Talk:Adaptive optimization

RE: reverse execution
Not original author, but I am pretty sure he doesn't mean reverse execution in theory. I agree with "which states that deoptimization is related to switch from optimized native code back to unoptimized interpreted code."

> Anyway, I'd be very interested in an implementation for the HotSpot JVM!

This is not an implementation of the HotSpot, while I do think that hot spot does have this (after all they do OSR) Here is a reference that explicitly uses the wording "reverse transition" to refer to "from compiled to interpreter".

https://dl.acm.org/citation.cfm?id=191116 — Preceding unsigned comment added by Erickomoto (talk • contribs) 21:51, 18 July 2018 (UTC)

reverse execution
The sentence "In some systems, notably the Java Virtual Machine, execution over a range of bytecode instructions can be provably reversed." sound as if especially the JVM supports reverse execution. Does it really? Or did the author mean: interpreted languages generally support reverse execution in theory?

I searched for information about reverse execution in HotSpot JVM but all I found was a glossary (glossary) which states that deoptimization is related to switch from optimized native code back to unoptimized interpreted code. There is just a research article on a prototypical implementation for Kaffe which supports reverse execution.

Anyway, I'd be very interested in an implementation for the HotSpot JVM!

BTW: According to google search results 'unwind' usually refers to exception handling and unwinding the call stack to find matching exception handlers or unwinding loops which seems to be an optimization technique.

examples of other implementations
it would be better to cite other implementations of this, pypy for example