Pedantic refactor

[adamsitnik]

@pgovind while reviewing dotnet#1445 I've come up with some pedantic refactoring ideas

since my code review was delayed by the 5.0 stuff and now I need these benchmarks to sign off 5.0 I decided to send a PR to your fork to speed up the process

PTAL

Sample results:

Method	Toolchain	Input	Mean	Ratio	Allocated
ToChars	netcoreapp3.1	EnglishAllAscii	17,728.09 ns	1.00	-
ToChars	netcoreapp5.0	EnglishAllAscii	17,702.73 ns	1.00	-
IsAscii	netcoreapp3.1	EnglishAllAscii	3,673.88 ns	1.00	-
IsAscii	netcoreapp5.0	EnglishAllAscii	3,430.67 ns	0.93	-
IsNormalized	netcoreapp3.1	EnglishAllAscii	133,466.28 ns	1.00	335120 B
IsNormalized	netcoreapp5.0	EnglishAllAscii	130,560.00 ns	0.98	335120 B
ToCharArray	netcoreapp3.1	EnglishAllAscii	59,725.94 ns	1.00	335116 B
ToCharArray	netcoreapp5.0	EnglishAllAscii	61,426.65 ns	1.02	335117 B
ToChars	netcoreapp3.1	EnglishMostlyAscii	72,505.19 ns	1.00	-
ToChars	netcoreapp5.0	EnglishMostlyAscii	71,528.86 ns	0.99	-
IsAscii	netcoreapp3.1	EnglishMostlyAscii	15.02 ns	1.00	-
IsAscii	netcoreapp5.0	EnglishMostlyAscii	11.92 ns	0.79	-
IsNormalized	netcoreapp3.1	EnglishMostlyAscii	333,324.30 ns	1.00	335128 B
IsNormalized	netcoreapp5.0	EnglishMostlyAscii	413,591.01 ns	1.24	335128 B
ToCharArray	netcoreapp3.1	EnglishMostlyAscii	183,631.69 ns	1.00	335126 B
ToCharArray	netcoreapp5.0	EnglishMostlyAscii	155,377.17 ns	0.85	335126 B
ToChars	netcoreapp3.1	Chinese	132,721.23 ns	1.00	-
ToChars	netcoreapp5.0	Chinese	132,349.89 ns	1.00	-
IsAscii	netcoreapp3.1	Chinese	27.36 ns	1.00	-
IsAscii	netcoreapp5.0	Chinese	24.92 ns	0.91	-
IsNormalized	netcoreapp3.1	Chinese	411,800.56 ns	1.00	155961 B
IsNormalized	netcoreapp5.0	Chinese	433,824.27 ns	1.05	155960 B
ToCharArray	netcoreapp3.1	Chinese	224,659.80 ns	1.00	155960 B
ToCharArray	netcoreapp5.0	Chinese	219,481.70 ns	0.98	155960 B
ToChars	netcoreapp3.1	Cyrillic	92,179.97 ns	1.00	-
ToChars	netcoreapp5.0	Cyrillic	90,463.51 ns	0.98	-
IsAscii	netcoreapp3.1	Cyrillic	30.15 ns	1.00	-
IsAscii	netcoreapp5.0	Cyrillic	34.55 ns	1.15	-
IsNormalized	netcoreapp3.1	Cyrillic	515,270.54 ns	1.00	133640 B
IsNormalized	netcoreapp5.0	Cyrillic	360,300.03 ns	0.70	133640 B
ToCharArray	netcoreapp3.1	Cyrillic	185,079.52 ns	1.00	133640 B
ToCharArray	netcoreapp5.0	Cyrillic	182,600.64 ns	0.99	133640 B
ToChars	netcoreapp3.1	Greek	147,888.70 ns	1.00	-
ToChars	netcoreapp5.0	Greek	146,859.71 ns	0.99	-
IsAscii	netcoreapp3.1	Greek	27.56 ns	1.00	-
IsAscii	netcoreapp5.0	Greek	33.26 ns	1.21	-
IsNormalized	netcoreapp3.1	Greek	727,875.98 ns	1.00	169352 B
IsNormalized	netcoreapp5.0	Greek	533,343.36 ns	0.73	169352 B
ToCharArray	netcoreapp3.1	Greek	291,101.71 ns	1.00	169352 B
ToCharArray	netcoreapp5.0	Greek	284,594.27 ns	0.98	169352 B

[pgovind]

cc @carlossanlop @kunalspathak

[carlossanlop]

@adamsitnik Shouldn't the folder names match the assembly name?

[carlossanlop]

I really don't mind (it helps to type less). Maybe I'm being too pedantic about it 😄 .

[adamsitnik]

@carlossanlop This is a very good question, however, I am not sure if there is a good answer to it.

The namespace, class name, method name, argument name and value are used to build benchmark ID which is used in the reporting system. Something like namespace.className.methoName(argName: argValue). Example: https://pvscmdupload.blob.core.windows.net/reports/allTestHistory/refs/heads/master_x64_Windows%2010.0.18362/Microsoft.Extensions.Primitives.StringSegmentBenchmark.EndsWith.html

So first of all, we need to choose good names because they are used as IDs to identify benchmarks: https://github.com/dotnet/performance/blob/master/docs/microbenchmark-design-guidelines.md#Benchmarks-are-Immutable

Here I hope that in 6.0 we are simply going to switch from Experimental to just not experimental and I would prefer to avoid using Experimental in the name.

The other problem is that this particular namespace is used to copy the output text files to output folder and the longer it is, the higher chance of hitting too long name path exceptions etc. So in general, the shorter names, the better.

[pgovind]

Merged. Thanks Adam!!

[pgovind]

@adamsitnik , these numbers are from an x64 run right (and no special envs)?

[adamsitnik]

these numbers are from an x64 run right (and no special envs)?

yes, these were Windows x64 numbers with default settings