Compare MetadataToken by == instead of XOR ^

[SergeiPavlov]

Because (a ^ b) == 0 if and only if a == b

[botinko]

Why it was like that before?

[botinko]

I found it wasdone by AU 971370b
I remember he did a lot of benchmarking for this case. Maybe xor approach is faster?

[SergeiPavlov]

I remember he did a lot of benchmarking for this case. Maybe xor approach is faster?

I think he just copy/pasted this pattern from some other project.
See:
5119c1f#diff-4debfe6bcc478710248f83f0466c51fe370e88f121efc6f3875a40095f08124aR943
&
5119c1f#diff-4debfe6bcc478710248f83f0466c51fe370e88f121efc6f3875a40095f08124aR949

He used both: == & ^ in the same function

I don't belive the compiler can generate more optimal code than single CMP instruction to compare two ints

[snaumenko-st]

I think that we can also change

ReferenceEquals(Module1, Module2)

with

Module1.ModuleHandle == Module2.ModuleHandle

[SergeiPavlov]

I think that we can also change

ReferenceEquals() looks better. There is no pointer dereference while Module1.ModuleHandle == Module2.ModuleHandle reads two values from RAM to compare

[snaumenko-st]

I think that we can also change

ReferenceEquals() looks better. There is no pointer dereference while Module1.ModuleHandle == Module2.ModuleHandle reads two values from RAM to compare

I did some benchmarks earlier and ModuleHandle comparison was 10% faster than reference. Probably reference comparison involves more checks than in case of structs, or it may be something else... Probably it's worth to repeat the tests again.