Performance issue with enumerate

[itsdfish]

I found a performance issue with enumerate. Here is my version information:

Julia Version 0.6.0 Commit 9036443 (2017-06-19 13:05 UTC) Platform Info: OS: Linux (x86_64-pc-linux-gnu) CPU: Intel(R) Core(TM) i7-4790K CPU @ 4.00GHz WORD_SIZE: 64 BLAS: libopenblas (USE64BITINT DYNAMIC_ARCH NO_AFFINITY Haswell) LAPACK: libopenblas64_ LIBM: libopenlibm LLVM: libLLVM-3.9.1 (ORCJIT, haswell)

The following code compares the execution time of enumerate and an iterator with a manually defined index. As you will see, enumerate is about 4-5 times slower.

using BenchmarkTools

type Person
  id::Int
end

function getID1(people)
  ids = fill(0,size(people,1))
  for (i,p) in enumerate(people)
    ids[i] = p.id
  end
  return ids
end

function getID2(people)
  ids = fill(0,size(people,1))
  i = 0
  for p in people
    i += 1
    ids[i] = p.id
  end
  return ids
end

people = [Person(id) for id in 1:1000]

v1 = @benchmarkable getID1(people)
tune!(v1)
r1 = run(v1)

v2 = @benchmarkable getID2(people)
tune!(v2)
r2 = run(v2)

Here are the results for enumerate:

BenchmarkTools.Trial: 
  memory estimate:  39.19 KiB
  allocs estimate:  1001
  --------------
  minimum time:     6.356 μs (0.00% GC)
  median time:      7.906 μs (0.00% GC)
  mean time:        11.842 μs (25.96% GC)
  maximum time:     529.366 μs (94.82% GC)
  --------------
  samples:          10000
  evals/sample:     5

And the results for the corresponding manual indexing:

BenchmarkTools.Trial: 
  memory estimate:  7.94 KiB
  allocs estimate:  1
  --------------
  minimum time:     1.886 μs (0.00% GC)
  median time:      2.088 μs (0.00% GC)
  mean time:        2.550 μs (10.38% GC)
  maximum time:     140.503 μs (87.76% GC)
  --------------
  samples:          10000
  evals/sample:     10

[KristofferC]

Dup of #16190 (which is dup of referenced issues therein).