More tweaks to reduce call stack size

[lahma]

Checking some hot paths and ensuring that the stack doesn't contain unnecessary items. The string-validate-input is pretty insane, though the tests still pass 🤷🏻‍♂️

/cc @KurtGokhan @CreepGin

Esprima.Benchmark.SunSpiderBenchmark

Diff	Method	FileName	Mean	Error	Gen 0	Gen 1	Gen 2	Allocated
Old	Run	3d-cube	217.72 ms	0.453 ms	3000.0000	333.3333	-	51,125 KB
New			206.21 ms (-5%)	0.372 ms	3000.0000 (0%)	333.3333 (0%)	-	51,125 KB (0%)
Old	Run	3d-morph	160.62 ms	0.226 ms	2000.0000	1000.0000	-	49,744 KB
New			158.71 ms (-1%)	0.510 ms	3000.0000 (+50%)	1250.0000 (+25%)	500.0000	49,745 KB (0%)
Old	Run	3d-raytrace	173.30 ms	1.197 ms	5000.0000	1000.0000	-	93,390 KB
New			172.14 ms (-1%)	2.352 ms	5000.0000 (0%)	1000.0000 (0%)	-	93,390 KB (0%)
Old	Run	access-binary-trees	86.68 ms	0.349 ms	4000.0000	1000.0000	-	78,472 KB
New			81.73 ms (-6%)	0.177 ms	4000.0000 (0%)	1000.0000 (0%)	-	78,472 KB (0%)
Old	Run	access-fannkuch	469.09 ms	1.075 ms	-	-	-	136 KB
New			455.82 ms (-3%)	2.175 ms	-	-	-	136 KB (0%)
Old	Run	access-nbody	197.97 ms	0.457 ms	3666.6667	-	-	63,853 KB
New			191.97 ms (-3%)	0.484 ms	3666.6667 (0%)	333.3333	-	63,850 KB (0%)
Old	Run	access-nsieve	157.01 ms	0.295 ms	1000.0000	-	-	21,518 KB
New			164.04 ms (+4%)	0.170 ms	1000.0000 (0%)	-	-	21,518 KB (0%)
Old	Run	bitop(...)-byte [24]	166.13 ms	0.398 ms	4000.0000	-	-	68,950 KB
New			163.70 ms (-1%)	0.509 ms	4000.0000 (0%)	-	-	68,950 KB (0%)
Old	Run	bitops-bits-in-byte	262.53 ms	0.422 ms	2000.0000	-	-	45,446 KB
New			244.74 ms (-7%)	0.389 ms	2000.0000 (0%)	-	-	45,446 KB (0%)
Old	Run	bitops-bitwise-and	178.28 ms	0.559 ms	5000.0000	-	-	93,439 KB
New			168.51 ms (-5%)	1.003 ms	5000.0000 (0%)	-	-	93,439 KB (0%)
Old	Run	bitops-nsieve-bits	212.82 ms	0.730 ms	3000.0000	1000.0000	-	54,856 KB
New			207.57 ms (-2%)	1.581 ms	3000.0000 (0%)	1000.0000 (0%)	-	54,856 KB (0%)
Old	Run	contr(...)rsive [21]	122.64 ms	0.226 ms	7400.0000	1600.0000	-	124,119 KB
New			116.13 ms (-5%)	0.287 ms	7400.0000 (0%)	1600.0000 (0%)	-	124,119 KB (0%)
Old	Run	crypto-aes	138.35 ms	0.317 ms	-	-	-	15,601 KB
New			135.67 ms (-2%)	0.378 ms	-	-	-	15,590 KB (0%)
Old	Run	crypto-md5	108.69 ms	0.273 ms	5000.0000	1000.0000	-	95,414 KB
New			108.42 ms (0%)	0.179 ms	5000.0000 (0%)	1000.0000 (0%)	-	95,414 KB (0%)
Old	Run	crypto-sha1	111.83 ms	0.311 ms	5000.0000	1000.0000	-	82,074 KB
New			107.91 ms (-4%)	0.302 ms	5000.0000 (0%)	1000.0000 (0%)	-	82,074 KB (0%)
Old	Run	date-format-tofte	109.06 ms	0.242 ms	2000.0000	-	-	48,549 KB
New			107.01 ms (-2%)	0.285 ms	2000.0000 (0%)	-	-	48,549 KB (0%)
Old	Run	date-format-xparb	58.78 ms	0.170 ms	1777.7778	666.6667	-	29,414 KB
New			59.77 ms (+2%)	0.073 ms	1777.7778 (0%)	666.6667 (0%)	-	29,414 KB (0%)
Old	Run	math-cordic	349.63 ms	1.410 ms	7000.0000	-	-	123,973 KB
New			337.03 ms (-4%)	1.368 ms	7000.0000 (0%)	-	-	123,973 KB (0%)
Old	Run	math-partial-sums	138.54 ms	0.279 ms	4000.0000	-	-	68,947 KB
New			134.10 ms (-3%)	0.537 ms	4000.0000 (0%)	-	-	68,947 KB (0%)
Old	Run	math-spectral-norm	136.91 ms	0.462 ms	4000.0000	-	-	67,176 KB
New			135.20 ms (-1%)	0.534 ms	4000.0000 (0%)	-	-	67,176 KB (0%)
Old	Run	regexp-dna	108.58 ms	0.399 ms	800.0000	800.0000	800.0000	17,672 KB
New			108.34 ms (0%)	0.383 ms	1000.0000 (+25%)	1000.0000 (+25%)	1000.0000 (+25%)	17,674 KB (0%)
Old	Run	string-base64	104.31 ms	0.330 ms	-	-	-	12,525 KB
New			103.66 ms (-1%)	0.365 ms	-	-	-	12,523 KB (0%)
Old	Run	string-fasta	204.44 ms	0.353 ms	10000.0000	333.3333	-	167,275 KB
New			197.59 ms (-3%)	0.456 ms	10000.0000 (0%)	333.3333 (0%)	-	167,275 KB (0%)
Old	Run	string-tagcloud	75.07 ms	0.790 ms	3142.8571	1285.7143	571.4286	48,439 KB
New			73.81 ms (-2%)	1.079 ms	3142.8571 (0%)	1285.7143 (0%)	571.4286 (0%)	48,438 KB (0%)
Old	Run	string-unpack-code	75.11 ms	0.249 ms	5000.0000	571.4286	-	84,573 KB
New			72.20 ms (-4%)	0.397 ms	5000.0000 (0%)	571.4286 (0%)	-	84,573 KB (0%)
Old	Run	strin(...)input [21]	1,271.89 ms	25.280 ms	696000.0000	692000.0000	691000.0000	6,352,002 KB
New			84.62 ms (-93%)	0.186 ms	2000.0000 (-100%)	1000.0000 (-100%)	-	44,618 KB (-99%)

[sebastienros]

What's up with the strin(...)input [21] benchmark?

[lahma]

What's up with the strin(...)input [21] benchmark?

I'll try to run the numbers at some point. This was a bit unexpected..

[lahma]

What's up with the strin(...)input [21] benchmark?

OK it's a change in behavior:

https://github.com/sebastienros/jint/pull/1277/files#diff-e715e3b319920d94963426777af53d921e568d5a187aa805a835d1ef4e6063dfL59

The function result is no longer eagerly cloned, in this case string builder (ConcatenatedString) isn't being materialized. I need to think if there is a case where it could be a problem, on assignment we usually want to ensure that it's properly materialized,.

[sebastienros]

Another point of view is that since it had such a big impact there, maybe there are other places with the same code that could be optimized the same way, as long as it's also valid.

[lahma]

So far I haven't found a reason this wouldn't generally be safe for string in this case. ConcatenatedString is only one currently using the Clone, I was afraid of ArgumentsInstance, which is pooled and can cause trouble (thanks test suite!). One thing to keep an eye on for sure.