[wasm] Don't generate unused imports in jiterpreter traces #83061
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Outdated suggestions cannot be applied.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.
This PR restructures trace generation so that we generate the trace body before the rest of the module, which allows us to identify which imports are actually used and only emit those. This frees up more space for the actual trace and probably also makes instantiating the module a bit faster since it will typically have 1-8 imports instead of the dozens it does now. This should help a lot for #82963, and it also will allow #82773 to potentially do direct calls (by dynamically adding new imports) instead of indirect calls.
This PR also fixes a bug that was causing the heat estimation mode to undercount failed traces.