fix: emit all allocas in the init block instead of the current block

[TomerStarkware]

Summary

Several places emitted llvm.alloca into the libfunc's own block rather than the function's init (pre-entry) block. When that block sits inside a loop, the alloca re-executes on every iteration and the stack grows without bound.

All allocas now go through the init block, which runs exactly once per function call:

• felt252_dict_squash: range-check and gas pointers.
• squashed_dict into_entries: result array slot.
• qm31 binary ops: lhs/rhs operand slots.
• RuntimeBindingsMeta::libfunc_qm31_bin_op: result slot. Now takes the LibfuncHelper instead of a bare &Module (same as dict_get); callers already passed the helper via deref.
• trace_dump::build_state_snapshot (with-trace-dump feature): per-variable snapshot slots emitted at every statement. Takes an explicit init block; both call sites in compiler.rs pass the pre-entry block.

Test plan

• cargo check with and without with-trace-dump
• cargo clippy --all-targets clean, cargo fmt applied
• cargo test --lib -- qm31 felt252_dict squashed_dict (41 passed)

🤖 Generated with Claude Code

[github-actions]

Benchmark results Main vs HEAD.

Base

Command	Mean [s]	Min [s]	Max [s]	Relative
base dict_insert.cairo (JIT)	1.548 ± 0.029	1.512	1.601	1.03 ± 0.02
base dict_insert.cairo (AOT)	1.496 ± 0.012	1.472	1.509	1.00

Head

Command	Mean [s]	Min [s]	Max [s]	Relative
head dict_insert.cairo (JIT)	1.766 ± 0.012	1.748	1.780	1.01 ± 0.01
head dict_insert.cairo (AOT)	1.748 ± 0.006	1.738	1.757	1.00

Base

Command	Mean [s]	Min [s]	Max [s]	Relative
base dict_snapshot.cairo (JIT)	1.363 ± 0.017	1.346	1.394	1.03 ± 0.02
base dict_snapshot.cairo (AOT)	1.330 ± 0.017	1.298	1.356	1.00

Head

Command	Mean [s]	Min [s]	Max [s]	Relative
head dict_snapshot.cairo (JIT)	1.593 ± 0.010	1.579	1.608	1.02 ± 0.01
head dict_snapshot.cairo (AOT)	1.565 ± 0.008	1.550	1.577	1.00

Base

Command	Mean [s]	Min [s]	Max [s]	Relative
base factorial_2M.cairo (JIT)	1.416 ± 0.016	1.391	1.444	1.00
base factorial_2M.cairo (AOT)	1.422 ± 0.029	1.384	1.484	1.00 ± 0.02

Head

Command	Mean [s]	Min [s]	Max [s]	Relative
head factorial_2M.cairo (JIT)	1.639 ± 0.009	1.624	1.649	1.01 ± 0.01
head factorial_2M.cairo (AOT)	1.620 ± 0.008	1.604	1.630	1.00

Base

Command	Mean [s]	Min [s]	Max [s]	Relative
base fib_2M.cairo (JIT)	1.379 ± 0.024	1.350	1.438	1.03 ± 0.02
base fib_2M.cairo (AOT)	1.335 ± 0.012	1.314	1.354	1.00

Head

Command	Mean [s]	Min [s]	Max [s]	Relative
head fib_2M.cairo (JIT)	1.593 ± 0.008	1.580	1.601	1.01 ± 0.01
head fib_2M.cairo (AOT)	1.572 ± 0.009	1.563	1.586	1.00

Base

Command	Mean [s]	Min [s]	Max [s]	Relative
base linear_search.cairo (JIT)	1.375 ± 0.017	1.354	1.406	1.04 ± 0.01
base linear_search.cairo (AOT)	1.327 ± 0.006	1.316	1.336	1.00

Head

Command	Mean [s]	Min [s]	Max [s]	Relative
head linear_search.cairo (JIT)	1.620 ± 0.005	1.611	1.628	1.02 ± 0.01
head linear_search.cairo (AOT)	1.587 ± 0.008	1.574	1.604	1.00

Base

Command	Mean [s]	Min [s]	Max [s]	Relative
base logistic_map.cairo (JIT)	1.359 ± 0.021	1.333	1.394	1.01 ± 0.02
base logistic_map.cairo (AOT)	1.341 ± 0.025	1.309	1.394	1.00

Head

Command	Mean [s]	Min [s]	Max [s]	Relative
head logistic_map.cairo (JIT)	1.597 ± 0.012	1.583	1.626	1.02 ± 0.01
head logistic_map.cairo (AOT)	1.562 ± 0.010	1.552	1.581	1.00

[github-actions]

Benchmarking results

Benchmark for program dict_insert

Open benchmarks

Command	Mean [s]	Min [s]	Max [s]	Relative
Cairo-vm (Rust, Cairo 1)	11.488 ± 0.058	11.363	11.555	5.99 ± 0.11
cairo-native (embedded AOT)	1.995 ± 0.036	1.910	2.030	1.04 ± 0.03
cairo-native (embedded JIT using LLVM's ORC Engine)	1.917 ± 0.034	1.874	1.969	1.00

Benchmark for program dict_snapshot

Open benchmarks

Command	Mean [ms]	Min [ms]	Max [ms]	Relative
Cairo-vm (Rust, Cairo 1)	561.7 ± 11.1	545.6	580.9	1.00
cairo-native (embedded AOT)	1755.1 ± 22.7	1725.9	1802.2	3.12 ± 0.07
cairo-native (embedded JIT using LLVM's ORC Engine)	1786.5 ± 18.0	1754.3	1810.3	3.18 ± 0.07

Benchmark for program factorial_2M

Open benchmarks

Command	Mean [s]	Min [s]	Max [s]	Relative
Cairo-vm (Rust, Cairo 1)	5.090 ± 0.031	5.046	5.143	2.77 ± 0.04
cairo-native (embedded AOT)	1.841 ± 0.021	1.796	1.879	1.00
cairo-native (embedded JIT using LLVM's ORC Engine)	1.871 ± 0.042	1.820	1.944	1.02 ± 0.03

Benchmark for program fib_2M

Open benchmarks

Command	Mean [s]	Min [s]	Max [s]	Relative
Cairo-vm (Rust, Cairo 1)	5.043 ± 0.036	4.999	5.101	2.91 ± 0.04
cairo-native (embedded AOT)	1.732 ± 0.022	1.697	1.768	1.00
cairo-native (embedded JIT using LLVM's ORC Engine)	1.755 ± 0.023	1.712	1.800	1.01 ± 0.02

Benchmark for program linear_search

Open benchmarks

Command	Mean [ms]	Min [ms]	Max [ms]	Relative
Cairo-vm (Rust, Cairo 1)	622.0 ± 12.3	605.9	648.2	1.00
cairo-native (embedded AOT)	1776.8 ± 30.8	1736.2	1842.6	2.86 ± 0.07
cairo-native (embedded JIT using LLVM's ORC Engine)	1802.9 ± 23.4	1774.0	1853.7	2.90 ± 0.07

Benchmark for program logistic_map

Open benchmarks

Command	Mean [ms]	Min [ms]	Max [ms]	Relative
Cairo-vm (Rust, Cairo 1)	519.7 ± 17.3	492.0	545.7	1.00
cairo-native (embedded AOT)	1741.1 ± 14.3	1715.0	1763.5	3.35 ± 0.11
cairo-native (embedded JIT using LLVM's ORC Engine)	1748.2 ± 34.3	1694.5	1805.8	3.36 ± 0.13