Improved performance benchmark

[rafaelsc]

• Add many new benchmarks
  • with many calls, with many classes
  • Singleton, Transient, Scoped, Mixed, Complex with simple and deep dependences
• Add all benchmark results to 'docs/benchmark'
• Update Readme benchmark results.

Some new benchmarks show that MEDI is still faster than jab in some cases.

[pakrym]

Hi @rafaelsc, thank you for your work. I'm a bit swamped by work so it might take me a bit to get through these benchmarks.

[pakrym]

This is quite interesting. Do you have an idea why complex resolution is so much slower in Jab?

[rafaelsc]

I tried to find it and didn't understand why yet.
Some could be explained as the lock {}, but I removed it and didn't see a big improvement. See #169

[rafaelsc]

I managed to beat the MEDI performance for this case.

• Removing locks and using LazyInitializer.EnsureInitialized
• Removing the TryAddDisposable() - (this could be done in compilation time instead of runtime)
• Replace all this.GetService<T>() for ((IServiceProvider<T>)this).GetService()

Method	NumbersOfCalls	NumbersOfClasses	Mean	Error	StdDev	Ratio	RatioSD	Gen0	Gen1	Allocated	Alloc Ratio
Jab	1	1	270.81 ns	63.19 ns	3.463 ns	1.00	0.00	0.0563	-	472 B	1.00
Improved_Jab	1	1	92.74 ns	65.84 ns	3.609 ns	0.34	0.01	0.0736	-	616 B	1.31
MEDI	1	1	157.47 ns	37.13 ns	2.035 ns	0.58	0.01	0.0870	0.0002	728 B	1.54
Jab	1	2	543.88 ns	229.26 ns	12.566 ns	1.00	0.00	0.1059	-	888 B	1.00
Improved_Jab	1	2	255.73 ns	213.70 ns	11.714 ns	0.47	0.03	0.1402	-	1176 B	1.32
MEDI	1	2	276.63 ns	129.49 ns	7.098 ns	0.51	0.02	0.1364	0.0005	1144 B	1.29
Jab	1	3	814.39 ns	65.88 ns	3.611 ns	1.00	0.00	0.1554	-	1304 B	1.00
Improved_Jab	1	3	389.66 ns	204.35 ns	11.201 ns	0.48	0.01	0.2074	0.0010	1736 B	1.33
MEDI	1	3	397.50 ns	134.14 ns	7.353 ns	0.49	0.01	0.1864	0.0010	1560 B	1.20
Jab	10	1	2,811.38 ns	127.55 ns	6.992 ns	1.00	0.00	0.5608	-	4720 B	1.00
Improved_Jab	10	1	932.41 ns	132.33 ns	7.253 ns	0.33	0.00	0.7362	-	6160 B	1.31
MEDI	10	1	1,517.62 ns	697.80 ns	38.249 ns	0.54	0.01	0.8698	0.0019	7280 B	1.54
Jab	10	2	5,394.83 ns	1,473.69 ns	80.778 ns	1.00	0.00	1.0605	-	8880 B	1.00
Improved_Jab	10	2	2,563.65 ns	542.04 ns	29.711 ns	0.48	0.01	1.4038	-	11760 B	1.32
MEDI	10	2	2,820.40 ns	1,988.28 ns	108.984 ns	0.52	0.03	1.3657	0.0038	11440 B	1.29
Jab	10	3	8,515.73 ns	1,780.40 ns	97.590 ns	1.00	0.00	1.5564	-	13040 B	1.00
Improved_Jab	10	3	3,791.29 ns	348.95 ns	19.127 ns	0.45	0.01	2.0752	0.0076	17360 B	1.33
MEDI	10	3	4,014.64 ns	1,245.71 ns	68.282 ns	0.47	0.00	1.8616	0.0076	15600 B	1.20
Jab	100	1	28,236.83 ns	12,613.72 ns	691.401 ns	1.00	0.00	5.6152	-	47200 B	1.00
Improved_Jab	100	1	10,499.51 ns	15,496.89 ns	849.437 ns	0.37	0.02	7.3547	-	61600 B	1.31
MEDI	100	1	16,673.16 ns	3,621.32 ns	198.497 ns	0.59	0.02	8.6975	0.0153	72800 B	1.54
Jab	100	2	56,434.78 ns	4,705.42 ns	257.920 ns	1.00	0.00	10.5591	-	88800 B	1.00
Improved_Jab	100	2	25,188.64 ns	6,705.33 ns	367.542 ns	0.45	0.00	14.0381	0.0305	117600 B	1.32
MEDI	100	2	27,463.55 ns	15,492.67 ns	849.206 ns	0.49	0.02	13.6719	0.0305	114400 B	1.29
Jab	100	3	84,041.73 ns	35,261.40 ns	1,932.796 ns	1.00	0.00	15.5029	-	130400 B	1.00
Improved_Jab	100	3	38,046.35 ns	11,730.31 ns	642.978 ns	0.45	0.02	20.7520	0.0610	173600 B	1.33
MEDI	100	3	42,855.80 ns	46,857.75 ns	2,568.432 ns	0.51	0.04	18.6157	0.0610	156000 B	1.20

[pakrym]

Quite surprised Jab allocates so much more..