Drop deferredNodes after checking them all.

[DanielRosenwasser]

Test to see if we can free up some memory once a file is fully type-checked.

Based on comment from #52605 (comment)

[DanielRosenwasser]

@typescript-bot perf test this faster

[typescript-bot]

Heya @DanielRosenwasser, I've started to run the abridged perf test suite on this PR at 8bc58c0. You can monitor the build here.

Update: The results are in!

[typescript-bot]

@DanielRosenwasser
The results of the perf run you requested are in!

Here they are:

Comparison Report - main..52606

Metric	main	52606	Delta	Best	Worst	p-value
Angular - node (v16.17.1, x64)
Memory used	359,220k (± 0.01%)	358,938k (± 0.00%)	-281k (- 0.08%)	358,923k	358,951k	p=0.005 n=6
Parse Time	4.18s (± 0.38%)	4.17s (± 0.12%)	~	4.17s	4.18s	p=0.418 n=6
Bind Time	1.23s (± 0.00%)	1.22s (± 0.52%)	-0.01s (- 0.81%)	1.21s	1.23s	p=0.008 n=6
Check Time	9.47s (± 0.31%)	9.46s (± 0.18%)	~	9.44s	9.49s	p=0.515 n=6
Emit Time	8.05s (± 0.16%)	8.08s (± 0.37%)	~	8.04s	8.12s	p=0.142 n=6
Total Time	22.93s (± 0.09%)	22.93s (± 0.15%)	~	22.88s	22.97s	p=1.000 n=6
Compiler-Unions - node (v16.17.1, x64)
Memory used	193,990k (± 0.03%)	194,513k (± 0.68%)	~	193,901k	197,219k	p=0.810 n=6
Parse Time	1.80s (± 0.46%)	1.80s (± 0.76%)	~	1.78s	1.82s	p=1.000 n=6
Bind Time	0.85s (± 0.48%)	0.85s (± 0.89%)	~	0.84s	0.86s	p=1.000 n=6
Check Time	10.45s (± 0.81%)	10.44s (± 0.44%)	~	10.38s	10.52s	p=0.808 n=6
Emit Time	3.06s (± 0.96%)	3.06s (± 0.96%)	~	3.02s	3.10s	p=0.622 n=6
Total Time	16.17s (± 0.54%)	16.15s (± 0.53%)	~	16.03s	16.29s	p=0.810 n=6
Monaco - node (v16.17.1, x64)
Memory used	343,278k (± 0.01%)	343,155k (± 0.01%)	-124k (- 0.04%)	343,111k	343,178k	p=0.005 n=6
Parse Time	3.15s (± 1.13%)	3.17s (± 1.91%)	~	3.10s	3.27s	p=0.467 n=6
Bind Time	1.12s (± 0.56%)	1.11s (± 0.00%)	-0.01s (- 0.89%)	1.11s	1.11s	p=0.008 n=6
Check Time	7.72s (± 0.31%)	7.74s (± 0.34%)	~	7.71s	7.78s	p=0.260 n=6
Emit Time	4.53s (± 0.61%)	4.52s (± 0.50%)	~	4.48s	4.54s	p=0.467 n=6
Total Time	16.52s (± 0.30%)	16.53s (± 0.32%)	~	16.45s	16.61s	p=0.687 n=6
TFS - node (v16.17.1, x64)
Memory used	299,769k (± 0.01%)	299,681k (± 0.01%)	-88k (- 0.03%)	299,665k	299,703k	p=0.005 n=6
Parse Time	2.49s (± 2.15%)	2.49s (± 1.18%)	~	2.44s	2.53s	p=0.936 n=6
Bind Time	1.26s (± 0.41%)	1.26s (± 0.96%)	~	1.25s	1.28s	p=0.533 n=6
Check Time	7.21s (± 0.47%)	7.22s (± 0.34%)	~	7.19s	7.26s	p=0.420 n=6
Emit Time	4.23s (± 0.86%)	4.24s (± 0.58%)	~	4.22s	4.29s	p=0.687 n=6
Total Time	15.18s (± 0.75%)	15.20s (± 0.20%)	~	15.15s	15.24s	p=1.000 n=6
material-ui - node (v16.17.1, x64)
Memory used	476,134k (± 0.01%)	475,946k (± 0.01%)	-188k (- 0.04%)	475,905k	475,975k	p=0.005 n=6
Parse Time	3.71s (± 0.24%)	3.71s (± 0.28%)	~	3.70s	3.73s	p=0.673 n=6
Bind Time	1.02s (± 0.00%)	1.02s (± 0.00%)	~	1.02s	1.02s	p=1.000 n=6
Check Time	18.13s (± 0.23%)	18.18s (± 0.32%)	~	18.12s	18.24s	p=0.127 n=6
Emit Time	0.00s (± 0.00%)	0.00s (± 0.00%)	~	0.00s	0.00s	p=1.000 n=6
Total Time	22.86s (± 0.17%)	22.92s (± 0.26%)	~	22.85s	22.98s	p=0.090 n=6
xstate - node (v16.17.1, x64)
Memory used	546,762k (± 0.02%)	546,701k (± 0.02%)	~	546,601k	546,876k	p=0.471 n=6
Parse Time	4.75s (± 0.35%)	4.76s (± 0.25%)	~	4.75s	4.78s	p=0.459 n=6
Bind Time	1.81s (± 3.98%)	1.81s (± 4.13%)	~	1.66s	1.85s	p=0.507 n=6
Check Time	3.08s (± 3.26%)	3.08s (± 2.94%)	~	3.02s	3.26s	p=1.000 n=6
Emit Time	0.09s (± 4.45%)	0.09s (± 0.00%)	~	0.09s	0.09s	p=0.405 n=6
Total Time	9.74s (± 0.56%)	9.74s (± 0.35%)	~	9.70s	9.80s	p=0.872 n=6

System

Machine Name	ts-ci-ubuntu
Platform	linux 5.4.0-135-generic
Architecture	x64
Available Memory	16 GB
Available Memory	15 GB
CPUs	4 × Intel(R) Core(TM) i7-4770 CPU @ 3.40GHz

Hosts

• node (v16.17.1, x64)

Scenarios

• Angular - node (v16.17.1, x64)
• Compiler-Unions - node (v16.17.1, x64)
• Monaco - node (v16.17.1, x64)
• TFS - node (v16.17.1, x64)
• material-ui - node (v16.17.1, x64)
• xstate - node (v16.17.1, x64)

Benchmark	Name	Iterations
Current	52606	6
Baseline	main	6

Developer Information:

Download Benchmark

[weswigham]

It's weird, because we can end up building up the list of deferred nodes piecewise while doing work requested for random other files, but once the file's checked it's definitely safe to discard the list - we certainly won't be referring to it again, anyway. We could probably clear out our unused identifier cache post-source-file-check, too, if we were a bit smarter about how we went about regenerating a list of unused identifier diagnostics in getSuggestionDiagnostics (it is basically redoing the checkUnusedIdentifiers call in checkSourceFile does, but reclassifying them as suggestions).