Greg Parker

JavaScript really made it. We now tweak CPUs to make it faster.

Remember Jazelle?

I know it seems insane, but IMO this is a good thing. A lot of other software does similar conversions, as well as some other languages.

Sorry, but AMD can do something like this, but ARM is a RISC architecture. It should focus on core instructions. That's the whole point of RISC. So basically, what this development means: ARM for high performance is dead. We need to use RISC-V instead.

It means they can't think on anything better to do with all the free transistors (cheaper to add a FET to a CPU than a grain of salt to a bag of salt). This won't change until we stop thinking imperatively.

Actually I this was more a marketing driven decision. A lot of tech media is using web-based benchmark tests... and now they are faster. There are also a lot of people who think it is a good idea to use JS for apps.

There are a lot of cases where sharing logic in JS for apps is great. Also, it’s reality and it’s smart to acknowledge that and make these existing (and new) apps faster.

Works pretty well here for us. Using one negative example doesn’t really reflect the market. Heck, even PDF has JavaScript logic, I happily take any CPU acceleration I can get.

Two examples. I also know a lot of other companies who struggled with that. But that's not the main argument against it: Inefficient is. Sorry, but we have to do things in a more efficient way. We don't have a second planet. And a Chat with 500Mb of memory isn't that efficient.

CPUs conditioned by a shitty programming language. Seeing is believing...

A great deal of use by mainstream users is the Web & the primary programming language of the Web JS. JS is prob one of the few prog langs you can count on will be on a PC & mobile. It only makes sense for the best interest of mainstream users for CPUs optimize the use of JS…

JavaScript is shit no matter what you say or do. If you adapt a processor's instruction set to a shit, you're doing it wrong.

If your CPU design performs some popular computation inefficiently simply because you think that computation is bad, you're doing it wrong. CPU design is not a morality play.

Does the pointer magic affect NaN-boxing?

Pointer authentication, you mean? It's opt-in—the simplest use only authenticates call stack return addresses—so you could continue to use un-authenticated pointers in your boxed values.

It's never been a better time to be a Javascript developer on iOS

Except for all the missing web standards in Safari

Yeah more like “there’s never been a better time to benchmark JavaScript on iOS”