Alexey Korovin @Temp1ar

@JetBrains developer Joined September 2008

Tweets

167
Followers

94
Following

109
Likes

132

Alexey Korovin @Temp1ar

7 months ago

@quanta_kt @RogerAlsing @JetBrainsRider @intellijidea Mac, I don't know if it's the same shortcut there

0 0 0 19 0

View Details

To celebrate #20YearsOfReSharper, I need to highlight my greatest achievement at JetBrains. 10 years ago, I wrote a blog post about Clippy for ReSharper. The best bit was that it was a real plugin, and you lucky people, it's back for R# 2024.1 and above! blog.jetbrains.com/dotnet/2014/04…

2 17 42 6K 1

View Details

Alexey Korovin @Temp1ar

3 years ago

@codeopinion Check out dotTrace's Line-By-Line mode blog.jetbrains.com/dotnet/2013/01… But keep in mind that you should'nt be analyzing time in such snapshots, because the overhead of line-by-line is massive.

0 0 1 125 0

View Details

Alexey Korovin @Temp1ar

3 years ago

@hypeartistmusic This is inlined target method body, specifically ` mov r14d, edi sub r14d, [rcx+8] ` Stuff before is just making sure that the object we are dealing with in this iteration is of type Sub1, if it's not - execution will go to the slow path calling the virtual method

1 0 2 26 0

View Details

Alexey Korovin @Temp1ar

3 years ago

@hypeartistmusic I haven't seen much of a de-virtualization disassemblies, how it is typically looks like?

1 0 0 20 0

View Details

Alexey Korovin @Temp1ar

3 years ago

@hypeartistmusic Yes, in all your benchmark the 86B version is executed according to the results table and it is faster than 48/52B version because it has an virtual method body instructions inlined directly into the loop (you can found `sub` instruction in the disasm of 86B code)

1 0 1 52 0

View Details

Alexey Korovin @Temp1ar

3 years ago

@hypeartistmusic Looks like your JIT was smart enough to inline virtual call in 86B version, that’s why it is so fast

1 0 0 17 0

View Details

Alexey Korovin @Temp1ar

3 years ago

@hypeartistmusic And also a big difference in code size, do you have a disassembly for 86B version? Whats the JIT version & OS?

1 0 0 24 0

View Details

Alexey Korovin @Temp1ar

3 years ago

@hypeartistmusic

1 0 1 21 0

View Details

Alexey Korovin @Temp1ar

3 years ago

@hypeartistmusic Something is off with your numbers. It says in the table that the code size is 86B, but in the disassembly it is 52B for M1. I also ran the benchmark on my PC and M2 is 20% slower, which is still surprising given the fact that the generated code is almost the same

1 0 0 28 0

View Details

Alexey Korovin @Temp1ar

3 years ago

@Pie_Daddy @Sharundaar Check out postfix completion jetbrains.com/help/resharper…

0 0 1 8 0

View Details

Alexey Korovin @Temp1ar

4 years ago

@TinkeringDaemon @VisualStudio @dotTrace Closest thing @dotTrace can do is to collect native allocations on Windows and show "unreleased" ones in Timeline profiling mode. This can be helpful to find memory leaks. jetbrains.com/help/profiler/…

0 0 0 33 0

View Details

Alexey Korovin @Temp1ar

4 years ago

@chnasarre @xoofx @LadislavBohm @SuperluminalSft @dottrace dotTrace Timeline mode have the thread diagram(no call frames on it indeed, only aggregated flame graph) and recently we’ve added native frames, kernel included. High frequency sampling we don’t have. 1kHz locked right now

1 0 4 0 0

View Details

Alexey Korovin @Temp1ar

4 years ago

@dominicfrei @hirschbenjamin @dotmemory @dottrace You linked documentation about console tools. They are self sufficient tools that don't require Rider installed. You can view the result from those tools in Rider though :)

2 0 1 0 0

View Details

Alexey Korovin @Temp1ar

4 years ago

@dominicfrei @hirschbenjamin @dotmemory @dottrace Not exactly. You can take and analyze dotTrace snapshots in Rider on all platforms. The features are limited, but maybe enough for your use case. jetbrains.com/help/rider/Ana…

1 0 2 0 0

View Details

Alexey Korovin @Temp1ar

5 years ago

@PeteBeech1 @dottrace It will "catch" every new .NET process that will start after you will start profiling. So the thread names should be there.

0 0 0 0 0

View Details

Alexey Korovin @Temp1ar

5 years ago

@PeteBeech1 @dottrace Hello, Pete. Yes, it is possible. You need to create .NET process profiling configuration as it described here jetbrains.com/help/profiler/… Also you may want to configure process filters to leave only necessary processes.