Speed Optimization for short living applications

[dadhi]

Speed Optimization for short living applications

[dadhi]

Here is the proof: benchmark, sut and the related epic #44

DryIoc v3:

CreateContainerAndRegisterServices_Then_FirstTimeOpenScopeAndResolve

                            Method |        Mean |      Error |     StdDev |  Ratio | RatioSD | Gen 0/1k Op | Gen 1/1k Op | Gen 2/1k Op | Allocated Memory/Op |
---------------------------------- |------------:|-----------:|-----------:|-------:|--------:|------------:|------------:|------------:|--------------------:|
 BmarkMicrosoftDependencyInjection |    128.3 us |   1.591 us |   1.489 us |   1.00 |    0.00 |     13.7939 |           - |           - |            58.66 KB |
                      BmarkAutofac |    425.8 us |   2.081 us |   1.947 us |   3.32 |    0.05 |     76.6602 |      2.4414 |           - |           354.91 KB |
                  BmarkAutofacMsDi |    434.9 us |   1.827 us |   1.709 us |   3.39 |    0.03 |     80.0781 |      0.4883 |           - |           371.22 KB |
                        BmarkGrace | 17,839.6 us |  99.023 us |  87.781 us | 138.99 |    1.72 |    156.2500 |     62.5000 |           - |           781.49 KB |
                    BmarkGraceMsDi | 20,568.8 us | 116.222 us | 108.714 us | 160.39 |    2.00 |    187.5000 |     93.7500 |     31.2500 |           954.13 KB |
                       BmarkDryIoc | 47,340.9 us | 175.076 us | 146.196 us | 368.47 |    3.79 |     90.9091 |           - |           - |            759.4 KB |
                   BmarkDryIocMsDi | 46,669.8 us | 302.005 us | 282.496 us | 363.92 |    4.44 |    181.8182 |     90.9091 |           - |           855.37 KB |

OpenScopeAndResolve

                            Method |      Mean |     Error |    StdDev | Ratio | RatioSD | Gen 0/1k Op | Gen 1/1k Op | Gen 2/1k Op | Allocated Memory/Op |
---------------------------------- |----------:|----------:|----------:|------:|--------:|------------:|------------:|------------:|--------------------:|
 BmarkMicrosoftDependencyInjection |  3.407 us | 0.0131 us | 0.0116 us |  1.00 |    0.00 |      0.8354 |           - |           - |             3.87 KB |
                        BmarkGrace |  3.954 us | 0.0102 us | 0.0095 us |  1.16 |    0.01 |      1.8921 |           - |           - |             8.73 KB |
                    BmarkGraceMsDi |  5.253 us | 0.0126 us | 0.0111 us |  1.54 |    0.01 |      2.2736 |           - |           - |            10.49 KB |
                       BmarkDryIoc | 25.944 us | 0.0288 us | 0.0270 us |  7.61 |    0.03 |      3.8757 |           - |           - |            17.88 KB |
                   BmarkDryIocMsDi | 29.552 us | 0.0841 us | 0.0746 us |  8.67 |    0.04 |      4.7302 |           - |           - |            21.94 KB |
                      BmarkAutofac | 40.997 us | 0.1914 us | 0.1790 us | 12.04 |    0.08 |     10.8643 |           - |           - |            50.23 KB |
                  BmarkAutofacMsDi | 51.114 us | 0.2980 us | 0.2788 us | 15.00 |    0.08 |     14.1602 |           - |           - |            65.39 KB |

DryIoc v4 (preview-02)

CreateContainerAndRegisterServices_Then_FirstTimeOpenScopeAndResolve

                            Method |        Mean |       Error |      StdDev |  Ratio | RatioSD | Gen 0/1k Op | Gen 1/1k Op | Gen 2/1k Op | Allocated Memory/Op |
---------------------------------- |------------:|------------:|------------:|-------:|--------:|------------:|------------:|------------:|--------------------:|
 BmarkMicrosoftDependencyInjection |    134.0 us |   2.1675 us |   2.0275 us |   1.00 |    0.00 |     13.7939 |      0.1221 |           - |            58.65 KB |
                       BmarkDryIoc |    145.3 us |   0.8318 us |   0.7374 us |   1.08 |    0.02 |     30.2734 |           - |           - |           140.35 KB |
                   BmarkDryIocMsDi |    161.6 us |   0.9626 us |   0.9004 us |   1.21 |    0.02 |     32.2266 |           - |           - |           149.19 KB |
                  BmarkAutofacMsDi |    665.2 us |   5.4804 us |   5.1264 us |   4.96 |    0.08 |    105.4688 |      3.9063 |           - |           487.64 KB |
                      BmarkAutofac |    673.2 us |   5.5615 us |   5.2022 us |   5.02 |    0.08 |    101.5625 |     18.5547 |           - |           470.39 KB |
                        BmarkGrace | 18,480.8 us | 100.5977 us |  89.1773 us | 137.83 |    2.35 |    156.2500 |     62.5000 |           - |           755.18 KB |
                    BmarkGraceMsDi | 21,640.4 us | 119.7309 us | 106.1383 us | 161.39 |    2.81 |    187.5000 |     93.7500 |     31.2500 |           926.88 KB |

OpenScopeAndResolve

                            Method |      Mean |     Error |    StdDev | Ratio | RatioSD | Gen 0/1k Op | Gen 1/1k Op | Gen 2/1k Op | Allocated Memory/Op |
---------------------------------- |----------:|----------:|----------:|------:|--------:|------------:|------------:|------------:|--------------------:|
 BmarkMicrosoftDependencyInjection |  3.222 us | 0.0138 us | 0.0129 us |  1.00 |    0.00 |      0.8354 |           - |           - |             3.87 KB |
                       BmarkDryIoc |  4.274 us | 0.0276 us | 0.0259 us |  1.33 |    0.01 |      1.9531 |           - |           - |             9.02 KB |
                   BmarkDryIocMsDi |  4.498 us | 0.0237 us | 0.0222 us |  1.40 |    0.01 |      1.9608 |           - |           - |             9.06 KB |
                        BmarkGrace |  4.604 us | 0.0271 us | 0.0254 us |  1.43 |    0.01 |      2.3499 |           - |           - |            10.85 KB |
                    BmarkGraceMsDi |  5.280 us | 0.0267 us | 0.0236 us |  1.64 |    0.01 |      2.2202 |           - |           - |            10.24 KB |
                      BmarkAutofac | 37.600 us | 0.5655 us | 0.5289 us | 11.67 |    0.16 |      9.3994 |      0.0610 |           - |            43.47 KB |
                  BmarkAutofacMsDi | 49.487 us | 0.4901 us | 0.4585 us | 15.36 |    0.13 |     13.3667 |      0.1221 |           - |            61.75 KB |

[EamonNerbonne]

It's conceivable that on .net core 2.1 you might get a similar boost by enabling tiered compilation.

It's certainly worth trying out; if that really works well, it might render the interpretation of expression trees themselves moot (because the JIT would be interpreting IL, and the conversion to IL is fairly fast with FastExpressionCompiler).

[Fruchuxs]

COMPlus_TieredCompilation

true	false
1,2543458	1,1927924
1,1643181	1,7478098
1,1927924	1,3724727
1,1572949	1,413623
avg	avg
1,1921878	1,431674475

Probes in seconds.
In my case it seems mostly faster, thanks for the info.

[Fruchuxs]

Yes, this is a general problem on my working machine (it's a laptop), as suggested here. I assume it's a CPU boost / throttling Problem. But my measurements gives an idea that COMPlus_TieredCompilation=true is faster at cold start.

[dadhi]

@Fruchuxs Hi, thanks for bringing it.

First, what platform did you target?

Next, there maybe a more work involved, considering that on some platforms DryIoc uses custom expression compiler.

But everything is possible. In FEC we probably immediately fallback to Expression.Compile when preferInterpretation: true is passed.

[Fruchuxs]

We also tested the FEC, but the speed up was in our cases negligible.
Thanks for your acceptance.

[dadhi]

@Fruchuxs, what platform did you test?

[Fruchuxs]

Oh sorry, forget to advise: .net core 2.1

[dadhi]

Hi @EamonNerbonne , are you suggesting to add <TieredCompilation>true</TieredCompilation> and check?

[EamonNerbonne]

Yeah, that would be a start ;-).

But I have no experience with that (new) feature, so can't tell you about any gotchas - I mean I hope there aren't any, but it's conceivable it doesn't work with dynamically generated IL or some such thing. Only one way to find out! ;-).

[Fruchuxs]

Will check but their measurements doesn't look promising.

[EamonNerbonne]

Just a little worried about the high variance in those numbers - it might just be chance.

[dadhi]

@Fruchuxs ,

Cool actually. I was sceptical, but it is a visible difference. Moreover, it affects dynamic methods as well.

Thanks for trying!

[dadhi]

#45 should help.

[Fruchuxs]

Looks nice, now we need this Version out. :P
Any release plans?

EDIT: Btw. really good work!

[enif-lee]

Wow... It's coooool