> On Jan 29, 2015, at 5:29 PM, Quincey Morris > <quinceymor...@rivergatesoftware.com> wrote: > > On Jan 29, 2015, at 17:18 , Roland King <r...@rols.org> wrote: >> >> I can find this >> http://lists.apple.com/archives/objc-language/2013/Apr/msg00119.html >> <http://lists.apple.com/archives/objc-language/2013/Apr/msg00119.html>, is >> that what you were remembering or is there something stronger my google-fu >> isn't finding, which wouldn't be for the first time? > > I think that’s it. A few months ago, as I said. ;) > > The only odd thing is “non-object”. Based on the rest of Greg’s statement, I > think it’s a typo for “non-struct”. I can’t think of a reason why an 8-byte > pointer would produce different results from an 8-byte int. But I don’t > really know.
"Non-object" is correct. `atomic` makes a big difference for a strong or weak property of object type because objects have retain counts. There is one exception to the "non-object scalar" claim. On 32-bit ARM non-object scalar properties that are larger than 4 bytes are slower when atomic. This includes types `double` and `long long`. The problem is that the 32-bit ARM ABI only aligns storage for those types at 4-byte boundaries, and the CPU does not guarantee atomicity for ordinary unaligned 8-byte memory access. Atomic accessors for those types need to do extra work. -- Greg Parker gpar...@apple.com Runtime Wrangler _______________________________________________ Cocoa-dev mailing list (Cocoa-dev@lists.apple.com) Please do not post admin requests or moderator comments to the list. Contact the moderators at cocoa-dev-admins(at)lists.apple.com Help/Unsubscribe/Update your Subscription: https://lists.apple.com/mailman/options/cocoa-dev/archive%40mail-archive.com This email sent to arch...@mail-archive.com
