On Friday, 27 May 2016 at 21:31:48 UTC, David Nadlinger wrote:
Just about the only reason I could think of for this to happen
is if the compiler fails to inline the range primitives from
std.array. Otherwise, the loops should be pretty much
equivalent to LLVM's optimiser.
This is so fundamental to D's strategic promise that we can't
afford to get it wrong.
— David
I agree, and it has been said on this forum that tight manual
loops are faster than the range paradigm (I think bearophile was
one of them, if I remember correctly). Maybe we should benchmark
this too, just in case.