Ilja A. Iwas

@iljawascoding subclassing is not impossible. It is just opt in now. Before it was performance opt in. I applaud to the move.

@iljawascoding I wonder how apple managed to make watchOS faster and and watches last longer without hardware update?

@icex33 @iljawascoding Apple Watch CPU is 500MHz w/32KB L1, 256KB L2. Dynamic dispatch was not an issue on 25Mhz 68040 w/4K cache.

@mpweiher @iljawascoding really? No cache misses on different subclasses? how big are the virtual tables?

@icex33 @iljawascoding I didn’t say dynamic dispatch didn’t exist. I said it wasn’t an issue. And you could always IMP-cache at call-site.

@mpweiher @iljawascoding everyone currently burning cycles just because they don't know better. Now you will have to explicitly say BURN :)

@icex33 @iljawascoding Yes, burning cycles on the altar of Swift. Usually several times slower, sometimes more, seldom ess.

@iljawascoding Disagree. Has nothing to do with compiler design, everything with API/language design. Knowing what is subclassable is good

@uliwitness My point is that somebody with a compiler engineering background has a different POV than an app developer.

@iljawascoding @uliwitness this was a community proposal, though

@NeoNacho @uliwitness Decision by echo-chamber committee?

@iljawascoding @uliwitness ¯\_(ツ)_/¯ — just saying the idea didn’t necessarily originate from the language devs