Seemingly large permformance regression on "^" power operator in >=1.6

Hello, I was doing a benchmark between Julia and a few other programming languages (Python, c++), when I was surprised to see Julia being slow compared to previous versions.

A quick benchmark between Julia 1.6.1 (1.6.0 gives same regression) and 1.5.0:

1.5.0
  167.168 ms (0 allocations: 0 bytes)
1.6.1
  852.693 ms (0 allocations: 0 bytes)

for the below code

using BenchmarkTools

function runLoop() 
    total_sum = 0.0
    for i in 0:10000000-1
        total_sum += i^0.1
    end
    return total_sum
end

println(VERSION)
# runLoop()
@btime runLoop()

Anyone have any idea why this might be?

1 Like

Aw thanks for the link, marked as solution