Am Sat, 01 Jun 2013 23:20:19 +0200 schrieb "renoX" <[email protected]>:
> On Saturday, 1 June 2013 at 18:23:31 UTC, Mehrdad wrote: > [cut] > > > > You shouldn't be using 32-bit indices on x64, that defeats the > > whole point of x64. > > Some would disagree: have a look at x32: > http://en.wikipedia.org/wiki/X32_ABI > Being able to use efficiently 64 bit doesn't mean that you *have* > to use 64bit when 32bit is enough.. > renoX No wait, the whole 'efficiency' point was that on x64 you don't have to waste memory with the native word size for indexes, but can go with 32-bits. So it saves memory. But Mehrdad argued that we have x64 to be able to address beyond 2^32 (bytes, items, objects). And 32-bit indexes in generic code are thus asking for overflow bugs on x64 one day. x32 adds nothing to the discussion as you'd always use 32-bit indexes there. There is no point in storing more than 2^32 items when the address space is 32-bit. (Exceptions: bit arrays, file offsets, etc.) -- Marco
