06:29 <LarstiQ> quotemstr: if I'm not wrong, because not every language wants to have C semantics

08:11 <mattip> Stumbled across this project using PyPy + cffi for a web framework. Benchmarks look good

08:11 <mattip> https://github.com/cirospaciari/socketify.py#zap-benchmarks

09:05 <arigato> quotemstr: arithmetic on INT doesn't work mostly by accident: when we started to develop pypy, most machines were still 32-bit, so we only allowed arithmetic on "Signed" which is equal to "INT". Later came the switch to 64-bit, and Signed is now 64-bit, and we never enabled arithmetic on the now-smaller "INT" (all RPython numbers are Signed)

09:49 <LarstiQ> luckily there are people who actually know to answer :)

14:02 <quotemstr> arigato: Thanks for the explanation!

14:02 <arigato> :-)

14:02 <quotemstr> I totally get being explicit instead of implicit --- C promotion rules can be arcane. I was just surprised, that's all. At least there's a nice error at compile time.

14:03 <quotemstr> Speaking of historical artifacts --- did any of the partial application machinery survive? One of the nice theoretical aspects of the Futamura projection thing is being able to take an interpreter and generate both a JIT and a compiled version of a given interpreted program

15:27 <cfbolz> quotemstr: our PE approach never really made it possible to produce a compiled version of a given program, because our PE was "lazy"

15:28 <cfbolz> it only explored some of the code paths eagerly, and then would wait for new information (eg a runtime type) to produce more code

15:28 <cfbolz> so it really was a *JIT*

15:34 <quotemstr> Ah