Crystal can't measure time below 1100 nanoseconds

[TheOnlyArtz]

Hey, I've done a little experiment with a friend where we tried to iterate the biggest number possible and check the time it takes, we go to UInt128, it went like:

iter : UInt128 = 0_u128

start = Time.monotonic
        UInt128::MAX.times {iter += 1}
elapsed = Time.monotonic - start


puts "Time took to iterate #{iter} times => #{elapsed.nanoseconds}"

The run took 1100 nanoseconds, I was pretty impressed, I thought it was pretty good, but then a friend ran in on Rust and it took 89 nanoseconds, I thought it's impossible and tried to run the crystal binary again (compiled using --release), I've got 1100 again.. and again, and again.

until eventually, I went down to UInt32 in order to be even "faster" and I still got the same result, 1100 nanoseconds, this must be an issue with Crystal, I tried the Time.measure method too and it gave me 1100 at the minimum too!

Runs:
UInt128:

UInt32:

[TheOnlyArtz]

Crystal 0.27.0 btw.

[asterite]

What's your OS? And what does running puts Time.measure { } with --release mode give you? I get times like 54 nanoseconds and 80 nanoseconds.

[TheOnlyArtz]

I ran it on a WSL, every time compiled with --release, it's Debian.

[Blacksmoke16]

Might be something to do with WSL then. On Debian proper, it's around 40-60 ns.

[straight-shoota]

This is probably #7377 and already fixed. Please try with 0.28.0

[TheOnlyArtz]

This is probably #7377 and already fixed. Please try with 0.28.0

Yep, it is fixed, thank you :)