Skip to content

Conversation

@jberthold
Copy link
Member

@jberthold jberthold commented Nov 18, 2025

Evaluates the static data lookup functions using the Haskell backend evaluator only.
The lookup functions are annotated with no-evaluators so they are never sent to the LLVM backend library for evaluation. Default equations are generated together with the defining equations from the SMIR JSON data.

  • For kmir prove-rs, the prover uploads an extra module (manually) with the lookup function equations. Unfortunately, this requires us to also use the prover's extra_module option to ensure that the equations are in scope for the proof.

KMIR.from_kompiled_kore(..., symbolic=True) only appends to the Haskell definition.kore, and uses the unmodified LLVM library. All clients of from_kompiled_kore will use the new HS-only evaluation automatically.

  • kmir section-edge uses KMIR.from_kompiled_kore(..., symbolic=True)
  • kmir run --haskell-backend uses KMIR.from_kompiled_kore(..., symbolic=True)
  • For kmir run without --haskell-backend, we still invoke llvm-kompile (within KMIR.from_kompiled_kore(..., symbolic=False) and use the LLVM interpreter with the lookup function equations included.

@jberthold
Copy link
Member Author

Unfortunately this causes core dumps (most likely in the LLVM library) in booster during p-token proofs

Sign up for free to join this conversation on GitHub. Already have an account? Sign in to comment

Labels

None yet

Projects

None yet

Development

Successfully merging this pull request may close these issues.

2 participants