nd LinearIndices prevents auto vectorization?

[N5N3]

Found this when tuning copyto_unalias!

Some benchmark

f1!(a,b) = @inbounds for i in LinearIndices(a)
        a[i+1] = b[i]
    end
f2!(a,b) = @inbounds @simd for i in LinearIndices(a)
        a[i] = b[i]
end

a = randn(8,8,8,8); b = similar(a);
@btime  f1!(b,a);   #  1.930 μs (0 allocations: 0 bytes)
@btime  f2!(b,a);   #  535.263 ns (0 allocations: 0 bytes)
a = randn(16,16,16); b = similar(a);
@btime  f1!(b,a);   #  1.300 μs (0 allocations: 0 bytes)
@btime  f2!(b,a);   #  552.105 ns (0 allocations: 0 bytes)
a = randn(64,64); b = similar(a);
@btime  f1!(b,a);   #  1.940 μs (0 allocations: 0 bytes)
@btime  f2!(b,a);   #  541.579 ns (0 allocations: 0 bytes)