When is it safe to use Base.OneTo?

General Usage Performance
1

Since Base.OneTo is not exported I am suspicious that there might be some gotchas with it.

I implemented a padded array interface to perform in-place r2c and c2r fft transformations.
For that I have to create views that are always something like (pseudo-code) @view a[begin:(end-2), : , : , ... , :] where begin is the first index of the first dimension.
As you can see below, using OneTos gives me the best performance.
I’ll be creating views of plain julia’s Arrays, so no array with custom indices. So, is it safe to use Base.OneTo directly?


julia> data = rand(130,128,128);

julia> a = view(data,1:128,1:128,1:128);

julia> b = view(data,1:128,:,:);

julia> c = view(data,Base.OneTo(128),Base.OneTo(128),Base.OneTo(128));

julia> using BenchmarkTools

julia> @btime sum($a)
  5.230 ms (0 allocations: 0 bytes)
1.0485345086366412e6

julia> @btime sum($b)
  4.549 ms (0 allocations: 0 bytes)
1.0485345086366412e6

julia> @btime sum($c)
  4.300 ms (0 allocations: 0 bytes)
1.0485345086366412e6

julia> versioninfo()
Julia Version 0.7.0-beta2.0
Commit b145832402* (2018-07-13 19:54 UTC)
Platform Info:
  OS: macOS (x86_64-apple-darwin17.6.0)
  CPU: Intel(R) Core(TM) i5-5287U CPU @ 2.90GHz
  WORD_SIZE: 64
  LIBM: libopenlibm
  LLVM: libLLVM-6.0.0 (ORCJIT, broadwell)
Environment:
  JULIA_NUM_THREADS = 2

Note that the : creates ranges of type Slice{OneTo{Int}}, so it is taking the advantages of OneTo…

1 Like