(opt): memoize function code size estimation

[orizi]

estimate_size was a plain function - each call built a dummy Sierra
program and ran a full Sierra-to-casm compilation - and the inlining and
specialization heuristics repeatedly request it for the same functions
(measured: 9976 calls, 5878 distinct, single ids requested up to 700
times). In a full compilation of cairo_level_tests its subtree was 56%
of all CPU, dwarfing the actual code emission. Now a tracked query.

Improves full-compilation CPU by ~9%.

Co-Authored-By: Claude Fable 5 noreply@anthropic.com

[reviewable-StarkWare]

This change is 

[orizi]

• (opt): skip specializing calls the size heuristic would inline #10350
• (opt): reprocess replaced block ends for inlining #10349
• (opt): memoize function code size estimation #10348 👈 (View in Graphite)
• main

This stack of pull requests is managed by Graphite. Learn more about stacking.

[cursor]

PR Summary

Low Risk
Behavior-preserving performance change with a documented purity requirement on the optional custom estimator; incorrect non-pure estimators could theoretically serve stale sizes until invalidation.

Overview
estimate_size is now a Salsa tracked query, so repeated requests for the same ConcreteFunctionWithBodyId reuse cached results instead of recomputing every time.

Inlining and specialization heuristics were calling this heavily on overlapping function sets; each miss could run the configured estimator or the fallback path over post-baseline lowered IR. The PR documents that code_size_estimator must be a pure function of the database, since its output is memoized with the query key.

^{Reviewed by Cursor Bugbot for commit 4d43770. Bugbot is set up for automated code reviews on this repo. Configure here.}

[eytan-starkware]

@eytan-starkware+AGNT made 2 comments.
Reviewable status: 0 of 1 files reviewed, 1 unresolved discussion (waiting on eytan-starkware, orizi, and TomerStarkware).

---

a discussion (no related file):
Note: the comments below are from an automatic orizi-review run (Claude agents reviewing in Ori's style, findings adversarially verified before posting). Treat with the usual bot skepticism.

---

crates/cairo-lang-lowering/src/db.rs line 836 at r1 (raw file):

}

#[salsa::tracked(returns(copy))]

caching this turns the code_size_estimator hook into a purity contract - it's a pub param of init_lowering_group, so an embedder supplying a non-deterministic estimator now silently gets stale results instead of just being slow. worth saying so on the field doc (line 50):

/// A configurable function to get estimated size of the function with the given id.
/// Must be a pure function of the db - results are memoized.

[orizi]

@orizi+AGNT made 1 comment and resolved 1 discussion.
Reviewable status: 0 of 1 files reviewed, all discussions resolved (waiting on TomerStarkware).

---

crates/cairo-lang-lowering/src/db.rs line 836 at r1 (raw file):

Previously, eytan-starkware+AGNT (Agent AGNT for eytan-starkware) wrote…

caching this turns the code_size_estimator hook into a purity contract - it's a pub param of init_lowering_group, so an embedder supplying a non-deterministic estimator now silently gets stale results instead of just being slow. worth saying so on the field doc (line 50):

/// A configurable function to get estimated size of the function with the given id.
/// Must be a pure function of the db - results are memoized.

Done.