Optimization problem: bulk Boolean operations on vectors
Walter Bright via Digitalmars-d
digitalmars-d at puremagic.com
Sat Dec 24 04:00:47 PST 2016
On 12/24/2016 2:28 AM, Johan Engelen wrote:
> Of course, but it's not codified in the source and thus the function is
> different from the foreach version.
>
> More importantly, I think it is broken. I find it very hard to reason about the
> ptrdiff_t version, because of all the potential overflows and the mixing of
> unsigned and signed math.
You are pedantically correct. It's possible someone would write a bizarre loop,
and compiler optimizers need to take that into account when doing loop
transformations.
This is one of the reasons why D's array syntax and semantics is superior - it
guarantees no overflow.
More information about the Digitalmars-d
mailing list