Using a large numeric array
I compared the performance of comparisons with zero and non zero comparison tolerance using
t←⎕t ⋄ ←x=1 ⋄ ⎕t-t
I expected a dramatic speedup for ⎕ct=0 over ⎕ct>0, because it is so much simpler.
But I did not see a consistent difference at all.
I concluded that must be no internal check for a comparison tolerance of 0. Is that right?
If not, couldn't there be a dramatic speedup?