search

maemre / cadmium

1 stars 0 forks source link

Possible research direction: JIT mode inference/speculation #3

Open maemre opened 5 years ago

maemre commented 5 years ago

I think discovering modes of predicates dynamically can allow us to perform some of Mercury's optimizations just-in-time while allowing the language to be dynamic. Gradual mode inference is another direction this research can go. In a gradual setting, we can infer as much mode information as possible and optimize the "moded" parts of the program, similar to how Julia uses types for performance.

maemre commented 5 years ago

This may tie in with the invariant inference research as we are trying to discover likely mode invariants here.