Conversation

I wonder when 32-bit systems (regardless of underlying arch (intel, arm, mips, etc.)) will become prohibitively more expensive than 64-bit for all markets.

Replying to

they'll just be more embedded the 32 bit systems literally take up less die space

Cy Schubert

Replying to

and

And 32-bit requires less stack and heap because pointers and other ints are half the size. This means less money spent on embedded systems memory which translates into more profits for manufacturers. Why do automakers still use ancient technology? That's why.

Replying to

and

Can happily use 32-bit pointers and integers on a 64-bit architecture too. As an example, the Android Runtime uses 32-bit pointers for Kotlin/Java pointers and most of the internal pointers. Running on 64-bit also doesn't result in non-FFI libraries using more 64-bit integers.

Replying to

C ended up with a weird approach to integers from portability to a lot of weird architectures with horrible things like segmentation. POSIX chose to deal with it in a way that ends up using far more 64-bit integers than alternatives. Lots of other languages don't really do it.

Replying to

Modern low-level languages tend to just have fixed-size integers and integers for sizes/addresses so only the ones for sizes get larger. Many older languages/libraries don't bother supporting > 2GB objects so they just use 32-bit integers so mostly only pointers end up 64-bit.

8:15 PM · Sep 20, 2021Twitter Web App