Function in array of functions not defined

New to Julia
1

Sorry, I dont know if this question was answered already, but I wasn’t capable to find it.
I want to iterate over data which I want to apply to different functions, collected in an Array. In this case, I want to apply the macro benchmark to it and call a function with the specific value and the specific function to create an entry in a dict.

valuesForRun = [10000,100000,1000000,10000000,100000000]
    functions::Array{Function} = [simple_loop_sum, threads_sum, sharedarray_parallel_sum, sharedarray_mapreduce, pmap_sum_nb, pmap_sum_b]
    for value in valuesForRun
        for i=1:length(functions)
            println(functions[i])
            println(typeof(functions[i]))
            fun::Function = functions[i]
            println(fun)
            run = @benchmark fun(value);
            createTestResultDict(run,fun,value);

I can iterate over data and functionArray and all println’s print the correct values, but if I try to call fun with a value, i get UndefVarError. Perhaps, there is easier way like map, also.

2

Try interpolating the function, too. Your code modified as an MWE:

using BenchmarkTools
valuesForRun = 10.^(5:8)
dummy1(n) = 1+n
dummy2(n) = sum(sqrt.(1:n))
functions = [dummy1, dummy2]
for value in valuesForRun
    for f in functions
        println(f)
        println(typeof(f))
        run = @benchmark $f($value);
        println(run)
    end
end

Also, abstract type declarations are probably not buying you anything in terms of performance, so they can be omitted unless you have another reason to use them.

1 Like
3

thank you for your advice,
can you please explain, why splicing inside the macro is necessary? I thought the function was unquoted and therefore dont need to be explicitly evaluated. Am I wrong?

4

See the BenchmarkTools.jl manual on interpolation, especially this:

Normally, you can’t use locally scoped variables in @benchmark or @benchmarkable, since all benchmarks are defined at the top-level scope by design. However, you can work around this by interpolating local variables into the benchmark expression:

and note that f is a local variable in your code.

1 Like
5

Thank your for your help

6

You are probably looking for this.