There’s also a check for method swizzling and other invalidation, assuming that there is cooperation from the runtime. Unless I’m misunderstanding what you mean by the selector changing?
Saagar Jha > On Dec 16, 2019, at 00:16, Jean-Daniel <mail...@xenonium.com> wrote: > > >> Le 16 déc. 2019 à 06:05, Saagar Jha <saa...@saagarjha.com >> <mailto:saa...@saagarjha.com>> a écrit : >> >> It’s been a while, but I just thought you both might be interested in some >> follow-up I did for this idea. I implemented it for fun in clang >> <https://github.com/saagarjha/expresscall> and it turns out that it’s a >> pretty decent performance win >> <https://saagarjha.com/blog/2019/12/15/bypassing-objc-msgsend/> over >> objc_msgSend, both because it dispatches faster and because the compiler can >> do a full inline through it. > > Yes, but you don't preserve the objc_msgSend semantic. > > If I understand you code correctly, all you do is checking if it is the same > ISA, which does not guarantee in anyway that the selector did not change > since the previous call. > > As classes are fully dynamic and methods can be swizzle, you must perform a > full IMP lookup for every calls, which complexly defeat the purpose of inline > caching. > _______________________________________________ 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
