Schedule tsc -b projects by dependency depth to reduce builder idle time on upstream projects by christianvuerings · Pull Request #64220 · microsoft/TypeScript · GitHub

Fixes #64222

tsc -b hands projects to builders in the depth-first order over references, and each builder blocks until the project's upstream is built. That order lists the whole subtree of the first reference before any sibling, so a builder that picks the root of a long chain sits idle while other builders work through the chain. On a large monorepo (19,430 composite projects, dependency graph 124 levels deep) only ~13 of 32 builders were busy on average during a cold build.

Change

computeScheduleOrder() stably sorts the topological order by dependency depth (no-upstream projects first, then their dependents, …) and rangeTask hands out projects in that order. When a builder picks a project at depth k, every project at a lower depth has already been picked up, though not necessarily finished, so this reduces avoidable blocking on upstream projects rather than eliminating it.
Results are reported by one goroutine walking Order(), so --verbose/--dry output and its order are unchanged and builders never block on reporting. This replaces the prevReporter/reportDone chain. A task drops its program once built (it is only needed for Testing.OnProgram), so finished-but-unreported tasks do not keep their programs alive.
graph_test.go asserts the schedule order for the existing graphs; ScheduleOrder() is exposed for that.

Results (cold full builds of the monorepo above, --builders 8, outputs deleted before each run, 4 runs each)

root references order	main	this PR
alphabetical	530 s median, peak RSS 24.7 GB	262 s median (2.0× faster), peak RSS 28.8 GB
pre-sorted by depth by hand	442 s	261 s

CPU time is identical (~3,150 s user), so the gain comes from less builder idle time; peak RSS is ~15% higher because more builders compile at once. --verbose output of the full build is byte-identical to main after stripping timestamps (98,000 lines).

Checks: go test -p 4 ./internal/execute/..., npx hereby lint, npx hereby check:format, go mod tidy -diff pass.

Disclosure: this patch was authored with an AI coding agent (Claude Code); I directed the investigation and the change, have read and understood the diff, and will handle review feedback myself.

🟢 Approval recommended

The implementation is coherent and adequately tested; only a minor documentation overstatement remains.

Pull request overview

Improves parallel tsc -b utilization by scheduling projects according to dependency depth while preserving output order.

Changes:

Adds stable dependency-depth scheduling.
Separates concurrent building from ordered reporting.
Tests schedule ordering across existing graph scenarios.

File summaries

File	Description
orchestrator.go	Implements scheduling and ordered reporting.
buildtask.go	Replaces chained reporting synchronization.
graph_test.go	Verifies expected schedule orders.

Review details

Files reviewed: 3/3 changed files
Comments generated: 1
Review effort level: Balanced

💡 Add a code-review agent skill for context-aware, tailored reviews. Learn more in the docs.

…on upstream projects Builders take projects from the topological order produced by the depth-first walk over references and block until each project's upstream is built. That order lists the whole subtree of the first reference before any sibling, so a builder that picks the root of a long chain sits idle while other builders work through the chain. Stably sort the schedule by dependency depth (projects with no upstream first), and report results from one goroutine that walks the original order so output is unchanged. A finished project releases its program before it is reported, so waiting-to-report tasks do not retain memory.

🔵 Needs a closer look

The concurrency-sensitive orchestration changes warrant final human review despite focused graph and existing parallel-build coverage.

Review details

Files reviewed: 3/3 changed files
Comments generated: 0 new
Review effort level: Balanced

🟢 Approval recommended

The scheduling, synchronization, memory release, and test updates are internally consistent.

Review details

Files reviewed: 3/3 changed files
Comments generated: 0 new
Review effort level: Balanced (auto)

Note

Copilot is running an experiment and ran this review at Balanced.

TypeScript Bot (@typescript-bot) perf test this faster

Starting jobs; this comment will be updated as builds start and complete.

Command	Status	Results
perf test this faster	✅ Started	👀 Results

Jake Bailey (@jakebailey)
The results of the perf run you requested are in!

Here they are:

tsc

Comparison Report - baseline..pr

Metric	baseline	pr	Delta	Best	Worst	p-value
Compiler-Unions - native
Errors	41	41	~	~	~	p=1.000 n=12
Symbols	115,480	115,480	~	~	~	p=1.000 n=12
Types	96,945	96,945	~	~	~	p=1.000 n=12
Memory Used	148,451k (± 0.52%)	148,337k (± 0.49%)	~	146,523k	150,217k	p=0.755 n=12
Memory Allocs	2,271,307 (± 0.00%)	2,271,233 (± 0.01%)	~	2,271,040	2,271,808	p=0.178 n=12
Config Time	0.000s	0.000s	~	~	~	p=1.000 n=12
Parse Time	0.041s (± 5.48%)	0.041s (± 3.36%)	~	0.037s	0.045s	p=0.832 n=12
Bind Time	0.012s (±12.12%)	0.012s (±10.73%)	~	0.010s	0.016s	p=0.976 n=12
Check Time	0.493s (± 0.87%)	0.495s (± 0.82%)	~	0.486s	0.507s	p=0.468 n=12
Emit Time	0.269s (± 1.51%)	0.273s (± 1.88%)	~	0.262s	0.289s	p=0.259 n=12
Total Time	0.822s (± 0.75%)	0.828s (± 0.70%)	~	0.816s	0.848s	p=0.155 n=12
angular-1 - native
Errors	3	3	~	~	~	p=1.000 n=12
Symbols	847,427 (± 0.11%)	847,074 (± 0.10%)	~	843,481	848,947	p=0.298 n=12
Types	249,851 (± 0.00%)	249,852 (± 0.00%)	~	249,849	249,854	p=0.296 n=12
Memory Used	793,309k (± 0.02%)	793,364k (± 0.08%)	~	792,034k	795,203k	p=0.920 n=12
Memory Allocs	12,953,050 (± 0.04%)	12,960,760 (± 0.10%)	~	12,948,081	13,023,673	p=0.242 n=12
Config Time	0.016s	0.016s (± 1.13%)	~	0.016s	0.017s	p=1.000 n=12
Parse Time	0.272s (± 2.63%)	0.271s (± 2.93%)	~	0.245s	0.285s	p=0.943 n=12
Bind Time	0.067s (±16.61%)	0.069s (±17.73%)	~	0.057s	0.110s	p=0.866 n=12
Check Time	0s	0s	~	~	~	p=1.000 n=12
Emit Time	1.646s (± 1.42%)	1.644s (± 1.58%)	~	1.607s	1.734s	p=0.514 n=12
Total Time	2.019s (± 1.41%)	2.018s (± 1.77%)	~	1.971s	2.132s	p=0.523 n=12
mui-docs - native
Errors	11,318 (± 0.04%)	11,315 (± 0.06%)	~	11,294	11,323	p=0.365 n=12
Symbols	4,082,255	4,082,255	~	~	~	p=1.000 n=12
Types	1,497,312	1,497,312	~	~	~	p=1.000 n=12
Memory Used	4,080,544k (± 0.05%)	4,080,706k (± 0.04%)	~	4,075,800k	4,085,056k	p=0.887 n=12
Memory Allocs	43,640,310 (± 0.05%)	43,635,675 (± 0.04%)	~	43,594,099	43,685,146	p=0.755 n=12
Config Time	0.016s (± 1.81%)	0.016s	~	~	~	p=0.217 n=12
Parse Time	0.553s (± 2.27%)	0.554s (± 1.58%)	~	0.531s	0.569s	p=1.000 n=12
Bind Time	0.002s	0.002s	~	~	~	p=1.000 n=12
Check Time	14.209s (± 0.36%)	14.172s (± 0.32%)	~	14.081s	14.304s	p=0.385 n=12
Emit Time	0.434s (± 0.27%)	0.437s (± 1.64%)	~	0.432s	0.473s	p=0.457 n=12
Total Time	15.942s (± 0.17%)	15.935s (± 0.11%)	~	15.894s	15.973s	p=0.434 n=12
strada-build-src - native
Errors	0	0	~	~	~	p=1.000 n=12
Symbols	1,394,252	1,394,252	~	~	~	p=1.000 n=12
Types	443,418	443,418	~	~	~	p=1.000 n=12
Memory Used	1,648,632k (± 0.22%)	1,660,833k (± 0.45%)	+12,201k (+ 0.74%)	1,650,157k	1,687,634k	p=0.001 n=12
Memory Allocs	97,069,284 (± 0.03%)	83,677,420 (± 0.05%)	🟩-13,391,864 (-13.80%)	83,516,709	83,789,528	p=0.000 n=12
Config Time	0.004s (± 7.93%)	0.004s (± 9.45%)	~	0.003s	0.005s	p=0.324 n=12
Parse Time	0.194s (± 1.83%)	0.270s (± 1.48%)	🔻+0.076s (+39.41%)	0.260s	0.280s	p=0.000 n=12
Bind Time	0.000s	0.000s (±217.90%)	~	0.000s	0.001s	p=1.000 n=12
Check Time	1.808s (± 0.53%)	1.797s (± 0.47%)	~	1.774s	1.820s	p=0.071 n=12
Emit Time	0.286s (± 6.96%)	0.304s (± 3.90%)	~	0.279s	0.341s	p=0.091 n=12
Total Time	24.849s (± 0.73%)	21.183s (± 0.75%)	🟩-3.666s (-14.75%)	20.882s	21.623s	p=0.000 n=12
strada-compiler - native
Errors	0	0	~	~	~	p=1.000 n=12
Symbols	337,531	337,531	~	~	~	p=1.000 n=12
Types	198,830	198,830	~	~	~	p=1.000 n=12
Memory Used	319,634k (± 0.07%)	319,497k (± 0.03%)	~	319,303k	319,719k	p=0.272 n=12
Memory Allocs	4,672,981 (± 0.01%)	4,672,809 (± 0.01%)	~	4,671,629	4,673,953	p=0.514 n=12
Config Time	0.001s	0.001s	~	~	~	p=1.000 n=12
Parse Time	0.114s (± 3.69%)	0.114s (± 3.80%)	~	0.102s	0.124s	p=0.991 n=12
Bind Time	0.000s	0.000s	~	~	~	p=1.000 n=12
Check Time	1.052s (± 0.41%)	1.048s (± 0.42%)	~	1.038s	1.062s	p=0.213 n=12
Emit Time	0.130s (±10.58%)	0.145s (± 7.17%)	~	0.110s	0.161s	p=0.110 n=12
Total Time	1.351s (± 0.96%)	1.357s (± 0.69%)	~	1.328s	1.375s	p=0.247 n=12
ts-pre-modules - native
Errors	87	87	~	~	~	p=1.000 n=12
Symbols	305,179	305,179	~	~	~	p=1.000 n=12
Types	181,876	181,876	~	~	~	p=1.000 n=12
Memory Used	277,164k (± 0.03%)	277,213k (± 0.03%)	~	277,022k	277,505k	p=0.401 n=12
Memory Allocs	1,641,729 (± 0.01%)	1,641,770 (± 0.01%)	~	1,641,343	1,642,609	p=0.966 n=12
Config Time	0.000s (±113.79%)	0.001s (±65.70%)	~	0.000s	0.001s	p=0.400 n=12
Parse Time	0.098s (± 3.97%)	0.102s (± 2.70%)	~	0.096s	0.108s	p=0.053 n=12
Bind Time	0.035s (± 9.39%)	0.040s (± 9.47%)	🔻+0.005s (+14.87%)	0.031s	0.050s	p=0.034 n=12
Check Time	0.839s (± 0.57%)	0.835s (± 0.38%)	~	0.828s	0.846s	p=0.224 n=12
Emit Time	0.000s	0.000s	~	~	~	p=1.000 n=12
Total Time	0.985s (± 0.83%)	0.990s (± 0.54%)	~	0.978s	1.001s	p=0.434 n=12
vscode - native
Errors	351	351	~	~	~	p=1.000 n=12
Symbols	8,856,348	8,856,348	~	~	~	p=1.000 n=12
Types	3,024,061	3,024,061	~	~	~	p=1.000 n=12
Memory Used	5,925,331k (± 0.02%)	5,924,969k (± 0.02%)	~	5,922,103k	5,928,825k	p=0.443 n=12
Memory Allocs	45,577,364 (± 0.03%)	45,578,133 (± 0.03%)	~	45,559,909	45,616,371	p=0.887 n=12
Config Time	0.064s (± 0.28%)	0.064s (± 0.48%)	~	0.064s	0.065s	p=0.317 n=12
Parse Time	1.687s (± 3.57%)	1.638s (± 2.82%)	~	1.558s	1.824s	p=0.291 n=12
Bind Time	0.499s (±15.78%)	0.553s (±14.77%)	~	0.370s	0.754s	p=0.298 n=12
Check Time	10.953s (± 1.56%)	10.829s (± 1.30%)	~	10.638s	11.301s	p=0.452 n=12
Emit Time	4.301s (±11.99%)	4.726s (± 9.07%)	~	3.346s	5.296s	p=0.198 n=12
Total Time	17.608s (± 2.12%)	17.921s (± 1.86%)	~	16.860s	18.388s	p=0.219 n=12
webpack - native
Errors	837	837	~	~	~	p=1.000 n=12
Symbols	1,125,807	1,125,807	~	~	~	p=1.000 n=12
Types	480,798	480,798	~	~	~	p=1.000 n=12
Memory Used	813,981k (± 0.02%)	814,021k (± 0.02%)	~	813,604k	814,429k	p=0.599 n=12
Memory Allocs	5,145,775 (± 0.03%)	5,145,299 (± 0.01%)	~	5,144,361	5,146,507	p=0.843 n=12
Config Time	0.008s	0.008s	~	~	~	p=1.000 n=12
Parse Time	0.237s (± 3.18%)	0.231s (± 2.45%)	~	0.219s	0.246s	p=0.224 n=12
Bind Time	0.062s (±16.90%)	0.069s (±14.91%)	~	0.048s	0.095s	p=0.310 n=12
Check Time	1.646s (± 0.62%)	1.638s (± 0.93%)	~	1.591s	1.667s	p=0.477 n=12
Emit Time	0.000s	0.000s	~	~	~	p=1.000 n=12
Total Time	1.987s (± 0.57%)	1.984s (± 0.91%)	~	1.943s	2.029s	p=0.723 n=12
xstate-main - native
Errors	0	0	~	~	~	p=1.000 n=12
Symbols	1,045,691	1,045,691	~	~	~	p=1.000 n=12
Types	381,854	381,854	~	~	~	p=1.000 n=12
Memory Used	611,209k (± 0.01%)	611,175k (± 0.01%)	~	610,873k	611,317k	p=0.701 n=12
Memory Allocs	4,826,717 (± 0.03%)	4,825,578 (± 0.03%)	~	4,823,153	4,830,289	p=0.178 n=12
Config Time	0.003s (± 6.23%)	0.003s (±10.34%)	~	0.002s	0.003s	p=0.590 n=12
Parse Time	0.150s (± 2.82%)	0.143s (± 3.94%)	~	0.131s	0.159s	p=0.053 n=12
Bind Time	0.040s (±14.69%)	0.050s (±16.32%)	🔻+0.010s (+24.17%)	0.033s	0.069s	p=0.049 n=12
Check Time	1.194s (± 0.59%)	1.191s (± 0.42%)	~	1.182s	1.210s	p=0.579 n=12
Emit Time	0.000s	0.000s	~	~	~	p=1.000 n=12
Total Time	1.400s (± 0.66%)	1.399s (± 0.34%)	~	1.388s	1.411s	p=0.522 n=12

System info unknown

Hosts

native

Scenarios

Compiler-Unions - native
angular-1 - native
mui-docs - native
strada-build-src - native
strada-compiler - native
ts-pre-modules - native
vscode - native
webpack - native
xstate-main - native

Benchmark	Name	Iterations
Current	pr	12
Baseline	baseline	12

Developer Information:

Download Benchmarks

Jake Bailey (@jakebailey) I tested this branch on a large monorepo and I do see performance gains, from an median of 38.24s down to 28.39s. While memory usage did increase a bit from ~5.4GB to ~6.9GB.

Most of the benchmarks used by the bot doesn't seem to use -b, where this optimization has the biggest impact.

Most of the benchmarks used by the bot doesn't seem to use -b, where this optimization has the biggest impact.

Yes, it's just one, xstate.

…g comments Compute each project's depth once and stably sort {config, depth} entries, so the comparator no longer normalizes paths or looks up tasks on every call. Describe depth ordering as reducing avoidable upstream blocking rather than eliminating it: a shallower project that has been picked up may not be done yet.

I feel like this PR is fine, though I know we had some prior concerns about changing build order due to odd things happening. But, I think with parallel building, that ship already sailed.

typescript-automation Bot added the For Uncommitted Bug PR for untriaged, rejected, closed or missing bug label Sep 9, 2026

Jake Bailey (jakebailey) requested a balanced review from Copilot September 9, 2026 22:46

Copilot started reviewing on behalf of Jake Bailey (jakebailey) September 9, 2026 22:46 View session

Copilot AI reviewed Sep 9, 2026

View reviewed changes

Comment thread tsc/internal/execute/build/orchestrator.go Outdated

Christian Vuerings (christianvuerings) force-pushed the build-schedule-by-depth branch from e754ddd to d8c7d1e Compare September 10, 2026 05:01

Jake Bailey (jakebailey) requested a balanced review from Copilot September 10, 2026 14:38

Copilot started reviewing on behalf of Jake Bailey (jakebailey) September 10, 2026 14:39 View session

Copilot AI reviewed Sep 10, 2026

View reviewed changes

Christian Vuerings (christianvuerings) marked this pull request as ready for review September 10, 2026 15:37

typescript-automation Bot added For Milestone Bug PRs that fix a bug with a specific milestone and removed For Uncommitted Bug PR for untriaged, rejected, closed or missing bug labels Sep 14, 2026

github-project-automation Bot added this to PR Backlog Sep 14, 2026

typescript-automation Bot assigned John Favret (johnfav03) Sep 14, 2026

github-project-automation Bot moved this to Not started in PR Backlog Sep 14, 2026

Jake Bailey (jakebailey) requested a balanced review from Copilot September 15, 2026 16:48

Copilot started reviewing on behalf of Jake Bailey (jakebailey) September 15, 2026 16:49 View session

Copilot AI reviewed Sep 15, 2026

View reviewed changes

Jake Bailey (jakebailey) reviewed Sep 21, 2026

View reviewed changes

Comment thread tsc/internal/execute/build/orchestrator.go Outdated

Comment thread tsc/internal/execute/build/orchestrator.go Outdated

Christian Vuerings (christianvuerings) changed the title Schedule tsc -b projects by dependency depth so builders do not idle on upstream projects Schedule tsc -b projects by dependency depth to reduce builder idle time on upstream projects Sep 22, 2026

Christian Vuerings (christianvuerings) requested a review from Jake Bailey (jakebailey) September 22, 2026 12:47

Jake Bailey (jakebailey) approved these changes Sep 22, 2026

View reviewed changes

github-project-automation Bot moved this from Not started to Needs merge in PR Backlog Sep 22, 2026

Jake Bailey (jakebailey) added this pull request to the merge queue Sep 22, 2026

Merged via the queue into microsoft:main with commit e4b844b Sep 22, 2026
28 checks passed

github-project-automation Bot moved this from Needs merge to Done in PR Backlog Sep 22, 2026

Jake Bailey (jakebailey) mentioned this pull request Oct 3, 2026

Restore dependency-depth build scheduling #64615

Open

Schedule tsc -b projects by dependency depth to reduce builder idle time on upstream projects by christianvuerings · Pull Request #64220 · microsoft/TypeScript · GitHub

Repository navigation

Schedule tsc -b projects by dependency depth to reduce builder idle time on upstream projects - #64220

Schedule tsc -b projects by dependency depth to reduce builder idle time on upstream projects#64220
Jake Bailey (jakebailey) merged 2 commits into
microsoft:mainfrom
christianvuerings:build-schedule-by-depth

Christian Vuerings (christianvuerings) commented Sep 9, 2026 •
edited
Loading

Uh oh!

Uh oh!

Copilot AI left a comment

Uh oh!

Uh oh!

Copilot AI left a comment

Uh oh!

Copilot AI left a comment

Uh oh!

Jake Bailey (jakebailey) commented Sep 15, 2026

Uh oh!

typescript-automation Bot commented Sep 15, 2026 •
edited
Loading

Uh oh!

Uh oh!

typescript-automation Bot commented Sep 15, 2026

Uh oh!

Hans-Albert Maritz (Freakazo) commented Sep 20, 2026

Uh oh!

Jake Bailey (jakebailey) commented Sep 21, 2026

Uh oh!

Uh oh!

Uh oh!

Jake Bailey (jakebailey) left a comment

Uh oh!

Uh oh!

Reviewers

Assignees

Labels

Projects

Milestone

Development

Uh oh!

5 participants

Schedule tsc -b projects by dependency depth to reduce builder idle time on upstream projects by christianvuerings · Pull Request #64220 · microsoft/TypeScript · GitHub

Repository navigation

Schedule tsc -b projects by dependency depth to reduce builder idle time on upstream projects - #64220

Conversation

Christian Vuerings (christianvuerings) commented Sep 9, 2026 • edited Loading Uh oh! There was an error while loading. Please reload this page.

Uh oh!

Uh oh!

Copilot AI left a comment

Choose a reason for hiding this comment

🟢 Approval recommended

Uh oh!

Uh oh!

Copilot AI left a comment

Choose a reason for hiding this comment

🔵 Needs a closer look

Uh oh!

Copilot AI left a comment

Choose a reason for hiding this comment

🟢 Approval recommended

Uh oh!

Jake Bailey (jakebailey) commented Sep 15, 2026

Uh oh!

typescript-automation Bot commented Sep 15, 2026 • edited Loading Uh oh! There was an error while loading. Please reload this page.

Uh oh!

Uh oh!

typescript-automation Bot commented Sep 15, 2026

tsc

Uh oh!

Hans-Albert Maritz (Freakazo) commented Sep 20, 2026

Uh oh!

Jake Bailey (jakebailey) commented Sep 21, 2026

Uh oh!

Uh oh!

Uh oh!

Jake Bailey (jakebailey) left a comment

Choose a reason for hiding this comment

Uh oh!

Uh oh!

Reviewers

Assignees

Labels

Projects

Milestone

Development

Uh oh!

5 participants

Christian Vuerings (christianvuerings) commented Sep 9, 2026 •
edited
Loading

typescript-automation Bot commented Sep 15, 2026 •
edited
Loading