Re:Future Directions for Optimizing Compilers
by wiz21c on Monday September 10, 2018 @09:43AM
(#17951724)
A bit off topic, but how does one frame the difference between "compiler level optimisation" and "high level optimisation" ? I can sure make a difference between optimising for CPU cycles and optimizing a QuickSort pivot search, but I have hard times generalizing/formalizing that idea. It seems important to me because it would be needed to set realistic/measurable objectives on compilers.
[ Parent
| on HN ]
Re:Future Directions for Optimizing Compilers
by CalChris on Monday September 10, 2018 @10:40AM
(#17952135)
I think that high level optimization is a synonym for machine independent optimization. The classic example would be common subexpression elimination. Similarly compiler level is a synonym for low level, machine dependent code generation. An example would be peephole optimization.
Front end vs back end.
Re:Future Directions for Optimizing Compilers
by masklinn on Monday September 10, 2018 @10:55AM
(#17952243)
I don't know, HLO could also be language-level optimisations (specific based on known language semantics), while LLO would be IR-level optimisations (more generic and reusable).
Re:Future Directions for Optimizing Compilers
by Q6T46nT668w6i3m on Monday September 10, 2018 @10:01AM
(#17951836)
In my opinion, it’s mostly ambiguous. Most compiler optimization researcher focuses on particular programming languages and particular targets rather than denotational or operational semantics and abstract machines.