Remove WL extractor, merge into hacker revive

[oflatt]

No description provided.

[Copilot]

Pull request overview

This PR removes the legacy WL/greedy extractor path and makes the tiger-based extraction pipeline the sole extraction mechanism, updating the CLI, schedules, and profiling harness accordingly.

Changes:

• Removed --use-tiger / --non-weakly-linear configuration plumbing and consolidated extraction through the tiger pipeline.
• Added a types_only() schedule for roundtrip checking to avoid canonicalization/always-run rewrites during that validation path.
• Updated profiling treatments and cleaned up related code (including moving is_dummy_ctx into from_egglog.rs).

Reviewed changes

Copilot reviewed 8 out of 8 changed files in this pull request and generated 3 comments.

Show a summary per file

File	Description
src/main.rs	Removes deprecated CLI flags and updates EggccConfig construction accordingly.
infra/profile.py	Drops removed CLI flags from benchmark invocations for tiger-related treatments.
dag_in_context/src/schedule.rs	Adds types_only() schedule and simplifies schedule construction; always runs non-weakly-linear in the parallel schedule.
dag_in_context/src/optimizations/hackers_delight.egg	Clarifies semantics around infinite loops / lowbit(0) in comments.
dag_in_context/src/linearity.rs	Removes the legacy linearity-checking helper module (greedy extractor-era).
dag_in_context/src/lib.rs	Deletes greedy-extractor integration and routes extraction/roundtrip through tiger; introduces local serialized_egraph()/has_debug_exprs().
dag_in_context/src/greedy_dag_extractor.rs	Removes the legacy greedy DAG extractor implementation (and its tests).
dag_in_context/src/from_egglog.rs	Inlines is_dummy_ctx() to remove dependency on the deleted greedy extractor module.

Comments suppressed due to low confidence (1)

dag_in_context/src/lib.rs:951

• When has_debug_exprs is true, the code logs that it will “extract them instead of original program” and then returns early after extract(). However, extract() no longer has a debug-expr extraction path (the previous greedy extractor handled DebugExpr). This likely makes the debug-expression visualization flow incorrect: the optimizer stops early but still extracts the normal functions. Consider either implementing DebugExpr extraction in the tiger path (or via egglog extraction over the DebugExpr relation) or removing the early-stop/log so behavior matches reality.

            let has_debug_exprs = has_debug_exprs(&serialized);
            if has_debug_exprs {
                log::info!(
                    "Program has debug expressions, extracting them instead of original program."
                );
            }
            let (iter_result, region_timings, extract_time) = extract(
                eggcc_config,
                &res,
                batch.clone(),
                &serialized,
                should_maintain_linearity,
            );

            eggcc_extraction_time += extract_time;
            eggcc_serialization_time += serialization_duration;

            extract_region_timings.extend(region_timings);

            // typecheck the program as a sanity check
            iter_result.typecheck();

            res = iter_result;

            if has_debug_exprs {
                log::info!("Program has debug expressions, stopping pass {}.", i);
                return Ok((
                    res,
                    EggccTimeStatistics {
                        eggcc_extraction_time,
                        eggcc_serialization_time,
                        extract_region_timings,
                    },
                ));
            }

---

💡 Add Copilot custom instructions for smarter, more guided reviews. Learn how to get started.

[FTRobbin]

No changes to the optimization rules themselves? Am I missing something?

[FTRobbin]

LGTM. Only a small question.

[FTRobbin]

Why do we need this?

[oflatt]

It was a bug our test suite revealed- we need those types for the extractor to work