Generators speed

New to Julia
2

You should not benchmark in global scope.

using BenchmarkTools

N = 10000000
A = randn(N)

sum1(A) = sum(A)
sum2(A) = sum(A[i] for i in 1:N)
sum3(A) = sum(a for a in A)

then

julia> @benchmark sum1($A)
BenchmarkTools.Trial: 
  memory estimate:  0 bytes
  allocs estimate:  0
  --------------
  minimum time:     4.576 ms (0.00% GC)
  median time:      5.744 ms (0.00% GC)
  mean time:        5.672 ms (0.00% GC)
  maximum time:     6.901 ms (0.00% GC)
  --------------
  samples:          881
  evals/sample:     1

julia> @benchmark sum2($A)
BenchmarkTools.Trial: 
  memory estimate:  112 bytes
  allocs estimate:  5
  --------------
  minimum time:     12.764 ms (0.00% GC)
  median time:      12.885 ms (0.00% GC)
  mean time:        12.972 ms (0.00% GC)
  maximum time:     16.242 ms (0.00% GC)
  --------------
  samples:          386
  evals/sample:     1

julia> @benchmark sum3($A)
BenchmarkTools.Trial: 
  memory estimate:  48 bytes
  allocs estimate:  3
  --------------
  minimum time:     12.761 ms (0.00% GC)
  median time:      12.874 ms (0.00% GC)
  mean time:        12.941 ms (0.00% GC)
  maximum time:     16.360 ms (0.00% GC)
  --------------
  samples:          387
  evals/sample:     1

so the difference is 2-3x. Above using v0.6-rc1.

1 Like