I’ve been working on extending Kitten’s permission system to support algebraic effects in some form. There are many tradeoffs, especially perf vs. generality. It’s easy to implement algebraic effects with first-class continuations if you’re willing to make all plain calls slower.
Conversation
A choice I made for performance reasons—namely, avoiding reification of the stack in memory, using traditional register allocation and interleaved data & call stacks—makes it hard to capture & restore. Heap-allocating stack frames makes it easier, but that impacts plain calls.
2
1
Replying to
Oh this is cool and exciting! Looking forward to hearing more about this!