wyhash-dotnet

Zero-allocation C# implementation of Wang Yi's 64-bit wyhash hash algorithm and wyrand PRNG.

wyhash is an extremely fast, portable hashing algorithm, and passes all of the SMHasher tests (which evaluates collision, dispersion and randomness qualities of hash functions).

Note wyhash-dotnet currently implements wyhash v1. v2 will be implemented once considered "stable".

Getting Started

Install the WyHash package from NuGet:

Install-Package WyHash

WyHash implements HashAlgorithm, so can be integrated into existing projects easily:

var seed = 42;
var hasher = WyHash64.Create(seed);

var result = hasher.ComputeHash(myData);

A common use case is getting the resulting hash as a 64-bit unsigned integer (ulong), so a static convenience method is also provided:

// Note that if no seed is specified, it default to 0
var seed = 42;
var result = WyHash64.ComputeHash64(myData, seed);

The wyrand PRNG is also implemented:

var rng = new WyRng(42);

// Generate pseudorandom values
ulong a64BitVal = rng.NextLong();
uint a32BitVal = rng.Next();

// Fill a buffer with pseudorandom bytes
var buffer = new byte[256];
rng.NextBytes(buffer);

// Span is also supported
Span<byte> spanBuffer = stackalloc byte[128];
rng.NextBytes(spanBuffer);

Performance & Future Work

Note wyhash-dotnet currently implements wyhash v1. v2 will be implemented once considered "stable".

Future improvements include support for Span and incremental hashing (useful for hashing streams).

At present (July 2019), wyhash is the fastest algorithm in the SMHasher benchmark.

On a dev laptop with 64GB RAM and an Intel Xeon CPU E3-1545M v5 2.90GHz CPU, this implementation can process data at a rate of around 5.5GB/s on .NET Core 3, or 3.3GB/s on .NET Core 2 or the .NET Framework - this is very fast.

The reason for the performance improvement on .NET Core 3+ is that it supports hardware intrinsics, and wyhash-dotnet uses the BMI2 MULX instruction to achieve faster 64-bit integer multiplication (on systems where BMI2 is available). Support for intrinsics won't make it into the .NET Framework, but will also be in the newly announced .NET 5.

Note that PInvokeing into a native DLL built using the reference C code (see the WyHash.Native project) achieves around 10.8GB/s (more or less RAM SPEED), so there is still work to do to bridge the performance gap between C# and native - I'm very much open to suggestions here!

Latest benchmarks (DataSize is the size of data hashed, in bytes):

BenchmarkDotNet=v0.13.5, OS=Windows 10 (10.0.19045.3086/22H2/2022Update)
AMD Ryzen 5 2600, 1 CPU, 12 logical and 6 physical cores
.NET SDK=7.0.107
  [Host]            : .NET 6.0.18 (6.0.1823.26907), X64 RyuJIT AVX2
  ShortRun-.NET 6.0 : .NET 6.0.18 (6.0.1823.26907), X64 RyuJIT AVX2

Job=ShortRun-.NET 6.0  Runtime=.NET 6.0  IterationCount=3  
LaunchCount=1  WarmupCount=3  

Method	DataSize	Mean	Error	StdDev	Min	Max	Ratio	Rank	Allocated	Alloc Ratio
TestXxHash	100	17.77 ns	1.179 ns	0.065 ns	17.71 ns	17.84 ns	0.97	2	-	NA
TestXxHashNative	100	21.89 ns	4.087 ns	0.224 ns	21.75 ns	22.14 ns	1.19	4	-	NA
TestWyHashNative	100	14.58 ns	1.336 ns	0.073 ns	14.53 ns	14.67 ns	0.79	1	-	NA
TestWyHash	100	18.40 ns	2.422 ns	0.133 ns	18.30 ns	18.55 ns	1.00	3	-	NA
TestXxHash	1024	89.38 ns	10.387 ns	0.569 ns	88.95 ns	90.02 ns	0.73	3	-	NA
TestXxHashNative	1024	81.17 ns	3.474 ns	0.190 ns	80.98 ns	81.36 ns	0.67	2	-	NA
TestWyHashNative	1024	64.38 ns	2.001 ns	0.110 ns	64.27 ns	64.49 ns	0.53	1	-	NA
TestWyHash	1024	121.77 ns	23.792 ns	1.304 ns	120.53 ns	123.13 ns	1.00	4	-	NA