uid | name |
---|---|
docs.diagnosers |
Diagnosers |
A diagnoser can attach to your benchmark and get some useful info.
The current Diagnosers are:
- GC and Memory Allocation (
MemoryDiagnoser
) which is cross platform, built-in and is not enabled by default anymore. Please see Adam Sitnik's blog post for all the details. - JIT Stats Diagnoser.
You can find this diagnoser in a separate package with diagnosers for Windows (
BenchmarkDotNet.Diagnostics.Windows
): - JIT Inlining Events (
InliningDiagnoser
). You can find this diagnoser in a separate package with diagnosers for Windows (BenchmarkDotNet.Diagnostics.Windows
): - JIT Tail Call Events (
TailCallDiagnoser
). You can find this diagnoser as well as the (InliningDiagnoser
) in a separate package with diagnosers for Windows (BenchmarkDotNet.Diagnostics.Windows
): Please see this post for all the details. - Hardware Counter Diagnoser.
You can find this diagnoser in a separate package with diagnosers for Windows (
BenchmarkDotNet.Diagnostics.Windows
): . Please see Adam Sitnik's blog post for all the details. - Disassembly Diagnoser. It allows you to disassemble the benchmarked code to asm, IL and C#/F#. Please see Adam Sitnik's blog post for all the details.
- ETW Profiler (
EtwProfiler
). It allows you to not only benchmark, but also profile the code. It's using TraceEvent, which internally uses ETW and exports all the information to a trace file. The trace file contains all of the stack traces captured by the profiler, PDBs to resolve symbols for both native and managed code and captured GC, JIT and CLR events. Please use one of the free tools: PerfView or Windows Performance Analyzer to analyze and visualize the data from trace file. You can find this diagnoser in a separate package with diagnosers for Windows (BenchmarkDotNet.Diagnostics.Windows
): Please see Adam Sitnik's blog post for all the details. - Concurrency Visualizer Profiler (
ConcurrencyVisualizerProfiler
) It usesEtwProfiler
to profile the code using ETW and create not only.etl
file but also a CVTrace file which can be opened by Concurrency Visualizer plugin from Visual Studio. Please see Adam Sitnik's blog post for all the details. - Native Memory Profiler (
NativeMemoryProfiler
) It usesEtwProfiler
to profile the code using ETW and adds the extra columnsAllocated native memory
andNative memory leak
. Please see Wojciech Nagórski's blog post for all the details. - Event Pipe Profiler (
EventPipeProfiler
). It is a cross-platform profiler that allows profile .NET code on every platform - Windows, Linux, macOS. Please see Wojciech Nagórski's blog post for all the details. - Threading Diagnoser (
ThreadingDiagnoser
) - .NET Core 3.0+ diagnoser that reports some Threading statistics. - Exception Diagnoser (
ExceptionDiagnoser
) - a diagnoser that reports the frequency of exceptions thrown during the operation.
Below is a sample output from the GC and Memory Allocation
diagnoser, note the extra columns on the right-hand side ("Gen 0", "Gen 1", "Gen 2" and "Allocated"):
Method | Mean | StdErr | Median | Gen 0 | Allocated |
----------------- |------------ |----------- |------------ |------- |---------- |
'new byte[10kB]' | 884.4896 ns | 46.3528 ns | 776.4237 ns | 0.1183 | 10 kB |
A config example:
private class Config : ManualConfig
{
public Config()
{
Add(MemoryDiagnoser.Default);
Add(new InliningDiagnoser());
Add(new EtwProfiler());
Add(ThreadingDiagnoser.Default);
Add(ExceptionDiagnoser.Default);
}
}
You can also use one of the following attributes (apply it on a class that contains Benchmarks):
[MemoryDiagnoser]
[InliningDiagnoser]
[TailCallDiagnoser]
[EtwProfiler]
[ConcurrencyVisualizerProfiler]
[NativeMemoryProfiler]
[ThreadingDiagnoser]
[ExceptionDiagnoser]
In BenchmarkDotNet, 1kB = 1024B, 1MB = 1024kB, and so on. The column Gen X means number of GC collections per 1000 operations for that generation.
- In order to not affect main results we perform a separate run if any diagnoser is used. That's why it might take more time to execute benchmarks.
- MemoryDiagnoser:
- In order to get the number of allocated bytes in cross platform way we are using
GC.GetAllocatedBytesForCurrentThread
which recently got exposed for netcoreapp1.1. That's why BenchmarkDotNet does not support netcoreapp1.0 from version 0.10.1. - MemoryDiagnoser is
99.5%
accurate about allocated memory when using default settings or Job.ShortRun (or any longer job than it).
- In order to get the number of allocated bytes in cross platform way we are using
- Threading Diagnoser:
- Works only for .NET Core 3.0+
- HardwareCounters:
- Windows 8+ only (we plan to add Unix support in the future)
- No Hyper-V (Virtualization) support
- Requires running as Admin (ETW Kernel Session)
- No
InProcessToolchain
support (#394)
- EtwProfiler, ConcurrencyVisualizerProfiler and NativeMemoryProfiler:
- Windows only
- Requires running as Admin (ETW Kernel Session)
- No
InProcessToolchain
support (#394)
- Disassembly Diagnoser:
- .NET Core disassembler works only on Windows
- Mono disassembler does not support recursive disassembling and produces output without IL and C#.
- Indirect calls are not tracked.
- To be able to compare different platforms, you need to target AnyCPU
<PlatformTarget>AnyCPU</PlatformTarget>
- To get the corresponding C#/F# code from disassembler you need to configure your project in following way:
<DebugType>pdbonly</DebugType>
<DebugSymbols>true</DebugSymbols>
[!includeIntroHardwareCounters]
[!includeIntroDisassemblyRyuJit]
[!includeIntroDisassembly]
[!includeIntroDisassemblyAllJits]
[!includeIntroDisassemblyDry]
[!includeIntroTailcall]
[!includeIntroJitStatsDiagnoser]
[!includeIntroNativeMemory]
[!includeIntroThreadingDiagnoser]
[!includeIntroExceptionDiagnoser]