perf: Avoid heap allocations in state, input buffers and byte hash path by vadorovsky · Pull Request #61 · Lightprotocol/light-poseidon

vadorovsky · 2026-09-20T08:36:00Z

See individual commits for details.

Current main:

poseidon_bn254_x5_1     time:   [24.091 µs 24.130 µs 24.180 µs]
poseidon_bn254_x5_2     time:   [34.334 µs 34.372 µs 34.420 µs]
poseidon_bn254_x5_3     time:   [46.463 µs 46.499 µs 46.544 µs]
poseidon_bn254_x5_4     time:   [65.584 µs 65.697 µs 65.855 µs]
poseidon_bn254_x5_5     time:   [84.879 µs 85.005 µs 85.172 µs]
poseidon_bn254_x5_6     time:   [113.84 µs 114.37 µs 114.97 µs]
poseidon_bn254_x5_7     time:   [144.51 µs 144.71 µs 144.95 µs]
poseidon_bn254_x5_8     time:   [176.00 µs 176.13 µs 176.28 µs]
poseidon_bn254_x5_9     time:   [205.66 µs 205.80 µs 205.95 µs]
poseidon_bn254_x5_10    time:   [272.40 µs 273.08 µs 273.98 µs]
poseidon_bn254_x5_11    time:   [298.14 µs 298.41 µs 298.69 µs]
poseidon_bn254_x5_12    time:   [378.33 µs 378.96 µs 379.70 µs]

This change + arrayvec migration:

poseidon_bn254_x5_1     time:   [22.893 µs 22.909 µs 22.925 µs]
                        change: [-5.0821% -4.9214% -4.7493%] (p = 0.00 < 0.05)
                        Performance has improved.
Found 7 outliers among 100 measurements (7.00%)
  4 (4.00%) high mild
  3 (3.00%) high severe

poseidon_bn254_x5_2     time:   [33.052 µs 33.081 µs 33.113 µs]
                        change: [-3.9722% -3.8519% -3.7366%] (p = 0.00 < 0.05)
                        Performance has improved.
Found 5 outliers among 100 measurements (5.00%)
  4 (4.00%) high mild
  1 (1.00%) high severe

poseidon_bn254_x5_3     time:   [45.248 µs 45.295 µs 45.355 µs]
                        change: [-2.6216% -2.5166% -2.4136%] (p = 0.00 < 0.05)
                        Performance has improved.
Found 10 outliers among 100 measurements (10.00%)
  4 (4.00%) high mild
  6 (6.00%) high severe

poseidon_bn254_x5_4     time:   [64.118 µs 64.156 µs 64.196 µs]
                        change: [-2.4788% -2.2897% -2.1257%] (p = 0.00 < 0.05)
                        Performance has improved.
Found 4 outliers among 100 measurements (4.00%)
  3 (3.00%) high mild
  1 (1.00%) high severe

poseidon_bn254_x5_5     time:   [83.835 µs 83.908 µs 83.999 µs]
                        change: [-1.3413% -1.1845% -1.0296%] (p = 0.00 < 0.05)
                        Performance has improved.
Found 7 outliers among 100 measurements (7.00%)
  5 (5.00%) high mild
  2 (2.00%) high severe

poseidon_bn254_x5_6     time:   [111.90 µs 111.99 µs 112.10 µs]
                        change: [-2.1066% -1.7171% -1.3534%] (p = 0.00 < 0.05)
                        Performance has improved.
Found 7 outliers among 100 measurements (7.00%)
  6 (6.00%) high mild
  1 (1.00%) high severe

poseidon_bn254_x5_7     time:   [143.53 µs 143.65 µs 143.79 µs]
                        change: [-0.9447% -0.7207% -0.5351%] (p = 0.00 < 0.05)
                        Change within noise threshold.
Found 2 outliers among 100 measurements (2.00%)
  1 (1.00%) high mild
  1 (1.00%) high severe

poseidon_bn254_x5_8     time:   [175.51 µs 175.66 µs 175.83 µs]
                        change: [-0.3432% -0.1513% +0.1045%] (p = 0.26 > 0.05)
                        No change in performance detected.
Found 6 outliers among 100 measurements (6.00%)
  3 (3.00%) high mild
  3 (3.00%) high severe

poseidon_bn254_x5_9     time:   [207.51 µs 207.86 µs 208.27 µs]
                        change: [+0.9627% +1.1469% +1.3456%] (p = 0.00 < 0.05)
                        Change within noise threshold.
Found 4 outliers among 100 measurements (4.00%)
  2 (2.00%) high mild
  2 (2.00%) high severe

poseidon_bn254_x5_10    time:   [272.48 µs 272.89 µs 273.35 µs]
                        change: [-1.3056% -0.5531% +0.0423%] (p = 0.12 > 0.05)
                        No change in performance detected.
Found 7 outliers among 100 measurements (7.00%)
  5 (5.00%) high mild
  2 (2.00%) high severe

poseidon_bn254_x5_11    time:   [297.25 µs 297.69 µs 298.21 µs]
                        change: [-0.3990% -0.2486% -0.0819%] (p = 0.00 < 0.05)
                        Change within noise threshold.
Found 6 outliers among 100 measurements (6.00%)
  1 (1.00%) high mild
  5 (5.00%) high severe

poseidon_bn254_x5_12    time:   [376.93 µs 377.70 µs 378.61 µs]
                        change: [-0.6526% -0.3721% -0.0959%] (p = 0.01 < 0.05)
                        Change within noise threshold.
Found 9 outliers among 100 measurements (9.00%)
  1 (1.00%) high mild
  8 (8.00%) high severe

The Poseidon state holds at most `MAX_X5_LEN` (13) field elements and the number of inputs is at most 12, so neither buffer needs dynamic allocation. `Poseidon.state` is now an `ArrayVec<F, MAX_X5_LEN>`. The per-round `collect()` in `apply_mds` no longer mallocs and frees a fresh `Vec` (one allocation pair per round, 64-74 per hash) - the replacement state is built on the stack and moved into place. `hash_bytes_be`/`hash_bytes_le` deserialize inputs into a single stack-resident `ArrayVec<F, MAX_INPUTS>` instead of building two intermediate `Vec`s (validation results and converted elements). This adds the `arrayvec` dependency and a public `MAX_INPUTS` constant. Two incidental refactors: the input-length check is extracted into `validate_inputs_length`, and `validate_bytes_length` now returns `Result<(), PoseidonError>` instead of `Result<&[u8], PoseidonError>` - the identity return was only used by the old map/collect pattern. Note that this last part is a public API change.

The byte-input path still performed small heap allocations per call. `bytes_to_prime_field_element_be/le` built a `num_bigint::BigUint` (a heap `Vec<u32>`) for every input. The value is now assembled directly in `F::BigInt` (a fixed-size stack array of limbs) from the bytes, most significant byte first, and checked with `F::from_bigint`, which performs the modulus check in ark-ff 0.5. `hash_bytes_be/le` converted the result via `to_bytes_be/le`, which return a `Vec<u8>`. A new `IntoHashBytes` extension trait for `PrimeField` serializes the result into the fixed-size output array directly from its limbs instead. This removes the `num-bigint` dependency. Edge-case behavior is preserved: short inputs and oversized inputs with leading zero bytes are still accepted, `BytesToBigInt` is still returned for values that do not fit in `F::BigInt`, and `InputLargerThanModulus` for values >= the modulus. Also drops a needless borrow in `hash` flagged by clippy. Benchmarked on AMD EPYC 7H12 64-core with the following results. Current `main`: ``` poseidon_bn254_x5_1 time: [24.091 µs 24.130 µs 24.180 µs] poseidon_bn254_x5_2 time: [34.334 µs 34.372 µs 34.420 µs] poseidon_bn254_x5_3 time: [46.463 µs 46.499 µs 46.544 µs] poseidon_bn254_x5_4 time: [65.584 µs 65.697 µs 65.855 µs] poseidon_bn254_x5_5 time: [84.879 µs 85.005 µs 85.172 µs] poseidon_bn254_x5_6 time: [113.84 µs 114.37 µs 114.97 µs] poseidon_bn254_x5_7 time: [144.51 µs 144.71 µs 144.95 µs] poseidon_bn254_x5_8 time: [176.00 µs 176.13 µs 176.28 µs] poseidon_bn254_x5_9 time: [205.66 µs 205.80 µs 205.95 µs] poseidon_bn254_x5_10 time: [272.40 µs 273.08 µs 273.98 µs] poseidon_bn254_x5_11 time: [298.14 µs 298.41 µs 298.69 µs] poseidon_bn254_x5_12 time: [378.33 µs 378.96 µs 379.70 µs] ``` This change + arrayvec migration: ``` poseidon_bn254_x5_1 time: [22.893 µs 22.909 µs 22.925 µs] change: [-5.0821% -4.9214% -4.7493%] (p = 0.00 < 0.05) Performance has improved. Found 7 outliers among 100 measurements (7.00%) 4 (4.00%) high mild 3 (3.00%) high severe poseidon_bn254_x5_2 time: [33.052 µs 33.081 µs 33.113 µs] change: [-3.9722% -3.8519% -3.7366%] (p = 0.00 < 0.05) Performance has improved. Found 5 outliers among 100 measurements (5.00%) 4 (4.00%) high mild 1 (1.00%) high severe poseidon_bn254_x5_3 time: [45.248 µs 45.295 µs 45.355 µs] change: [-2.6216% -2.5166% -2.4136%] (p = 0.00 < 0.05) Performance has improved. Found 10 outliers among 100 measurements (10.00%) 4 (4.00%) high mild 6 (6.00%) high severe poseidon_bn254_x5_4 time: [64.118 µs 64.156 µs 64.196 µs] change: [-2.4788% -2.2897% -2.1257%] (p = 0.00 < 0.05) Performance has improved. Found 4 outliers among 100 measurements (4.00%) 3 (3.00%) high mild 1 (1.00%) high severe poseidon_bn254_x5_5 time: [83.835 µs 83.908 µs 83.999 µs] change: [-1.3413% -1.1845% -1.0296%] (p = 0.00 < 0.05) Performance has improved. Found 7 outliers among 100 measurements (7.00%) 5 (5.00%) high mild 2 (2.00%) high severe poseidon_bn254_x5_6 time: [111.90 µs 111.99 µs 112.10 µs] change: [-2.1066% -1.7171% -1.3534%] (p = 0.00 < 0.05) Performance has improved. Found 7 outliers among 100 measurements (7.00%) 6 (6.00%) high mild 1 (1.00%) high severe poseidon_bn254_x5_7 time: [143.53 µs 143.65 µs 143.79 µs] change: [-0.9447% -0.7207% -0.5351%] (p = 0.00 < 0.05) Change within noise threshold. Found 2 outliers among 100 measurements (2.00%) 1 (1.00%) high mild 1 (1.00%) high severe poseidon_bn254_x5_8 time: [175.51 µs 175.66 µs 175.83 µs] change: [-0.3432% -0.1513% +0.1045%] (p = 0.26 > 0.05) No change in performance detected. Found 6 outliers among 100 measurements (6.00%) 3 (3.00%) high mild 3 (3.00%) high severe poseidon_bn254_x5_9 time: [207.51 µs 207.86 µs 208.27 µs] change: [+0.9627% +1.1469% +1.3456%] (p = 0.00 < 0.05) Change within noise threshold. Found 4 outliers among 100 measurements (4.00%) 2 (2.00%) high mild 2 (2.00%) high severe poseidon_bn254_x5_10 time: [272.48 µs 272.89 µs 273.35 µs] change: [-1.3056% -0.5531% +0.0423%] (p = 0.12 > 0.05) No change in performance detected. Found 7 outliers among 100 measurements (7.00%) 5 (5.00%) high mild 2 (2.00%) high severe poseidon_bn254_x5_11 time: [297.25 µs 297.69 µs 298.21 µs] change: [-0.3990% -0.2486% -0.0819%] (p = 0.00 < 0.05) Change within noise threshold. Found 6 outliers among 100 measurements (6.00%) 1 (1.00%) high mild 5 (5.00%) high severe poseidon_bn254_x5_12 time: [376.93 µs 377.70 µs 378.61 µs] change: [-0.6526% -0.3721% -0.0959%] (p = 0.01 < 0.05) Change within noise threshold. Found 9 outliers among 100 measurements (9.00%) 1 (1.00%) high mild 8 (8.00%) high severe ```

ananas-block · 2026-09-22T13:47:17Z

Both of these come from the same root cause: the Circom-specific constants became universal capacities.

1. hash_bytes_* panics on too many inputs. The macro fills ArrayVec<F, MAX_INPUTS> (12) before hash validates the count, so 13 valid 32-byte inputs hit CapacityError instead of InvalidNumberOfInputs. The existing test loops 1..13, one short of the first overflow. Validate the count before filling the buffer.

2. Poseidon::new loses widths above 13. state: ArrayVec<F, MAX_X5_LEN> makes the Circom x5 bound the capacity for every hasher; on main it only gated new_circom, and state capacity came from params.width. Custom parameters with width = 14 pass the input check and then panic on the last push. Either keep generic widths, or make the capacity explicit and the constructor fallible.

Generated with Claude Code

vadorovsky · 2026-09-22T15:56:49Z

Good catch, quite embarrassing from my side. We can focus on landing your PR.

vadorovsky force-pushed the vad/arrayvec branch from 42762d0 to e039a09 Compare September 21, 2026 07:03

vadorovsky marked this pull request as ready for review September 21, 2026 07:28

vadorovsky added 2 commits September 22, 2026 15:20

vadorovsky force-pushed the vad/arrayvec branch from e039a09 to fb624a6 Compare September 22, 2026 13:20

ananas-block mentioned this pull request Sep 22, 2026

perf: zero-heap-allocation Poseidon hash path #64

Draft

Navigation Menu

perf: Avoid heap allocations in state, input buffers and byte hash path - #61

perf: Avoid heap allocations in state, input buffers and byte hash path#61
vadorovsky wants to merge 2 commits into
Lightprotocol:mainfrom
vadorovsky:vad/arrayvec

vadorovsky commented Sep 20, 2026 •

edited

Loading

Uh oh!

ananas-block commented Sep 22, 2026

Uh oh!

vadorovsky commented Sep 22, 2026

Uh oh!

Reviewers

Assignees

Labels

Projects

Milestone

Development

Uh oh!

2 participants

Conversation

vadorovsky commented Sep 20, 2026 • edited Loading Uh oh! There was an error while loading. Please reload this page.

Uh oh!

Uh oh!

ananas-block commented Sep 22, 2026

Uh oh!

vadorovsky commented Sep 22, 2026

Uh oh!

Reviewers

Assignees

Labels

Projects

Milestone

Development

Uh oh!

2 participants

vadorovsky commented Sep 20, 2026 •

edited

Loading