Sasha Goldshtein, the guru of .NET Performance, will speak at the .NET conference in St. Petersburg
In June, two world-class .NET programming stars, Dino Esposito and Sasha Goldshtein, will come to our DotNext conference.
Both of our supergosts are famous for two things:
I have already written about Dino a couple of times, and I will probably write this time too. And today I will tell you about Sasha.
Sasha Goldstein is the leading global performance expert on the .NET platform, eight-time Microsoft MVP, the author of the excellent (and, probably, the only) serious .NET performance book “Pro .NET Performance” ( eng , rus ).
')
Sasha is our great success: he came to Moscow in December at the Moscow DotNext at the invitation of Andrey Akinshin aka DreamWalker , our friend, also .NET MVP, a member of the DotNext program committee. Sasha delighted the audience, despite the fact that he spoke in English. Sasha lives in Israel, speaks Russian perfectly, but, by his own admission, he speaks better on technical topics in English. And in his case - this is not a show off or show off. As I understood from a conversation with him, he simply never spoke on technical topics in Russian before :)
I am very impressed with his reports. As a person who is upsetting from studying runtimes, I adore his report on how .NET compilers work with vectorization ( System.Numerics.Vectors ) using SIMD instructions:
A month later in St. Petersburg on DotNext Sasha will make two reports.
PerfView is a free compact tool from the CLR team that helps in researching the performance of .NET applications. It offers several unique features for in-depth study of voluminous Performance-data. This tool can even be used to analyze performance problems on ARM devices.
In this report, we will use PerfView to periodically measure and improve CPU time, wall-clock time and memory usage (yes, PerfView can help you find memory leaks!).
The phrase “Memory Model” usually instills developers with horror, drives them to sleep or even puts them in a coma. A deep understanding of intricate links between the compiler, the processor and the memory subsystem is necessary if you are going to write low-level synchronization, code without locks, or debugging, trying to find a race in your multi-threaded code. In this report, we will go down into the abyss of CLR and C ++ memory models, affectionately called SC-DRF (see Shipilev , starting on slide 32). We'll talk about volatile and atomic variables, look at the examples of the re-eringing that the compiler and the processor do, and see how multi-threaded code, which at first (and even second) look seems correct, really falls apart on non-Intel-ov processors.
In reality, this can give you a bit of trouble if, for example, you write and test the code on Intel, and in production you live in a cloud and at some point decide to move to ARM, PowerPC or something else. In our time, this is an increasingly frequent story. Well, or all is more prosy: you have a multithreaded code for Windows RT or the upcoming Win 10 for ARM. Then you should definitely figure it out.
This is how hardcore it is. Well, the full list of conference reports can be found here .
PS: Misha Shcherbakov and I, aka yu5k3 , the leader of the SPb .NET Community , are planning to call Sasha to speak at the mit the day before DotNext or a day later. So not only follow the DotNext news, but also the news community too.
Both of our supergosts are famous for two things:
- they are authors of beautiful books;
- they are great speakers.
I have already written about Dino a couple of times, and I will probably write this time too. And today I will tell you about Sasha.
Sasha Goldstein / Sasha Goldshtein
Sasha Goldstein is the leading global performance expert on the .NET platform, eight-time Microsoft MVP, the author of the excellent (and, probably, the only) serious .NET performance book “Pro .NET Performance” ( eng , rus ).
')
Sasha is our great success: he came to Moscow in December at the Moscow DotNext at the invitation of Andrey Akinshin aka DreamWalker , our friend, also .NET MVP, a member of the DotNext program committee. Sasha delighted the audience, despite the fact that he spoke in English. Sasha lives in Israel, speaks Russian perfectly, but, by his own admission, he speaks better on technical topics in English. And in his case - this is not a show off or show off. As I understood from a conversation with him, he simply never spoke on technical topics in Russian before :)
I am very impressed with his reports. As a person who is upsetting from studying runtimes, I adore his report on how .NET compilers work with vectorization ( System.Numerics.Vectors ) using SIMD instructions:
A month later in St. Petersburg on DotNext Sasha will make two reports.
PerfView: Measure and Improve Your App's Performance For Free
PerfView is a free compact tool from the CLR team that helps in researching the performance of .NET applications. It offers several unique features for in-depth study of voluminous Performance-data. This tool can even be used to analyze performance problems on ARM devices.
In this report, we will use PerfView to periodically measure and improve CPU time, wall-clock time and memory usage (yes, PerfView can help you find memory leaks!).
The C ++ and CLR Memory Models
The phrase “Memory Model” usually instills developers with horror, drives them to sleep or even puts them in a coma. A deep understanding of intricate links between the compiler, the processor and the memory subsystem is necessary if you are going to write low-level synchronization, code without locks, or debugging, trying to find a race in your multi-threaded code. In this report, we will go down into the abyss of CLR and C ++ memory models, affectionately called SC-DRF (see Shipilev , starting on slide 32). We'll talk about volatile and atomic variables, look at the examples of the re-eringing that the compiler and the processor do, and see how multi-threaded code, which at first (and even second) look seems correct, really falls apart on non-Intel-ov processors.
In reality, this can give you a bit of trouble if, for example, you write and test the code on Intel, and in production you live in a cloud and at some point decide to move to ARM, PowerPC or something else. In our time, this is an increasingly frequent story. Well, or all is more prosy: you have a multithreaded code for Windows RT or the upcoming Win 10 for ARM. Then you should definitely figure it out.
This is how hardcore it is. Well, the full list of conference reports can be found here .
PS: Misha Shcherbakov and I, aka yu5k3 , the leader of the SPb .NET Community , are planning to call Sasha to speak at the mit the day before DotNext or a day later. So not only follow the DotNext news, but also the news community too.
Source: https://habr.com/ru/post/282890/
All Articles