Cache last computed flow type for identifiers and 'this'

[ahejlsberg]

Experiment to gauge effects on performance.

[ahejlsberg]

@typescript-bot test this
@typescript-bot user test this inline
@typescript-bot run dt
@typescript-bot perf test faster
@typescript-bot test top100

[typescript-bot]

Heya @ahejlsberg, I've started to run the extended test suite on this PR at cc1dd83. You can monitor the build here.

[typescript-bot]

Heya @ahejlsberg, I've started to run the parallelized Definitely Typed test suite on this PR at cc1dd83. You can monitor the build here.

Update: The results are in!

[typescript-bot]

Heya @ahejlsberg, I've started to run the diff-based user code test suite on this PR at cc1dd83. You can monitor the build here.

Update: The results are in!

[typescript-bot]

Heya @ahejlsberg, I've started to run the abridged perf test suite on this PR at cc1dd83. You can monitor the build here.

Update: The results are in!

[typescript-bot]

Heya @ahejlsberg, I've started to run the diff-based top-repos suite on this PR at cc1dd83. You can monitor the build here.

Update: The results are in!

[typescript-bot]

@ahejlsberg Here are the results of running the user test suite comparing main and refs/pull/53893/merge:

There were infrastructure failures potentially unrelated to your change:

• 1 instance of "Package install failed"

Otherwise...

Everything looks good!

[typescript-bot]

Heya @ahejlsberg, I've run the RWC suite on this PR - assuming you're on the TS core team, you can view the resulting diff here.

[typescript-bot]

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

Here they are:

Comparison Report - main..53893

Metric	main	53893	Delta	Best	Worst	p-value
Angular - node (v16.17.1, x64)
Memory used	364,628k (± 0.00%)	364,646k (± 0.01%)	~	364,591k	364,703k	p=0.378 n=6
Parse Time	3.52s (± 0.43%)	3.52s (± 0.38%)	~	3.50s	3.54s	p=0.931 n=6
Bind Time	1.18s (± 0.44%)	1.19s (± 0.83%)	~	1.17s	1.20s	p=0.191 n=6
Check Time	9.50s (± 0.34%)	9.50s (± 0.39%)	~	9.45s	9.55s	p=0.808 n=6
Emit Time	7.92s (± 0.39%)	7.93s (± 0.42%)	~	7.89s	7.98s	p=0.567 n=6
Total Time	22.12s (± 0.12%)	22.14s (± 0.31%)	~	22.08s	22.23s	p=0.936 n=6
Compiler-Unions - node (v16.17.1, x64)
Memory used	192,609k (± 0.04%)	193,173k (± 0.69%)	~	192,537k	195,891k	p=0.298 n=6
Parse Time	1.60s (± 0.56%)	1.58s (± 1.46%)	~	1.54s	1.60s	p=0.114 n=6
Bind Time	0.82s (± 0.63%)	0.82s (± 0.92%)	~	0.81s	0.83s	p=0.784 n=6
Check Time	10.24s (± 0.72%)	10.19s (± 0.75%)	~	10.10s	10.33s	p=0.125 n=6
Emit Time	2.99s (± 0.58%)	2.98s (± 0.51%)	~	2.96s	3.00s	p=0.192 n=6
Total Time	15.66s (± 0.58%)	15.57s (± 0.59%)	~	15.45s	15.73s	p=0.077 n=6
Monaco - node (v16.17.1, x64)
Memory used	345,776k (± 0.01%)	345,721k (± 0.01%)	~	345,691k	345,772k	p=0.128 n=6
Parse Time	2.73s (± 0.92%)	2.72s (± 0.44%)	~	2.71s	2.74s	p=0.803 n=6
Bind Time	1.09s (± 0.77%)	1.08s (± 0.48%)	~	1.08s	1.09s	p=0.533 n=6
Check Time	7.84s (± 0.36%)	7.81s (± 0.41%)	~	7.76s	7.85s	p=0.210 n=6
Emit Time	4.44s (± 1.22%)	4.43s (± 0.36%)	~	4.41s	4.45s	p=1.000 n=6
Total Time	16.08s (± 0.48%)	16.05s (± 0.27%)	~	15.99s	16.10s	p=0.520 n=6
TFS - node (v16.17.1, x64)
Memory used	300,053k (± 0.00%)	299,988k (± 0.01%)	-64k (- 0.02%)	299,965k	300,007k	p=0.005 n=6
Parse Time	2.16s (± 0.70%)	2.16s (± 0.48%)	~	2.14s	2.17s	p=1.000 n=6
Bind Time	1.24s (± 0.61%)	1.24s (± 0.79%)	~	1.23s	1.25s	p=0.554 n=6
Check Time	7.21s (± 0.44%)	7.15s (± 0.39%)	-0.06s (- 0.88%)	7.11s	7.18s	p=0.006 n=6
Emit Time	4.38s (± 0.58%)	4.35s (± 0.86%)	~	4.30s	4.41s	p=0.227 n=6
Total Time	14.98s (± 0.28%)	14.90s (± 0.45%)	-0.08s (- 0.55%)	14.79s	14.99s	p=0.044 n=6
material-ui - node (v16.17.1, x64)
Memory used	481,137k (± 0.01%)	481,149k (± 0.02%)	~	481,027k	481,290k	p=0.936 n=6
Parse Time	3.26s (± 0.27%)	3.26s (± 0.46%)	~	3.24s	3.28s	p=0.676 n=6
Bind Time	0.95s (± 0.67%)	0.95s (± 0.54%)	~	0.95s	0.96s	p=0.386 n=6
Check Time	17.83s (± 0.80%)	17.67s (± 0.28%)	-0.16s (- 0.93%)	17.63s	17.76s	p=0.024 n=6
Emit Time	0.00s (± 0.00%)	0.00s (± 0.00%)	~	0.00s	0.00s	p=1.000 n=6
Total Time	22.05s (± 0.63%)	21.88s (± 0.30%)	-0.17s (- 0.77%)	21.83s	22.00s	p=0.019 n=6
xstate - node (v16.17.1, x64)
Memory used	559,334k (± 0.01%)	559,322k (± 0.02%)	~	559,208k	559,456k	p=0.689 n=6
Parse Time	3.99s (± 0.16%)	4.00s (± 0.30%)	~	3.98s	4.01s	p=0.512 n=6
Bind Time	1.78s (± 0.42%)	1.77s (± 0.56%)	~	1.76s	1.78s	p=0.109 n=6
Check Time	3.02s (± 0.55%)	3.01s (± 0.79%)	~	2.98s	3.04s	p=0.466 n=6
Emit Time	0.09s (± 4.45%)	0.09s (± 4.45%)	~	0.09s	0.10s	p=1.000 n=6
Total Time	8.88s (± 0.21%)	8.87s (± 0.29%)	~	8.84s	8.91s	p=0.683 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	53893	6
Baseline	main	6

Developer Information:

Download Benchmark

[ahejlsberg]

Hmm, doesn't appear there's much to be saved here. I know that the cache catches about 10% of the calls to getFlowTypeOfReference in the compiler-with-unions test. But of course it won't help much if the CFA graphs are shallow, which appears to be the case.

[jakebailey]

Still seems like a win, though! The other benchmarks appear to have gotten better, but the difference is so small that it could technically be attributed to chance (hence, hidden).

[typescript-bot]

@ahejlsberg Here are the results of running the top-repos suite comparing main and refs/pull/53893/merge:

Everything looks good!

[typescript-bot]

Hey @ahejlsberg, the results of running the DT tests are ready.
Everything looks the same!
You can check the log here.