I am going through this ML tutorial on Julia, and I noticed something odd. Output of the following script using BenchmarkTools using Flux @btime f(x) = x^2+1 @btime df(x) = gradient(f,x,nest=true)[1] is this: 0.908 ns (0 allocations: 0 bytes) 0.908 ns (0 allocations: 0 bytes) But it takes …

@btime takes 5 seconds to run because it will run your function repeatedly for about 5 seconds (to get accurate runtime measurements).

Real runtime vs. @btime output

jafar.isbarov January 23, 2022, 8:57am 11

@mkitti Thanks. I will give that a try.

Topic		Replies	Views
@time vs @btime Performance	5	17352	March 22, 2018
Benchmarking with @time @btime and subsequent runs return shorter execution time New to Julia benchmark , benchmarktools	8	2112	December 31, 2020
Save @btime output New to Julia question , benchmarktools	7	3392	April 1, 2023
@btime eats my variables New to Julia benchmarktools	5	2743	May 1, 2019
.== performance regression New to Julia	10	1043	September 4, 2017