Add benchmarks for Gen.sample

[ghost]

This PR puts the examples from #135 into a proper benchmark, in support of #371.

/cc @moodmosaic @TysonMN

[ghost]

BenchmarkDotNet=v0.13.1, OS=debian 10 (container)
Intel Core i7-9750H CPU 2.60GHz, 8 CPU, 8 logical and 8 physical cores
.NET Core SDK=3.1.413
  [Host]        : .NET Core 3.1.19 (CoreCLR 4.700.21.41101, CoreFX 4.700.21.41603), X64 RyuJIT DEBUG
  .NET Core 3.1 : .NET Core 3.1.19 (CoreCLR 4.700.21.41101, CoreFX 4.700.21.41603), X64 RyuJIT

Job=.NET Core 3.1  Runtime=.NET Core 3.1  

Method	Mean	Error	StdDev
GenSampleLargeHedgehog	130.46 ms	1.178 ms	0.984 ms
GenSampleLargeFsCheck	41.73 ms	0.216 ms	0.191 ms

[ghost]

Also, as a fun side note I removed bigint again (again, naively just replacing bigint with int64) and got the following:

BenchmarkDotNet=v0.13.1, OS=debian 10 (container)
Intel Core i7-9750H CPU 2.60GHz, 8 CPU, 8 logical and 8 physical cores
.NET Core SDK=3.1.413
  [Host]        : .NET Core 3.1.19 (CoreCLR 4.700.21.41101, CoreFX 4.700.21.41603), X64 RyuJIT DEBUG
  .NET Core 3.1 : .NET Core 3.1.19 (CoreCLR 4.700.21.41101, CoreFX 4.700.21.41603), X64 RyuJIT

Job=.NET Core 3.1  Runtime=.NET Core 3.1  

Method	Mean	Error	StdDev
GenSampleLargeHedgehog	52.42 ms	0.695 ms	0.616 ms
GenSampleLargeFsCheck	42.26 ms	0.352 ms	0.330 ms

[moodmosaic]

I need to refresh my memory on the choice for bigint, perhaps because of the stuff in Numeric.fs 🤔

[TysonMN]

Add Params like in this file. I am thinking powers of 10 would be good choices.

And this slightly contradicts my other comment that said to put the quantity to be sampled in the method name. With this change, don't put the quantity in the name.

[ghost]

I need to refresh my memory on the choice for bigint, perhaps because of the stuff in Numeric.fs 🤔

I think it was out of concern for generating min/max ranges for uint64/int64? I remember seeing a comment from you on some issue or PR about it. Either way, I think we can safely replace it 🤷🏼‍♂️

[ghost]

Add Params like in this file. I am thinking powers of 10 would be good choices.

And this slightly contradicts my other comment that said to put the quantity to be sampled in the method name. With this change, don't put the quantity in the name.

BenchmarkDotNet=v0.13.1, OS=debian 10 (container)
Intel Core i7-9750H CPU 2.60GHz, 8 CPU, 8 logical and 8 physical cores
.NET Core SDK=3.1.413
  [Host]        : .NET Core 3.1.19 (CoreCLR 4.700.21.41101, CoreFX 4.700.21.41603), X64 RyuJIT DEBUG
  .NET Core 3.1 : .NET Core 3.1.19 (CoreCLR 4.700.21.41101, CoreFX 4.700.21.41603), X64 RyuJIT

Job=.NET Core 3.1  Runtime=.NET Core 3.1

Method	N	Mean	Error	StdDev	Median
HedgehogGenSampleInt32	100	139.42 μs	1.612 μs	1.508 μs	139.31 μs
FsCheckGenSampleInt32	100	23.39 μs	0.065 μs	0.054 μs	23.38 μs
HedgehogGenSampleInt32	1000	1,313.18 μs	17.493 μs	16.363 μs	1,304.07 μs
FsCheckGenSampleInt32	1000	169.81 μs	3.191 μs	4.260 μs	170.33 μs
HedgehogGenSampleInt32	10000	13,045.48 μs	185.140 μs	173.180 μs	13,081.98 μs
FsCheckGenSampleInt32	10000	1,878.52 μs	31.893 μs	29.833 μs	1,886.14 μs
HedgehogGenSampleInt32	100000	130,364.13 μs	1,713.186 μs	1,602.515 μs	130,159.68 μs
FsCheckGenSampleInt32	100000	27,397.14 μs	542.102 μs	1,309.237 μs	26,793.29 μs

For reference, here's the same benchmarks without bigint:

BenchmarkDotNet=v0.13.1, OS=debian 10 (container)
Intel Core i7-9750H CPU 2.60GHz, 8 CPU, 8 logical and 8 physical cores
.NET Core SDK=3.1.413
  [Host]        : .NET Core 3.1.19 (CoreCLR 4.700.21.41101, CoreFX 4.700.21.41603), X64 RyuJIT DEBUG
  .NET Core 3.1 : .NET Core 3.1.19 (CoreCLR 4.700.21.41101, CoreFX 4.700.21.41603), X64 RyuJIT

Job=.NET Core 3.1  Runtime=.NET Core 3.1  

Method	N	Mean	Error	StdDev
HedgehogGenSampleInt32	100	58.38 μs	1.010 μs	1.349 μs
FsCheckGenSampleInt32	100	23.51 μs	0.071 μs	0.066 μs
HedgehogGenSampleInt32	1000	500.94 μs	2.716 μs	2.407 μs
FsCheckGenSampleInt32	1000	167.69 μs	2.665 μs	2.362 μs
HedgehogGenSampleInt32	10000	5,187.75 μs	27.047 μs	22.586 μs
FsCheckGenSampleInt32	10000	1,864.81 μs	35.988 μs	36.957 μs
HedgehogGenSampleInt32	100000	50,216.08 μs	913.421 μs	1,551.057 μs
FsCheckGenSampleInt32	100000	26,470.79 μs	357.376 μs	334.289 μs

There is the data graphed, green is Hedgehog+BigInt, Blue is Hedgehog+Int64, Black is FsCheck.

[TysonMN]

This looks great. Excellent work @adam-becker!

[TysonMN]

I think this PR is ready to merge

[TysonMN]

@adam-becker, do you think this PR is ready to merge?

[dharmaturtle]

I don't have any experience with BenchmarkDotNet - is it possible to use it to write a test that will fail if, say, HedgehogGenSampleInt32 N=100 takes 1 second? Not a blocker to merging - I'm just wondering.

[TysonMN]

I don't think so. The point of a BenchmarkDotNet test is to look at the report it generates.