aiur: verify u32 addition from hinted sum - #556
Merged
Conversation
Codegen regenerated; all 71 kernel FFT pins and the shard-pipeline aggregate re-measured: every pin drops (median -3.7%, up to -7.9% on the string/byte-heavy checks), shard aggregate -9.9%. aiur (cargo + prove suite incl. parity), ixvm, multi-stark and recursive-verifier suites pass; fmt, clippy, deny clean.
This comment was marked as outdated.
This comment was marked as outdated.
This comment was marked as outdated.
This comment was marked as outdated.
Codegen regenerated; all 71 kernel FFT pins and the shard-pipeline aggregate re-measured: every pin drops again (median -1.5%, up to -3.4% on the string/byte-heavy checks), shard aggregate -4.3% on top of the hinted-adder commit (-13.8% combined). aiur (cargo + prove suite incl. parity), ixvm, multi-stark and recursive-verifier suites pass; fmt, clippy, deny clean.
This comment was marked as outdated.
This comment was marked as outdated.
This comment was marked as outdated.
This comment was marked as outdated.
gabriel-barrett
force-pushed
the
perf/u32-add-hint
branch
from
August 13, 2026 14:53
24d3175 to
52f9ff7
Compare
Member
Author
|
!benchmark aiur-recursive fresh |
|
| constant | recursive-prove-time (main) | recursive-prove-time (PR) | Δ% | recursive-peak-ram (main) | recursive-peak-ram (PR) | Δ% | recursive-proof-size (main) | recursive-proof-size (PR) | Δ% | recursive-verify-time (main) | recursive-verify-time (PR) | Δ% | recursive-execute-time (main) | recursive-execute-time (PR) | Δ% | recursive-fft-cost (main) | recursive-fft-cost (PR) | Δ% | prove-time (main) | prove-time (PR) | Δ% | proof-size (main) | proof-size (PR) | Δ% | verify-time (main) | verify-time (PR) | Δ% | peak-ram (main) | peak-ram (PR) | Δ% |
|---|---|---|---|---|---|---|---|---|---|---|---|---|---|---|---|---|---|---|---|---|---|---|---|---|---|---|---|---|---|---|
Nat.add_comm |
26.994 s | 21.577 s | -20.1% (1.25× faster) 🟢 | 81.01 GiB | 58.98 GiB | -27.2% (1.37× smaller) 🟢 | 5.48 MiB | 5.30 MiB | -3.3% 🟢 | 33.2 ms | 37.0 ms | +11.3% (1.11× slower) |
4.203 s | 3.741 s | -11.0% (1.12× faster) 🟢 | 148.08B | 113.85B | -23.1% (1.30× fewer) 🟢 | 1.012 s | 1.054 s | +4.2% |
9.11 MiB | 8.93 MiB | -2.0% | 57.4 ms | 53.0 ms | -7.7% (1.08× faster) 🟢 | 4.09 GiB | 4.11 GiB | +0.5% |
Plus the trailing integration: BytecodeEval's new hint arms index the value map via getD (no Inhabited G instance is exposed for `!`), the u32_sum helpers and the codegen template emit clippy-clean casts (u128::from / masked try_from), codegen regenerated, and all 71 kernel FFT pins + the shard aggregate re-measured (median -0.4%, shard -1.3% on top of the fused adder). aiur (cargo + prove suite incl. parity), ixvm, multi-stark and recursive-verifier suites pass; fmt, clippy, deny clean.
gabriel-barrett
force-pushed
the
perf/u32-add-hint
branch
from
August 13, 2026 15:07
52f9ff7 to
7cff837
Compare
Member
Author
|
!benchmark aiur fresh |
|
| constant | prove-time (main) | prove-time (PR) | Δ% | throughput (const/s) (main) | throughput (const/s) (PR) | Δ% | peak-ram (main) | peak-ram (PR) | Δ% | execute-time (main) | execute-time (PR) | Δ% | verify-time (main) | verify-time (PR) | Δ% | proof-size (main) | proof-size (PR) | Δ% | fft-cost (main) | fft-cost (PR) | Δ% |
|---|---|---|---|---|---|---|---|---|---|---|---|---|---|---|---|---|---|---|---|---|---|
Array.extract_append |
39.061 s | 38.016 s | -2.7% | 39.480 | 40.560 | +2.7% | 77.01 GiB | 73.74 GiB | -4.3% 🟢 | 9.909 s | 9.661 s | -2.5% | 138.9 ms | 130.7 ms | -5.9% (1.06× faster) 🟢 | 21.93 MiB | 21.59 MiB | -1.6% | 141.66B | 136.02B | -4.0% 🟢 |
ByteArray.utf8DecodeChar?_utf8EncodeChar_append |
39.708 s | 37.083 s | -6.6% (1.07× faster) 🟢 | 67.770 | 72.570 | +7.1% (1.07× faster) 🟢 | 79.86 GiB | 73.12 GiB | -8.4% (1.09× smaller) 🟢 | 9.778 s | 9.594 s | -1.9% | 139.5 ms | 133.0 ms | -4.6% 🟢 | 22.03 MiB | 21.69 MiB | -1.6% | 151.82B | 138.80B | -8.6% (1.09× fewer) 🟢 |
Char.ofOrdinal_le_of_le |
30.010 s | 28.793 s | -4.1% 🟢 | 88.170 | 91.900 | +4.2% 🟢 | 61.91 GiB | 58.56 GiB | -5.4% (1.06× smaller) 🟢 | 6.711 s | 6.636 s | -1.1% | 135.9 ms | 132.5 ms | -2.5% | 21.98 MiB | 21.64 MiB | -1.6% | 108.99B | 99.42B | -8.8% (1.10× fewer) 🟢 |
Vector.extract_append._proof_2 |
21.876 s | 21.111 s | -3.5% 🟢 | 59.520 | 61.670 | +3.6% 🟢 | 41.37 GiB | 39.66 GiB | -4.1% 🟢 | 5.331 s | 5.330 s | -0.0% | 133.4 ms | 136.5 ms | +2.3% | 21.64 MiB | 21.30 MiB | -1.6% | 81.14B | 77.08B | -5.0% (1.05× fewer) 🟢 |
_private.Init.Data.Range.Polymorphic.SInt.0.Int64.instRxcHasSize_eq |
18.413 s | 17.092 s | -7.2% (1.08× faster) 🟢 | 98.410 | 106.010 | +7.7% (1.08× faster) 🟢 | 38.17 GiB | 34.79 GiB | -8.9% (1.10× smaller) 🟢 | 3.636 s | 3.522 s | -3.1% 🟢 | 135.3 ms | 141.9 ms | +4.9% |
21.82 MiB | 21.47 MiB | -1.6% | 60.56B | 54.85B | -9.4% (1.10× fewer) 🟢 |
String.split |
17.416 s | 16.172 s | -7.1% (1.08× faster) 🟢 | 101.340 | 109.140 | +7.7% (1.08× faster) 🟢 | 36.10 GiB | 32.70 GiB | -9.4% (1.10× smaller) 🟢 | 3.384 s | 3.307 s | -2.3% | 133.8 ms | 131.7 ms | -1.6% | 22.02 MiB | 21.68 MiB | -1.6% | 55.29B | 50.25B | -9.1% (1.10× fewer) 🟢 |
List.mergeSort |
12.612 s | 11.817 s | -6.3% (1.07× faster) 🟢 | 114.810 | 122.530 | +6.7% (1.07× faster) 🟢 | 25.42 GiB | 23.73 GiB | -6.6% (1.07× smaller) 🟢 | 2.475 s | 2.356 s | -4.8% (1.05× faster) 🟢 | 133.5 ms | 134.4 ms | +0.7% | 21.87 MiB | 21.53 MiB | -1.6% | 40.03B | 36.10B | -9.8% (1.11× fewer) 🟢 |
Vector.append |
4.123 s | 3.951 s | -4.2% 🟢 | 117.630 | 122.740 | +4.3% 🟢 | 7.48 GiB | 6.67 GiB | -10.8% (1.12× smaller) 🟢 | 689.6 ms | 668.1 ms | -3.1% 🟢 | 126.0 ms | 135.1 ms | +7.3% (1.07× slower) |
20.47 MiB | 20.13 MiB | -1.7% | 8.83B | 8.02B | -9.2% (1.10× fewer) 🟢 |
Nat.gcd_comm |
3.299 s | 3.200 s | -3.0% | 118.230 | 121.880 | +3.1% 🟢 | 5.81 GiB | 5.75 GiB | -1.1% | 531.7 ms | 531.4 ms | -0.0% | 123.2 ms | 124.4 ms | +1.0% | 20.24 MiB | 19.90 MiB | -1.7% | 5.93B | 5.25B | -11.5% (1.13× fewer) 🟢 |
String.append |
2.409 s | 2.390 s | -0.8% | 126.190 | 127.180 | +0.8% | 5.84 GiB | 4.64 GiB | -20.6% (1.26× smaller) 🟢 | 418.0 ms | 399.0 ms | -4.5% 🟢 | 117.2 ms | 123.1 ms | +5.1% (1.05× slower) |
19.51 MiB | 19.17 MiB | -1.8% | 3.27B | 2.90B | -11.1% (1.13× fewer) 🟢 |
Int.gcd |
2.014 s | 2.010 s | -0.2% | 103.280 | 103.480 | +0.2% | 4.53 GiB | 4.53 GiB | +0.0% | 349.8 ms | 349.7 ms | -0.0% | 125.3 ms | 117.4 ms | -6.3% (1.07× faster) 🟢 | 19.07 MiB | 18.73 MiB | -1.8% | 2.10B | 1.87B | -10.9% (1.12× fewer) 🟢 |
Nat.sub_le_of_le_add |
1.842 s | 1.818 s | -1.3% | 92.300 | 93.530 | +1.3% | 4.65 GiB | 4.61 GiB | -0.7% | 341.9 ms | 343.2 ms | +0.4% | 114.3 ms | 113.7 ms | -0.5% | 19.42 MiB | 19.08 MiB | -1.8% | 1.79B | 1.60B | -10.5% (1.12× fewer) 🟢 |
Nat.add_comm |
1.055 s | 1.044 s | -1.0% | 39.820 | 40.230 | +1.0% | 3.61 GiB | 3.84 GiB | +6.3% (1.06× larger) |
253.3 ms | 248.5 ms | -1.9% | 105.9 ms | 106.0 ms | +0.1% | 17.60 MiB | 17.25 MiB | -1.9% | 292.23M | 271.37M | -7.1% (1.08× fewer) 🟢 |
arthurpaulino
approved these changes
Aug 13, 2026
This file contains hidden or bidirectional Unicode text that may be interpreted or compiled differently than what appears below. To review, open the file in an editor that reveals hidden Unicode characters.
Learn more about bidirectional Unicode characters
Sign up for free
to join this conversation on GitHub.
Already have an account?
Sign in to comment
2 participants
Add this suggestion to a batch that can be applied as a single commit.This suggestion is invalid because no changes were made to the code.Suggestions cannot be applied while the pull request is closed.Suggestions cannot be applied while viewing a subset of changes.Only one suggestion per line can be applied in a batch.Add this suggestion to a batch that can be applied as a single commit.Applying suggestions on deleted lines is not supported.You must change the existing code in this line in order to create a valid suggestion.Outdated suggestions cannot be applied.This suggestion has been applied or marked resolved.Suggestions cannot be applied from pending reviews.Suggestions cannot be applied on multi-line comments.Suggestions cannot be applied while the pull request is queued to merge.Suggestion cannot be applied right now. Please check back later.
u32_addoptimizationInstead of defining
u32addition as a bunch ofu8addition, we can define it through the native addition:u32_add(x, y), forx: u32, y: u32is az: u32such thatx + y - zis either0or2^32, that is,such that
(x + y - z)/2^32is a boolean. We also extend that to addition on 3 numbers:u32_add3(x, y, z)is aw: u32such that(x + y + z - w)/2^32is either0,1, or2