Well all my benchmarks (even using the exact code you created) show different results. I get the results:
if1: 3725
if2: 3616
if3: 3034
if1 is if (1 == 1) { haltdef }
if2 is if 1 == 1 { haltdef }
if3 is if (1 == 1) haltdef
So from that test it is clear that the () version is faster, so if you are saying your results show that the one without () are faster, it would prove my argument that your tests are flawed. I have done the test about 50 times now, and each time if3 comes out being the fastest.
