Hello, I am Laszlo

Performance of Value Tuples in C# - 2

06/16/2018

In the previous post I have shown how value tuples differ from having multiple out parameters. The IL code has been shown and some micro-benchmarking have been done.In this post I would like re-visit one very interesting question around the performance and the generated IL Code. When we know for sure, that one of the parameters of a method is not needed, we can use a (int success, _) syntax to tell the C# compiler that we are only interested in the first parameter of the returned value tuple.

Here is the IL when we have both parameters used:

The previous post has shown that the average of deconstructing both parameters takes 18959 ms in that given test environment.

Find out more

Performance of Value Tuples in C# - 1

06/03/2018

C# 7 has introduced lots of new features. One of them is the new ValueTuples

Instead of discussing the compiler syntax and usage of value tuples, I suggest to search on the web, here is one great post C# 7 Series, Part 1: Value Tuples about it.

In this post I will do a small micro-benchmarking and also look at IL Code got generated for multiple return types.I will use an example where I have a possibility to return two values, but previously we have used one result parameter and out parameter: a TryParse method. It is made up sample code though, probably you would never have this business logic in a real application.

The Out parameter

Find out more

The Is, As and the Is-As - WinDBG

05/11/2018

In the previous post, I have investigated the performance implications of using the is and as keywords, including the new pattern matching syntax for the is keyword. Let me reference a previous work in this topic from Sasha Goldshtein, which motivated to do the investigation on the new pattern matching syntax of the is keyword: Micro-Benchmarking Done Wrong, And For The Wrong Reasons

In the previous post, I have measured performance with micro-benchmarking and took a look at the code, through ILSpy.

In this post I will take a look at the same code (invoking a non-inlined Work() method) but this time, I will use WinDBG and compare the native code generated by the JIT compiler.

JIT compiler's output may change with version updates of the framework, as this is an implementation detail to the CLR. The generated code also depends on the architecture of the system where we run our code. My test machine is x64 and uses RyuJIT compiler.

Find out more

The Is, As and the Is-As

04/25/2018

C# has recently introduced with C# 7 a couple of new syntax sugars. I this post, I will investigate the new is keyword from IL standpoint and from performance standpoint. I will investigate by analyzing the generated IL Code and by micro benchmarking the instructions.

In my tests I repeat and measure the time it takes to run these instructions in a loop of 1000000000. I elimininate the best and worst case, and calculate an average on the rest.

The book Pro .Net Performance has a chapter on micro benchmarking the is and as keywords. It starts with a false assumption that the 'as' keyword is faster compared to the 'is', while the real conclusion of the chapter would be to consider all details of micro benchmarking to avoid making false assumption.I will have a simple 'My' class created for the tests. Note, the NoInlining attribute on the Work method to avoid any optimization when invoking the method.

public class My
{
  public int a = 0;
  
  [MethodImpl(MethodImplOptions.NoInlining)]
  public void Work() {}
}

Find out more

Cost of running a chatbot

04/14/2018

I have been using bot framework to run an unpublished chatbot. My chatbot is unpublished, as it is only used by a handful of people for testing purposes. As of end of last year, bot framework went live, and depricated the previous way to store the bot's state. we need to provide a way to manage data state, we can implement IBotDataStore, or just simply use one of the provided solutions. I went by using the latter option, using the CosmosDB. The reason was that I wanted to take a closer look at Cosmos DB, which has great features. After a couple of months I took a look the detailed costs in my bill, and I was striked me by the fact, that CosmosDB is expensive for a chatbot being a test phase.

A pre-tax of nearly 50 Euros seems quite high for an unpublished chatbot having a couple of requests a day.

I changed to table store instead, it meant to give another implementation to IBotDataStore:

Find out more