Fast and with zero allocations implementation of msgpack specification.
When you're sure what you're doing, use WriteXXXX methods or ReadXXXX, they are faster, but can throw exceptions. TryXXXX methods are more convinient if you're not sure about anything.
Numbers worth more than thousands of words. Below results of serialization of array of one hundred of large uints. As we can see, code in C#(MsgPackSpecArray) roughly twice as slow as native code. Can we go faster? Yes, we can, look to PointerBigEndian version: it's slower than C version only by 10%. But you usually don't want to right code like below. Let's hope that CLR team will improve Span a lot more. And may be some more bounds checking will be eliminated.
BenchmarkDotNet=v0.11.0, OS=debian 9
Intel Core i7-6700HQ CPU 2.60GHz (Skylake), 1 CPU, 8 logical and 8 physical cores
.NET Core SDK=2.1.300
[Host] : .NET Core 2.1.0 (CoreCLR 4.6.26515.07, CoreFX 4.6.26515.06), 64bit RyuJIT
DefaultJob : .NET Core 2.1.0 (CoreCLR 4.6.26515.07, CoreFX 4.6.26515.06), 64bit RyuJIT
Method | Mean | Error | StdDev | Median | Q3 | Scaled | ScaledSD | Allocated |
---------------------- |----------:|---------:|---------:|----------:|----------:|-------:|---------:|----------:|
MsgPackSpecArray | 170.75 ns | 3.391 ns | 6.286 ns | 169.92 ns | 175.26 ns | 1.75 | 0.07 | 0 B |
PointerBigEndian | 103.19 ns | 1.181 ns | 1.047 ns | 102.86 ns | 103.49 ns | 1.06 | 0.02 | 0 B |
CArray | 97.69 ns | 1.723 ns | 1.611 ns | 97.08 ns | 98.57 ns | 1.00 | 0.00 | 0 B |
CppArray | 87.18 ns | 2.086 ns | 1.951 ns | 86.90 ns | 88.96 ns | 0.89 | 0.02 | 0 B |Pointers in C# code:
public unsafe void PointerBigEndian()
{
fixed (byte* pointer = &_buffer[0])
{
pointer[0] = DataCodes.Array16;
Unsafe.WriteUnaligned(ref pointer[1], BinaryPrimitives.ReverseEndianness(length));
for (var i = 0u; i < length; i++)
{
pointer[3 + 5 * i] = DataCodes.UInt32;
Unsafe.WriteUnaligned(ref pointer[3 + 5 * i + 1], BinaryPrimitives.ReverseEndianness(baseInt));
}
}
}