Why findall() function is super slow?

General Usage Performance
1

Here is an example and I’m just wondering why it takes 51 mS to do this considering the input size is very small.

aa = [1,2,3,4,5,6]
bb = [2,5]
@time findall(x->x==bb, aa)

julia> @time findall(x->in(x, bb), aa)
  0.051113 seconds (78.99 k allocations: 4.014 MiB)
2-element Array{Int64,1}:
 2
 5
2

It is compiling the findall function. Run the timing twice in the same session and you will see it is much faster the second time. I would suggest using the BenchmarkTools package for microbenchmarks, as it takes care of that and a few other things.

1 Like
3

Separately, I’ve found in is pretty slow if the collection isn’t a set – you’ll get better performance converting bb to a set first.

4

No, that’s not it. See This post. OP, it’s because the anonymous function is compiling, and you compile it again every time you run that call in global scope.

Won’t be a problem if you have that command in a function.

8 Likes
5

Yes, I noticed that it’s a lot faster when it is used inside a function.

6

Nitpick: it’s “ms”. “mS” is millisiemens, a unit of conductance.

2 Likes