On Tue, Dec 28, 2021 at 2:54 PM Warren Weckesser <[email protected]> wrote:
> On 12/28/21, Lev Maximov <[email protected]> wrote: > > On Tue, Dec 28, 2021 at 3:43 PM Evgeni Burovski > > <[email protected]> > > wrote: > > > >> Very nice overview! > >> > >> One question and one suggestion: > >> > >> 1. Is integer wraparound guaranteed for signed ints, or is it an > >> implementation detail? For unsigned ints, sure, it's straight from a C > >> standard; what about signed types however. > >> > > Signed ints wraparound in just the same way as unsigned, both in C and in > > NumPy. Added an illustration. > > Overflow of *signed* ints in the C language is *undefined behavior*. > In practice, most compilers might do what you expect, but the > wrap-around behavior is not guaranteed and should not be relied on. > > Warren > > There used to be one's complement machines. I expect Numpy will break badly if it is compiled for anything that doesn't use two's complement. Chuck
_______________________________________________ NumPy-Discussion mailing list -- [email protected] To unsubscribe send an email to [email protected] https://mail.python.org/mailman3/lists/numpy-discussion.python.org/ Member address: [email protected]
