,
6 tweets,
2 min read
Read on Twitter
All is now live in Beta 6!
You need to reinstall your apps (restore from iCloud is the easiest way) to get the repackaged app version.
Then have fun with 2nd/3rd party apps. And yes, the speed improvement happens on all devices (from the 6s to the Xs, iPads too).
You need to reinstall your apps (restore from iCloud is the easiest way) to get the repackaged app version.
Then have fun with 2nd/3rd party apps. And yes, the speed improvement happens on all devices (from the 6s to the Xs, iPads too).
And I didn’t post any side by side video (there are already some online) because it’s really a delight best experienced for oneself.
When all of this was enabled fully internally I spent half a day just doing A/B launches with the excitement of a kid on Christmas Day.
When all of this was enabled fully internally I spent half a day just doing A/B launches with the excitement of a kid on Christmas Day.
Some generic advices (that we made in the App Launch WWDC talk) that will take full advantages of this improvement.
For frameworks you use rarely:
- do not dlopen;
- do not use static initializers, but lazy ones (pthread_once/dispatch_once)
1/2
For frameworks you use rarely:
- do not dlopen;
- do not use static initializers, but lazy ones (pthread_once/dispatch_once)
1/2
Mind what the objc runtime tells you:
- avoid duplicate class names like the plague;
- try to put categories on the same class in the same binary (if you extend say NSString 3 times, do it from a single binary if possible)
- avoid swizzling during launch (or at all really)
2/2
- avoid duplicate class names like the plague;
- try to put categories on the same class in the same binary (if you extend say NSString 3 times, do it from a single binary if possible)
- avoid swizzling during launch (or at all really)
2/2
(by do not dlopen what I mean here is: just link them, the linking and runtime initializations of unused frameworks has become ridiculously low, but the cost is still the full thing at dlopen time, and doesn't benefit of dirty memory reduction optimizations and similar)
