GetHashCode seems to only use the first field of a struct to calculate hash code

[miguep]

With a struct like this:

    struct MyStruct
        {
            public float a,b;
        };

if I try the following:

            MyStruct myStruct1 = new MyStruct { b= 1.0f, a = 2.0f};
            MyStruct myStruct2 = new MyStruct { b= 1.0f, a = 1.0f};

            int hash = myStruct1.GetHashCode();
            int hash2 = myStruct2.GetHashCode();

hash and hash2 are different, but trying this:

            MyStruct myStruct1 = new MyStruct { a= 1.0f, b = 2.0f};
            MyStruct myStruct2 = new MyStruct { a= 1.0f, b = 1.0f};

            int hash = myStruct1.GetHashCode();
            int hash2 = myStruct2.GetHashCode();

hash and hash2 are the same.

This behavior is different from the one in .NET Framework

[danmoseley]

@jkotas was this an intentional change? I suppose some line may have to be drawn otherwise computing a hash for a struct could be arbitrarily expensive.

[jkotas]

Yes, this is side-effect of using functionally correct hash algorithm for float/double: https://github.com/dotnet/coreclr/pull/13164/files#diff-ca326b8cd58d6642f56aa054c221c22aR2717

[jkotas]

some line may have to be drawn otherwise computing a hash for a struct could be arbitrarily expensive.

Right, the built-in hash algorithm for structs has number of trade offs, and it is pretty slow too. It is best to always implement GetHashCode on structs explicitly.