Clear TD3 sides when not reusing wpoint incrementally
#1010
Conversation
This file contains bidirectional Unicode text that may be interpreted or compiled differently than what appears below. To review, open the file in an editor that reveals hidden Unicode characters.
Learn more about bidirectional Unicode characters
Keeping sides means globals become wpoints on first side effect again.
sim642
commented
Mar 6, 2023
| HM.clear data.wpoint; | ||
| HM.clear data.sides |
There was a problem hiding this comment.
The real change is just this line being added.
michael-schwarz
approved these changes
Mar 22, 2023
stilscher
approved these changes
Mar 22, 2023
sim642
added a commit
to sim642/opam-repository
that referenced
this pull request
Sep 13, 2023
CHANGES: * Add `setjmp`/`longjmp` analysis (goblint/analyzer#887, goblint/analyzer#970, goblint/analyzer#1015, goblint/analyzer#1019). * Refactor race analysis to lazy distribution (goblint/analyzer#1084, goblint/analyzer#1089, goblint/analyzer#1136, goblint/analyzer#1016). * Add thread-unsafe library function call analysis (goblint/analyzer#723, goblint/analyzer#1082). * Add mutex type analysis and mutex API analysis (goblint/analyzer#800, goblint/analyzer#839, goblint/analyzer#1073). * Add interval set domain and string literals domain (goblint/analyzer#901, goblint/analyzer#966, goblint/analyzer#994, goblint/analyzer#1048). * Add affine equalities analysis (goblint/analyzer#592). * Add use-after-free analysis (goblint/analyzer#1050, goblint/analyzer#1114). * Add dead code elimination transformation (goblint/analyzer#850, goblint/analyzer#979). * Add taint analysis for partial contexts (goblint/analyzer#553, goblint/analyzer#952). * Add YAML witness validation via unassume (goblint/analyzer#796, goblint/analyzer#977, goblint/analyzer#1044, goblint/analyzer#1045, goblint/analyzer#1124). * Add incremental analysis rename detection (goblint/analyzer#774, goblint/analyzer#777). * Fix address sets unsoundness (goblint/analyzer#822, goblint/analyzer#967, goblint/analyzer#564, goblint/analyzer#1032, goblint/analyzer#998, goblint/analyzer#1031). * Fix thread escape analysis unsoundness (goblint/analyzer#939, goblint/analyzer#984, goblint/analyzer#1074, goblint/analyzer#1078). * Fix many incremental analysis issues (goblint/analyzer#627, goblint/analyzer#836, goblint/analyzer#835, goblint/analyzer#841, goblint/analyzer#932, goblint/analyzer#678, goblint/analyzer#942, goblint/analyzer#949, goblint/analyzer#950, goblint/analyzer#957, goblint/analyzer#955, goblint/analyzer#954, goblint/analyzer#960, goblint/analyzer#959, goblint/analyzer#1004, goblint/analyzer#558, goblint/analyzer#1010, goblint/analyzer#1091). * Fix server mode for abstract debugging (goblint/analyzer#983, goblint/analyzer#990, goblint/analyzer#997, goblint/analyzer#1000, goblint/analyzer#1001, goblint/analyzer#1013, goblint/analyzer#1018, goblint/analyzer#1017, goblint/analyzer#1026, goblint/analyzer#1027). * Add documentation for configuration JSON schema and OCaml API (goblint/analyzer#999, goblint/analyzer#1054, goblint/analyzer#1055, goblint/analyzer#1053). * Add many library function specifications (goblint/analyzer#962, goblint/analyzer#996, goblint/analyzer#1028, goblint/analyzer#1079, goblint/analyzer#1121, goblint/analyzer#1135, goblint/analyzer#1138). * Add OCaml 5.0 support (goblint/analyzer#1003, goblint/analyzer#945, goblint/analyzer#1162).
Sign up for free
to join this conversation on GitHub.
Already have an account?
Sign in to comment
Add this suggestion to a batch that can be applied as a single commit.
This suggestion is invalid because no changes were made to the code.
Suggestions cannot be applied while the pull request is closed.
Suggestions cannot be applied while viewing a subset of changes.
Only one suggestion per line can be applied in a batch.
Add this suggestion to a batch that can be applied as a single commit.
Applying suggestions on deleted lines is not supported.
You must change the existing code in this line in order to create a valid suggestion.
Outdated suggestions cannot be applied.
This suggestion has been applied or marked resolved.
Suggestions cannot be applied from pending reviews.
Suggestions cannot be applied on multi-line comments.
Suggestions cannot be applied while the pull request is queued to merge.
Suggestion cannot be applied right now. Please check back later.
Even though our default behavior is not reusing
wpointincrementally, we were reusingsides, which is used for turning global constraint variables into widening points. This meant that even when we didn't consider such variable a widening point after incremental loading, it would become one after a single side effect. This is different from the usual counting achieved viasides.There's no real value in reusing
sidesif we're trying to be more accurate by not reusingwpoint. Reusingsidesis not necessary for termination because variables will get re-added if necessary.I don't know if this has any real-world improvement for incremental precision. Theoretically it is possible, although it only makes a difference if the incremental run makes just a few side effects from a particular node.
Also includes changes for debug output related to the issue. The accidental reuse was hard to notice because no stats about
sideswere being printed. I suspected a problem because for incremental runs the number of widening points was much greater than at the end of the original run.