should exceed what these V8 numbers show.

Notes

  • All core/player unit tests and lint pass; no behavioral changes.
  • The batch set and resolver copy-on-write share one implementation
    (data/owned-set-in.ts) that mirrors timm's setIn semantics; both are guarded so the
    common single-write path is never penalized.
  • Numbers measured by stashing each change and re-recording the baseline on the same
    machine (the CI benchmark job compares each PR against the recorded baseline
    automatically).
  • Also includes a fix for updating the baseline metrics on main build

Change Type (required)

Indicate the type of change your pull request is:

  • patch
  • minor
  • major
  • N/A

Does your PR have any documentation updates?

  • Updated docs
  • No Update needed
  • Unable to update docs

Release Notes

Performance improvements to Player Core. For full explanation see original PR.

@KetanReddy KetanReddy requested a review from a team as a code owner June 4, 2026 19:05
@KetanReddy KetanReddy added the patch Increment the patch version when merged label Jun 4, 2026
@intuit-svc

intuit-svc commented Jun 4, 2026 •
edited
Loading

Copy link
Copy Markdown
Contributor

Benchmark Results

Comparison against baseline from main. ⚠️ = regression (>10% slower), ✅ = improvement (>5% faster)

core/player

Benchmark Current Baseline Change
core/player/src/tests/player.bench.ts > Player.start > minimal flow (cold start) 11.41K ops/s — new
core/player/src/tests/player.bench.ts > Player data set > set bound data triggers re-resolve 179.62K ops/s — new
core/player/src/tests/string-resolver.bench.ts > resolveExpressionsInString > single expression 1.07M ops/s — new
core/player/src/tests/string-resolver.bench.ts > resolveExpressionsInString > embedded expressions 449.34K ops/s — new
core/player/src/tests/string-resolver.bench.ts > resolveDataRefsInString > single ref 774.99K ops/s — new
core/player/src/tests/string-resolver.bench.ts > resolveDataRefsInString > multiple refs 272.75K ops/s — new
core/player/src/tests/string-resolver.bench.ts > resolveDataRefsInString > no refs (fast exit) 5.58M ops/s — new
core/player/src/tests/string-resolver.bench.ts > resolveDataRefs object > flat (30 keys) 45.72K ops/s — new
core/player/src/tests/string-resolver.bench.ts > resolveDataRefs object > nested (depth 3) 125.47K ops/s — new
core/player/src/binding/tests/binding.bench.ts > BindingInstance construction > from array (numeric + string segments) 1.53M ops/s — new
core/player/src/binding/tests/binding.bench.ts > BindingInstance construction > from string 1.31M ops/s — new
core/player/src/binding/tests/binding.bench.ts > BindingParser.parse (simple) > cache hit 1.61M ops/s — new
core/player/src/binding/tests/binding.bench.ts > BindingParser.parse (simple) > cache cold (fast-path) 929.22K ops/s — new
core/player/src/binding/tests/binding.bench.ts > BindingParser.parse (complex query) > foo.pets[name='frodo'].type 382.03K ops/s — new
core/player/src/binding/__tests__/parser.bench.ts > parser benchmarks > Resolving binding: foo.bar 947.43K ops/s 879.70K ops/s +7.7% ✅
core/player/src/binding/__tests__/parser.bench.ts > parser benchmarks > Resolving binding: foo.pets.1.name 586.44K ops/s 567.31K ops/s +3.4%
core/player/src/binding/__tests__/parser.bench.ts > parser benchmarks > Resolving binding: foo.pets.01.name 566.52K ops/s 548.71K ops/s +3.2%
core/player/src/binding/__tests__/parser.bench.ts > parser benchmarks > Resolving binding: foo.pets['01'].name 493.78K ops/s 489.72K ops/s +0.8%
core/player/src/binding/__tests__/parser.bench.ts > parser benchmarks > Resolving binding: foo.pets[01].name 545.79K ops/s 512.39K ops/s +6.5% ✅
core/player/src/binding/__tests__/parser.bench.ts > parser benchmarks > Resolving binding: foo.pets[name = "frodo"].type 323.16K ops/s 301.85K ops/s +7.1% ✅
core/player/src/binding/__tests__/parser.bench.ts > parser benchmarks > Resolving binding: foo.pets["name" = "sprinkles"].type 251.38K ops/s 239.32K ops/s +5.0% ✅
core/player/src/binding/__tests__/parser.bench.ts > parser benchmarks > Resolving binding: foo.pets["isDog" = false].type 322.23K ops/s 305.91K ops/s +5.3% ✅
core/player/src/binding/__tests__/parser.bench.ts > parser benchmarks > Resolving binding: foo.pets["isDog" = true].type 323.72K ops/s 310.29K ops/s +4.3%
core/player/src/binding/__tests__/parser.bench.ts > binding creation benchmarks > Resolving binding: foo.bar 634.80K ops/s 624.19K ops/s +1.7%
core/player/src/binding/__tests__/parser.bench.ts > binding creation benchmarks > Resolving binding: foo.pets.1.name 402.47K ops/s 404.21K ops/s -0.4%
core/player/src/binding/__tests__/parser.bench.ts > binding creation benchmarks > Resolving binding: foo.pets.01.name 407.87K ops/s 387.24K ops/s +5.3% ✅
core/player/src/binding/__tests__/parser.bench.ts > binding creation benchmarks > Resolving binding: foo.pets['01'].name 356.55K ops/s 360.02K ops/s -1.0%
core/player/src/binding/__tests__/parser.bench.ts > binding creation benchmarks > Resolving binding: foo.pets[01].name 401.74K ops/s 389.29K ops/s +3.2%
core/player/src/binding/__tests__/parser.bench.ts > binding creation benchmarks > Resolving binding: foo.pets[name = "frodo"].type 254.08K ops/s 240.57K ops/s +5.6% ✅
core/player/src/binding/__tests__/parser.bench.ts > binding creation benchmarks > Resolving binding: foo.pets["name" = "sprinkles"].type 205.45K ops/s 192.56K ops/s +6.7% ✅
core/player/src/binding/__tests__/parser.bench.ts > binding creation benchmarks > Resolving binding: foo.pets["isDog" = false].type 256.00K ops/s 226.09K ops/s +13.2% ✅
core/player/src/binding/__tests__/parser.bench.ts > binding creation benchmarks > Resolving binding: foo.pets["isDog" = true].type 259.72K ops/s 235.14K ops/s +10.5% ✅
core/player/src/data/tests/local-model.bench.ts > LocalModel get > get shallow 3.43M ops/s — new
core/player/src/data/tests/local-model.bench.ts > LocalModel get > get deep 2.92M ops/s — new
core/player/src/data/tests/local-model.bench.ts > LocalModel get > get array index 3.03M ops/s — new
core/player/src/data/tests/local-model.bench.ts > LocalModel set > set existing shallow 735.68K ops/s — new
core/player/src/data/tests/local-model.bench.ts > LocalModel set > set deep 450.16K ops/s — new
core/player/src/data/tests/local-model.bench.ts > LocalModel set > set new key 565.27K ops/s — new
core/player/src/data/tests/local-model.bench.ts > LocalModel set > set batch (15) 44.02K ops/s — new
core/player/src/data/tests/local-model.bench.ts > LocalModel delete > delete object key 124.41K ops/s — new
core/player/src/data/tests/local-model.bench.ts > LocalModel delete > delete array index 381.49K ops/s — new
core/player/src/expressions/__tests__/performance.bench.ts > Expression Parsing/Execution Benchmark > Parsing: {{foo}} = 1 + 3 (sync) 518.99K ops/s 441.41K ops/s +17.6% ✅
core/player/src/expressions/__tests__/performance.bench.ts > Expression Parsing/Execution Benchmark > Parsing: {{foo}} = 1 + 3 (async) 402.25K ops/s 369.56K ops/s +8.8% ✅
core/player/src/expressions/__tests__/performance.bench.ts > Expression Parsing/Execution Benchmark > Parsing: conditional(true, true, false) (sync) 519.47K ops/s 460.66K ops/s +12.8% ✅
core/player/src/expressions/__tests__/performance.bench.ts > Expression Parsing/Execution Benchmark > Parsing: conditional(true, true, false) (async) 453.77K ops/s 398.91K ops/s +13.8% ✅
core/player/src/expressions/__tests__/performance.bench.ts > Expression Parsing/Execution Benchmark > Parsing: {{foo}} = conditional({{bar}} > 0, true, false) (sync) 236.04K ops/s 195.52K ops/s +20.7% ✅
core/player/src/expressions/__tests__/performance.bench.ts > Expression Parsing/Execution Benchmark > Parsing: {{foo}} = conditional({{bar}} > 0, true, false) (async) 216.27K ops/s 170.11K ops/s +27.1% ✅
core/player/src/expressions/__tests__/performance.bench.ts > Expression Parsing/Execution Benchmark > Parsing: {{foo}} = conditional(conditional(true = false, false, true), conditional(false = false, true, false), conditional(true = true, false, true)) (sync) 174.59K ops/s 161.81K ops/s +7.9% ✅
core/player/src/expressions/__tests__/performance.bench.ts > Expression Parsing/Execution Benchmark > Parsing: {{foo}} = conditional(conditional(true = false, false, true), conditional(false = false, true, false), conditional(true = true, false, true)) (async) 168.71K ops/s 146.70K ops/s +15.0% ✅
core/player/src/expressions/__tests__/performance.bench.ts > Expression Parsing/Execution Benchmark > Parsing: {{foo}} = await(asyncTestFunction(1)) (sync) N/A N/A N/A
core/player/src/expressions/__tests__/performance.bench.ts > Expression Parsing/Execution Benchmark > Parsing: {{foo}} = await(asyncTestFunction(1)) (async) 295.16K ops/s 259.21K ops/s +13.9% ✅
core/player/src/expressions/__tests__/performance.bench.ts > Expression Parsing/Execution Benchmark > Parsing: {{foo}} = asyncTestFunction(1) (sync) 368.45K ops/s 300.53K ops/s +22.6% ✅
core/player/src/expressions/__tests__/performance.bench.ts > Expression Parsing/Execution Benchmark > Parsing: {{foo}} = asyncTestFunction(1) (async) 326.89K ops/s 287.64K ops/s +13.6% ✅
core/player/src/expressions/__tests__/performance.bench.ts > Expression Parsing/Execution Benchmark > Parsing: asyncTestFunction(1) (sync) 938.56K ops/s 849.70K ops/s +10.5% ✅
core/player/src/expressions/__tests__/performance.bench.ts > Expression Parsing/Execution Benchmark > Parsing: asyncTestFunction(1) (async) 731.01K ops/s 664.21K ops/s +10.1% ✅
core/player/src/expressions/__tests__/performance.bench.ts > Expression Parsing/Execution Benchmark > Parsing: {{foo}} = conditional(!{{bar}} == false, await(asyncTestFunction(1)), false) (sync) 203.76K ops/s 179.69K ops/s +13.4% ✅
core/player/src/expressions/__tests__/performance.bench.ts > Expression Parsing/Execution Benchmark > Parsing: {{foo}} = conditional(!{{bar}} == false, await(asyncTestFunction(1)), false) (async) 186.94K ops/s 160.62K ops/s +16.4% ✅
core/player/src/schema/tests/schema.bench.ts > SchemaController getType (uncached binding) > shallow binding 473.93K ops/s — new
core/player/src/schema/tests/schema.bench.ts > SchemaController getType (uncached binding) > deep binding (array+record, 6 segments) 162.06K ops/s — new
core/player/src/schema/tests/schema.bench.ts > SchemaController getApparentType (cached binding) > repeat same binding 3.05M ops/s — new
core/player/src/controllers/data/tests/controller.bench.ts > DataController set > set no-op 1.09M ops/s — new
core/player/src/controllers/data/tests/controller.bench.ts > DataController set > set changed scalar 505.05K ops/s — new
core/player/src/controllers/data/tests/controller.bench.ts > DataController set > set changed object 451.51K ops/s — new
core/player/src/controllers/data/tests/controller.bench.ts > DataController set > set batch (15, all changed) 21.57K ops/s — new
core/player/src/controllers/data/tests/controller.bench.ts > DataController get > get hot 2.77M ops/s — new
core/player/src/view/resolver/__tests__/index.bench.ts > resolver benchmarks > initial resolve 663.49 ops/s 627.37 ops/s +5.8% ✅
core/player/src/view/resolver/__tests__/index.bench.ts > resolver benchmarks > Resolving from cache 20.92K ops/s 18.15K ops/s +15.3% ✅
core/player/src/view/resolver/__tests__/index.bench.ts > resolver benchmarks > data changes 3.07K ops/s 2.97K ops/s +3.4%
core/player/src/view/resolver/__tests__/index.bench.ts > resolver benchmarks > data changes slow 649.58 ops/s 583.21 ops/s +11.4% ✅

plugins/async-node/core

Benchmark Current Baseline Change
plugins/async-node/core/src/__tests__/index.bench.ts > async node benchmarks > Resolve Async Node 1 times 14.55K ops/s 4.92K ops/s +195.7% ✅
plugins/async-node/core/src/__tests__/index.bench.ts > async node benchmarks > Resolve Async Node 5 times 14.50K ops/s 5.74K ops/s +152.3% ✅
plugins/async-node/core/src/__tests__/index.bench.ts > async node benchmarks > Resolve Async Node 10 times 10.30K ops/s 4.09K ops/s +151.7% ✅
plugins/async-node/core/src/__tests__/index.bench.ts > async node benchmarks > Resolve Async Node 50 times 3.39K ops/s 1.48K ops/s +129.9% ✅
plugins/async-node/core/src/__tests__/index.bench.ts > async node benchmarks > Resolve Async Node 100 times 1.72K ops/s 897.89 ops/s +92.1% ✅
plugins/async-node/core/src/__tests__/transform.bench.ts > async transform benchmarks > Resolve Async Node 1 times 7.35K ops/s 2.93K ops/s +150.9% ✅
plugins/async-node/core/src/__tests__/transform.bench.ts > async transform benchmarks > Resolve Async Node 5 times 8.29K ops/s 3.12K ops/s +165.4% ✅
plugins/async-node/core/src/__tests__/transform.bench.ts > async transform benchmarks > Resolve Async Node 10 times 7.15K ops/s 2.68K ops/s +167.1% ✅
plugins/async-node/core/src/__tests__/transform.bench.ts > async transform benchmarks > Resolve Async Node 50 times 2.91K ops/s 1.31K ops/s +122.8% ✅
plugins/async-node/core/src/__tests__/transform.bench.ts > async transform benchmarks > Resolve Async Node 100 times 1.66K ops/s 767.69 ops/s +115.7% ✅

react/player

Benchmark Current Baseline Change
react/player/src/asset/__tests__/index.bench.tsx > ReactAsset benchmarks > Render asset nested in 1 ReactAssets 625.58 ops/s 346.43 ops/s +80.6% ✅
react/player/src/asset/__tests__/index.bench.tsx > ReactAsset benchmarks > Bubble errors nested in 1 ReactAssets 1.13K ops/s 421.19 ops/s +168.1% ✅
react/player/src/asset/__tests__/index.bench.tsx > ReactAsset benchmarks > Render asset nested in 5 ReactAssets 617.85 ops/s 298.87 ops/s +106.7% ✅
react/player/src/asset/__tests__/index.bench.tsx > ReactAsset benchmarks > Bubble errors nested in 5 ReactAssets 1.11K ops/s 424.04 ops/s +161.4% ✅
react/player/src/asset/__tests__/index.bench.tsx > ReactAsset benchmarks > Render asset nested in 10 ReactAssets 626.49 ops/s 323.13 ops/s +93.9% ✅
react/player/src/asset/__tests__/index.bench.tsx > ReactAsset benchmarks > Bubble errors nested in 10 ReactAssets 885.06 ops/s 355.28 ops/s +149.1% ✅
react/player/src/asset/__tests__/index.bench.tsx > ReactAsset benchmarks > Render asset nested in 50 ReactAssets 504.47 ops/s 262.79 ops/s +92.0% ✅
react/player/src/asset/__tests__/index.bench.tsx > ReactAsset benchmarks > Bubble errors nested in 50 ReactAssets 282.74 ops/s 125.72 ops/s +124.9% ✅
react/player/src/asset/__tests__/index.bench.tsx > ReactAsset benchmarks > Render asset nested in 100 ReactAssets 321.64 ops/s 199.87 ops/s +60.9% ✅
react/player/src/asset/__tests__/index.bench.tsx > ReactAsset benchmarks > Bubble errors nested in 100 ReactAssets 120.88 ops/s 54.91 ops/s +120.2% ✅

@codecov

codecov Bot commented Jun 4, 2026 •
edited
Loading

Copy link
Copy Markdown

Bundle Report

Changes will increase total bundle size by 42.84kB (0.76%) ⬆️. This is within the configured threshold ✅

Detailed changes
Bundle name Size Change
plugins/common-expressions/core 410.12kB 3.52kB (0.87%) ⬆️
plugins/check-path/core 423.97kB 3.52kB (0.84%) ⬆️
plugins/stage-revert-data/core 387.65kB 3.52kB (0.92%) ⬆️
plugins/async-node/core 461.92kB 3.52kB (0.77%) ⬆️
plugins/markdown/core 671.2kB 3.52kB (0.53%) ⬆️
plugins/reference-assets/core 467.16kB 3.52kB (0.76%) ⬆️
plugins/beacon/core 405.34kB 3.52kB (0.88%) ⬆️
plugins/metrics/core 441.8kB 3.52kB (0.8%) ⬆️
plugins/common-types/core 484.58kB 3.52kB (0.73%) ⬆️
core/player 965.47kB 11.18kB (1.17%) ⬆️

Affected Assets, Files, and Routes:

view changes for bundle: plugins/markdown/core

Assets Changed:

Asset Name Size Change Total Size Change (%)
MarkdownPlugin.native.js 3.52kB 646.04kB 0.55%
view changes for bundle: plugins/reference-assets/core

Assets Changed:

Asset Name Size Change Total Size Change (%)
ReferenceAssetsPlugin.native.js 3.52kB 437.26kB 0.81%
view changes for bundle: plugins/common-types/core

Assets Changed:

Asset Name Size Change Total Size Change (%)
CommonTypesPlugin.native.js 3.52kB 413.41kB 0.86%
view changes for bundle: plugins/async-node/core

Assets Changed:

Asset Name Size Change Total Size Change (%)
AsyncNodePlugin.native.js 3.52kB 414.07kB 0.86%
view changes for bundle: core/player

Assets Changed:

Asset Name Size Change Total Size Change (%)
Player.native.js 3.52kB 406.71kB 0.87%
cjs/index.cjs 2.57kB 190.54kB 1.37%
index.legacy-esm.js 2.54kB 184.11kB 1.4%
index.mjs 2.54kB 184.11kB 1.4%
view changes for bundle: plugins/metrics/core

Assets Changed:

Asset Name Size Change Total Size Change (%)
MetricsPlugin.native.js 3.52kB 409.46kB 0.87%
view changes for bundle: plugins/check-path/core

Assets Changed:

Asset Name Size Change Total Size Change (%)
CheckPathPlugin.native.js 3.52kB 394.49kB 0.9%
view changes for bundle: plugins/common-expressions/core

Assets Changed:

Asset Name Size Change Total Size Change (%)
CommonExpressionsPlugin.native.js 3.52kB 388.88kB 0.91%
view changes for bundle: plugins/stage-revert-data/core

Assets Changed:

Asset Name Size Change Total Size Change (%)
StageRevertDataPlugin.native.js 3.52kB 380.83kB 0.93%
view changes for bundle: plugins/beacon/core

Assets Changed:

Asset Name Size Change Total Size Change (%)
BeaconPlugin.native.js 3.52kB 390.87kB 0.91%

@codecov

codecov Bot commented Jun 4, 2026

Copy link
Copy Markdown

Codecov Report

✅ All modified and coverable lines are covered by tests.
✅ Project coverage is 0.00%. Comparing base (238aa82) to head (fd1a8f6).
⚠️ Report is 1 commits behind head on main.

Additional details and impacted files 
@@     Coverage Diff     @@
##   main   #876   +/-   ##
===========================
===========================

☔ View full report in Codecov by Harness.
📢 Have feedback on the report? Share it here.

🚀 New features to boost your workflow:
  • ❄️ Test Analytics: Detect flaky tests, report on failures, and find test suite problems.

@KetanReddy

KetanReddy commented Jun 4, 2026

Copy link
Copy Markdown
Member Author

Ran the JVM perf tests locally because we don't have them set up in CI yet and here are the results

Benchmark Runtime Mode Score (base) Error (base) Score (PR) Error (PR) Units Improvement
BenchPlayerCreation.createHeadlessPlayer Hermes avgt 4.164 0.476 5.327 0.694 ms/op 127.93%
BenchPlayerCreation.createHeadlessPlayer J2V8 avgt 12.176 1.084 13.076 0.739 ms/op 107.39%
BenchPlayerCreation.createJSPlayer Hermes avgt 14.834 0.240 15.499 1.487 ms/op 104.48%
BenchPlayerCreation.createJSPlayer J2V8 avgt 8.507 0.081 9.559 0.582 ms/op 112.37%
BenchPlayerFlow.firstViewUpdate Hermes avgt 6.899 0.291 8.742 0.919 ms/op 126.71%
BenchPlayerFlow.firstViewUpdate J2V8 avgt 16.443 0.655 19.274 0.669 ms/op 117.22%
BenchRuntimeCreation.createHeadlessPlayer Hermes avgt 4.851 0.843 6.383 0.565 ms/op 131.58%
BenchRuntimeCreation.createHeadlessPlayer J2V8 avgt 13.294 1.338 15.605 0.631 ms/op 117.38%
BenchRuntimeCreation.createRuntime Hermes avgt 0.676 0.016 0.755 0.017 ms/op 111.69%
BenchRuntimeCreation.createRuntime J2V8 avgt 0.712 0.040 0.776 0.236 ms/op 108.99%

KVSRoyal
KVSRoyal reviewed Jun 17, 2026
View reviewed changes

@KVSRoyal KVSRoyal left a comment •
edited
Loading

Copy link
Copy Markdown
Member

Choose a reason for hiding this comment

The reason will be displayed to describe this comment to others. Learn more.

Overall, looks fine. Might be helpful to add more a few more comments flagging where things have been done one way for performance to prevent unintended changes later.

Comment thread core/player/src/binding/binding.ts
) {
const split = Array.isArray(raw) ? raw : raw.split(".");
this.split = split.map((segment) => {
const normalized: RawBindingSegment[] = new Array(split.length);

@KVSRoyal KVSRoyal Jun 17, 2026

Copy link
Copy Markdown
Member

Choose a reason for hiding this comment

The reason will be displayed to describe this comment to others. Learn more.

I'm thinking it might be helpful to drop a comment here that doing the manual for-loop is for better performance. Might help prevent this accidentally getting changed later. Thoughts?

Comment thread core/player/src/__tests__/player.bench.ts
@@ -0,0 +1,95 @@
import { bench, describe } from "vitest";

@KVSRoyal KVSRoyal Jun 17, 2026

Copy link
Copy Markdown
Member

Choose a reason for hiding this comment

The reason will be displayed to describe this comment to others. Learn more.

Unrelated to this line; just making reply-able:

  1. Is this PR targeting memory performance only? (Not speed?)

  2. Are the before and after flipped in your table? Because this table is showing that everything has a significant increase in memory, which seems unintended?

    Benchmark Before After Change
    LocalModel set batch (15) 30.8K 114.7K +272%

Sign up for free to join this conversation on GitHub. Already have an account? Sign in to comment

Reviewers

@KVSRoyal KVSRoyal KVSRoyal left review comments

At least 1 approving review is required to merge this pull request.

Assignees

No one assigned

Labels

patch Increment the patch version when merged

Projects

None yet

Milestone

No milestone

Development

Successfully merging this pull request may close these issues.

3 participants

@KetanReddy @intuit-svc @KVSRoyal