Skip to content
New issue

Have a question about this project? Sign up for a free GitHub account to open an issue and contact its maintainers and the community.

By clicking “Sign up for GitHub”, you agree to our terms of service and privacy statement. We’ll occasionally send you account related emails.

Already on GitHub? Sign in to your account

Solver TD3-💨 with Side-Effect Gas #1358

Open
michael-schwarz opened this issue Feb 9, 2024 · 0 comments · May be fixed by #1442
Open

Solver TD3-💨 with Side-Effect Gas #1358

michael-schwarz opened this issue Feb 9, 2024 · 0 comments · May be fixed by #1442

Comments

@michael-schwarz
Copy link
Member

We have quite often suspected excessive widening at unknowns that receive side-effects (e.g. function entry nodes) as a cause of unnecessary precision-loss.

In a brainstorming session with @stilscher and @hseidl we came up with a possible solution. Instead of making an unknown a point after having received 2 increasing side-effects, one could generalize this to allow n increasing side-effects before making the unknown a widening point.

This should be able to do in an extensible way by replacing the point data structure by a customizable module that receives the information whether a side-effects increased the abstract value and is queried instead of point.
This might also cleanup many of the existing options we already have concerning handling widening of side-effects.

The same logic could in principle also apply to other unknowns, which would mostly concern loop heads by delaying widening. This seems like one would get it for free after implementing the idea above.

(The title of the issue is a pun on gas 💨, maybe ⛽ would be the more serious emoji, provided emoji can be serious)


The question then is whether @stilscher wants to do this herself or whether we should give it to a Master's student. Another alternative would be giving it to Ali if he wants to start contributing to the solver as a project to catch a foothold before moving on to the 🧵-TD3 (#644).

@stilscher stilscher self-assigned this Feb 9, 2024
@michael-schwarz michael-schwarz linked a pull request May 8, 2024 that will close this issue
Sign up for free to join this conversation on GitHub. Already have an account? Sign in to comment
Projects
None yet
Development

Successfully merging a pull request may close this issue.

2 participants