nmake, your test code is too small to be used as a benchmark. You have to use at least 5..10 seconds compile times in order to get more representative results.

JohnFound, for general benchmarks, I agree with you, for most practical purposes I agree, in most assemblers it is true, but in macro assemblers it is not the size that determines compile time.

My benchmark is not a typical benchmark, it is an abnormal benchmark, something bad happens when it should not

Since no one else seems to see this effect then we are unable to help you find the reason. Perhaps if you posted your source someone else can reproduce the results.

I have some examples, where fasmw.exe compiles twice as long as fasm.exe. But this behaviour is stable between different versions.

You do? I thought I was all alone in this.

But quite frankly, it doesn't matter much with this "phenomenon", because the latest version solved it for me.

nmake
You are alone, because I experience no performance improvement related to the latest version.

Oh well thanks for all help