fix(values): encode BoundedInt in its compact representation; deref boxed returns by TomerStarkware · Pull Request #1657 · starkware-libs/cairo_native

TomerStarkware · 2026-08-18T07:40:37Z

Stacked on #1655 and the nested-Box/Nullable PR.

Problem

Two remaining marshaling asymmetries:

BoundedInt encode: to_ptr wrote the un-biased value into a 32-byte felt slot, while the native representation (and what from_ptr decodes) is the compact one — value - lower stored in get_integer_layout(repr_bit_width()). Nested in an aggregate (top-level args are still gated by the arch.rs panic, implement AbiArgument::to_ptr for Value::BoundedInt #1217), it was both silently mis-decoded and overran its slot: Array<BoundedInt<3, 10>> has a 1-byte stride and got 32 bytes written per element. The existing unit test test_to_ptr_bounded_int_valid asserted the buggy encoding.
Boxed returns: parse_result passed a returned Box<T>'s return-pointer slot straight to from_ptr without dereferencing it. Any function using a non-ZST builtin returns through a return pointer, so e.g. fn(...) -> Box<felt252> using pedersen decoded two raw heap addresses as the payload.

Fix

Add RangeExt::repr_encode / repr_decode as the single implementation of the compact encoding (felt-wrapping subtraction so negative lower bounds round-trip). to_ptr uses the encode side — allocating the actual repr_bit_width layout and validating against the type's range — and both existing decode duplicates (Value::from_ptr and the register path in parse_result) now share the decode side.
parse_result's Box arm dereferences the return-pointer slot once, mirroring the Nullable arm.
The arch.rs BoundedInt panic (implement AbiArgument::to_ptr for Value::BoundedInt #1217) stays, with a TODO noting it can now be implemented on top of repr_encode.

Tests

VM-vs-native: boxed return forced through the return pointer (pedersen builtin).
Round-trip unit tests: Array<BoundedInt<3, 10>> asserting the 1-byte biased element buffer ([0, 4, 6] for values [3, 7, 9]), and a negative-lower range (BoundedInt<-5, 5>).
test_to_ptr_bounded_int_valid corrected: value 16 in BoundedInt<10, 510> encodes as [6, 0] in 2 bytes (and its embedded range fixed to the parsed [10, 511)).

🤖 Generated with Claude Code

This change is

github-actions · 2026-08-18T07:41:03Z

✅ Code is now correctly formatted.

orizi

@orizi reviewed 6 files and all commit messages, and made 1 comment.
Reviewable status: complete! all files reviewed, all discussions resolved (waiting on TomerStarkware).

…oxed returns Two remaining marshaling asymmetries: - `Value::to_ptr` wrote a BoundedInt as the un-biased value in a 32-byte felt slot, while the native representation (and `from_ptr`) is the compact one: `value - lower` stored in `repr_bit_width()` bits. A BoundedInt nested in an aggregate (top-level arguments are still gated by the arch.rs panic, #1217) was both silently mis-decoded and overran its slot, corrupting neighboring elements — `Array<BoundedInt<3, 10>>` returned garbage. - `parse_result` passed a returned `Box<T>`'s return-pointer slot straight to `from_ptr` without dereferencing it. Any function that uses a non-ZST builtin returns through a return pointer, so e.g. `fn(...) -> Box<felt252>` using pedersen decoded two raw heap addresses as the payload. Mirror the Nullable arm's deref. Add `RangeExt::repr_encode`/`repr_decode` as the single implementation of the compact encoding (felt-wrapping subtraction so negative lower bounds round-trip) and use the decode side from both existing duplicates (`Value::from_ptr` and the register path in `parse_result`). The `to_ptr` arm now also validates against the type's range rather than only the value's embedded one. Fixes `test_to_ptr_bounded_int_valid`, which asserted the buggy 32-byte encoding ([16, ...] instead of `value - lower` = [6, 0] in the 2-byte 9-bit representation), and its embedded range, which didn't match the type's (`BoundedInt<10, 510>` parses to the range `[10, 511)`). Adds a VM-vs-native test for a boxed return forced through the return pointer, and round-trip unit tests for `Array<BoundedInt<3, 10>>` (asserting the 1-byte biased element buffer) and a negative-lower range. Co-Authored-By: Claude Fable 5 <noreply@anthropic.com>

github-actions · 2026-08-23T12:43:23Z

Benchmarking results

Benchmark for program `dict_insert`

Open benchmarks

Command	Mean [s]	Min [s]	Max [s]	Relative
`Cairo-vm (Rust, Cairo 1)`	5.767 ± 0.053	5.695	5.888	5.56 ± 0.06
`cairo-native (embedded AOT)`	1.037 ± 0.006	1.030	1.047	1.00
`cairo-native (embedded JIT using LLVM's ORC Engine)`	1.044 ± 0.005	1.039	1.052	1.01 ± 0.01

Benchmark for program `dict_snapshot`

Open benchmarks

Command	Mean [ms]	Min [ms]	Max [ms]	Relative
`Cairo-vm (Rust, Cairo 1)`	306.2 ± 1.8	302.5	308.4	1.00
`cairo-native (embedded AOT)`	929.8 ± 8.4	918.9	942.5	3.04 ± 0.03
`cairo-native (embedded JIT using LLVM's ORC Engine)`	946.7 ± 3.3	942.8	951.8	3.09 ± 0.02

Benchmark for program `factorial_2M`

Open benchmarks

Command	Mean [s]	Min [s]	Max [s]	Relative
`Cairo-vm (Rust, Cairo 1)`	2.499 ± 0.038	2.465	2.584	2.60 ± 0.04
`cairo-native (embedded AOT)`	0.963 ± 0.004	0.957	0.970	1.00
`cairo-native (embedded JIT using LLVM's ORC Engine)`	0.973 ± 0.003	0.968	0.976	1.01 ± 0.01

Benchmark for program `fib_2M`

Open benchmarks

Command	Mean [s]	Min [s]	Max [s]	Relative
`Cairo-vm (Rust, Cairo 1)`	2.419 ± 0.029	2.386	2.484	2.59 ± 0.04
`cairo-native (embedded AOT)`	0.933 ± 0.009	0.924	0.957	1.00
`cairo-native (embedded JIT using LLVM's ORC Engine)`	0.945 ± 0.003	0.940	0.950	1.01 ± 0.01

Benchmark for program `linear_search`

Open benchmarks

Command	Mean [ms]	Min [ms]	Max [ms]	Relative
`Cairo-vm (Rust, Cairo 1)`	332.4 ± 3.9	326.4	339.0	1.00
`cairo-native (embedded AOT)`	945.5 ± 2.9	942.4	952.5	2.84 ± 0.03
`cairo-native (embedded JIT using LLVM's ORC Engine)`	960.7 ± 4.4	954.8	966.7	2.89 ± 0.04

Benchmark for program `logistic_map`

Open benchmarks

Command	Mean [ms]	Min [ms]	Max [ms]	Relative
`Cairo-vm (Rust, Cairo 1)`	282.3 ± 1.5	279.8	284.2	1.00
`cairo-native (embedded AOT)`	932.9 ± 4.9	924.4	939.8	3.30 ± 0.02
`cairo-native (embedded JIT using LLVM's ORC Engine)`	951.4 ± 3.4	948.0	957.2	3.37 ± 0.02

github-actions · 2026-08-23T12:58:15Z

Benchmark results Main vs HEAD.

Base

Command	Mean [s]	Min [s]	Max [s]	Relative
`base dict_insert.cairo (JIT)`	1.827 ± 0.017	1.808	1.853	1.01 ± 0.02
`base dict_insert.cairo (AOT)`	1.812 ± 0.027	1.769	1.847	1.00

Head

Command	Mean [s]	Min [s]	Max [s]	Relative
`head dict_insert.cairo (JIT)`	1.879 ± 0.023	1.846	1.914	1.01 ± 0.01
`head dict_insert.cairo (AOT)`	1.852 ± 0.015	1.833	1.885	1.00

Base

Command	Mean [s]	Min [s]	Max [s]	Relative
`base dict_snapshot.cairo (JIT)`	1.637 ± 0.018	1.615	1.670	1.03 ± 0.01
`base dict_snapshot.cairo (AOT)`	1.588 ± 0.008	1.575	1.602	1.00

Head

Command	Mean [s]	Min [s]	Max [s]	Relative
`head dict_snapshot.cairo (JIT)`	1.709 ± 0.019	1.683	1.739	1.01 ± 0.01
`head dict_snapshot.cairo (AOT)`	1.686 ± 0.016	1.657	1.718	1.00

Base

Command	Mean [s]	Min [s]	Max [s]	Relative
`base factorial_2M.cairo (JIT)`	1.680 ± 0.013	1.658	1.706	1.01 ± 0.01
`base factorial_2M.cairo (AOT)`	1.658 ± 0.009	1.643	1.674	1.00

Head

Command	Mean [s]	Min [s]	Max [s]	Relative
`head factorial_2M.cairo (JIT)`	1.768 ± 0.007	1.758	1.778	1.01 ± 0.01
`head factorial_2M.cairo (AOT)`	1.752 ± 0.018	1.721	1.775	1.00

Base

Command	Mean [s]	Min [s]	Max [s]	Relative
`base fib_2M.cairo (JIT)`	1.626 ± 0.008	1.610	1.637	1.01 ± 0.01
`base fib_2M.cairo (AOT)`	1.614 ± 0.017	1.593	1.641	1.00

Head

Command	Mean [s]	Min [s]	Max [s]	Relative
`head fib_2M.cairo (JIT)`	1.757 ± 0.014	1.733	1.778	1.01 ± 0.01
`head fib_2M.cairo (AOT)`	1.733 ± 0.010	1.711	1.743	1.00

Base

Command	Mean [s]	Min [s]	Max [s]	Relative
`base linear_search.cairo (JIT)`	1.655 ± 0.009	1.641	1.673	1.00 ± 0.01
`base linear_search.cairo (AOT)`	1.654 ± 0.020	1.627	1.683	1.00

Head

Command	Mean [s]	Min [s]	Max [s]	Relative
`head linear_search.cairo (JIT)`	1.776 ± 0.017	1.752	1.811	1.03 ± 0.01
`head linear_search.cairo (AOT)`	1.731 ± 0.012	1.708	1.742	1.00

Base

Command	Mean [s]	Min [s]	Max [s]	Relative
`base logistic_map.cairo (JIT)`	1.668 ± 0.028	1.633	1.712	1.01 ± 0.02
`base logistic_map.cairo (AOT)`	1.645 ± 0.020	1.621	1.686	1.00

Head

Command	Mean [s]	Min [s]	Max [s]	Relative
`head logistic_map.cairo (JIT)`	1.752 ± 0.017	1.726	1.792	1.04 ± 0.01
`head logistic_map.cairo (AOT)`	1.685 ± 0.011	1.669	1.701	1.00

TomerStarkware force-pushed the tomer/values_nested_box_nullable branch from 9bed290 to c09a4f1 Compare August 18, 2026 09:43

TomerStarkware force-pushed the tomer/values_bounded_int_box_return branch from bff9388 to 44e5111 Compare August 18, 2026 09:43

TomerStarkware force-pushed the tomer/values_nested_box_nullable branch from c09a4f1 to b7dd582 Compare August 18, 2026 15:39

TomerStarkware force-pushed the tomer/values_bounded_int_box_return branch 2 times, most recently from 86ba869 to cc5f00f Compare August 19, 2026 16:07

TomerStarkware force-pushed the tomer/values_nested_box_nullable branch from b7dd582 to 3d5c68c Compare August 19, 2026 16:07

TomerStarkware requested a review from orizi August 20, 2026 12:28

orizi approved these changes Aug 20, 2026

View reviewed changes

TomerStarkware force-pushed the tomer/values_nested_box_nullable branch from 3d5c68c to e42ae5e Compare August 23, 2026 10:33

Base automatically changed from tomer/values_nested_box_nullable to main August 23, 2026 11:54

TomerStarkware force-pushed the tomer/values_bounded_int_box_return branch from cc5f00f to e9f2176 Compare August 23, 2026 12:21

TomerStarkware added this pull request to the merge queue Aug 23, 2026

Merged via the queue into main with commit 58ddd21 Aug 23, 2026
15 checks passed

TomerStarkware deleted the tomer/values_bounded_int_box_return branch August 23, 2026 13:46

Navigation Menu

Uh oh!

fix(values): encode BoundedInt in its compact representation; deref boxed returns - #1657

fix(values): encode BoundedInt in its compact representation; deref boxed returns#1657
TomerStarkware merged 1 commit into
mainfrom
tomer/values_bounded_int_box_return

TomerStarkware commented Aug 18, 2026 •

edited by orizi

Loading

Uh oh!

github-actions Bot commented Aug 18, 2026 •

edited

Loading

Uh oh!

orizi left a comment

Uh oh!

github-actions Bot commented Aug 23, 2026

Uh oh!

github-actions Bot commented Aug 23, 2026

Uh oh!

Uh oh!

Reviewers

Assignees

Labels

Projects

Milestone

Development

Uh oh!

2 participants

Uh oh!

Conversation

TomerStarkware commented Aug 18, 2026 • edited by orizi Loading Uh oh! There was an error while loading. Please reload this page.

Uh oh!

Problem

Fix

Tests

Uh oh!

github-actions Bot commented Aug 18, 2026 • edited Loading Uh oh! There was an error while loading. Please reload this page.

Uh oh!

Uh oh!

orizi left a comment

Choose a reason for hiding this comment

Uh oh!

github-actions Bot commented Aug 23, 2026

Benchmarking results

Benchmark for program dict_insert

Benchmark for program dict_snapshot

Benchmark for program factorial_2M

Benchmark for program fib_2M

Benchmark for program linear_search

Benchmark for program logistic_map

Uh oh!

github-actions Bot commented Aug 23, 2026

Benchmark results Main vs HEAD.

Base

Head

Base

Head

Base

Head

Base

Head

Base

Head

Base

Head

Uh oh!

Uh oh!

Reviewers

Assignees

Labels

Projects

Milestone

Development

Uh oh!

2 participants

TomerStarkware commented Aug 18, 2026 •

edited by orizi

Loading

github-actions Bot commented Aug 18, 2026 •

edited

Loading

Benchmark for program `dict_insert`

Benchmark for program `dict_snapshot`

Benchmark for program `factorial_2M`

Benchmark for program `fib_2M`

Benchmark for program `linear_search`

Benchmark for program `logistic_map`