Update multi-stark for hybrid CUDA scheduling - #612
Draft
arthurpaulino wants to merge 1 commit into
Draft
Conversation
arthurpaulino
force-pushed
the
ap/gpu
branch
from
September 4, 2026 19:51
2581c11 to
1de9691
Compare
Member
Author
|
!benchmark fresh |
Contributor
|
| constant | execute-time (main) | execute-time (PR) | Δ% | prove-time (main) | prove-time (PR) | Δ% | throughput (const/s) (main) | throughput (const/s) (PR) | Δ% | peak-ram (main) | peak-ram (PR) | Δ% | proof-size (main) | proof-size (PR) | Δ% | verify-time (main) | verify-time (PR) | Δ% | fft-cost (main) | fft-cost (PR) | Δ% |
|---|---|---|---|---|---|---|---|---|---|---|---|---|---|---|---|---|---|---|---|---|---|
ByteArray.utf8DecodeChar?_utf8EncodeChar_append |
8.813 s | 8.852 s | +0.4% | 29.304 s | 29.377 s | +0.2% | 94.700 | 94.460 | -0.3% | 72.85 GiB | 72.83 GiB | -0.0% | 11.11 MiB | 11.11 MiB | +0.0% | 63.2 ms | 62.9 ms | -0.5% | 136.01B | 136.01B | +0.0% |
Char.ofOrdinal_le_of_le |
6.844 s | 6.823 s | -0.3% | 24.993 s | 25.849 s | +3.4% |
110.550 | 106.890 | -3.3% |
65.85 GiB | 65.91 GiB | +0.1% | 11.11 MiB | 11.11 MiB | +0.0% | 62.8 ms | 65.5 ms | +4.2% |
104.21B | 104.21B | +0.0% |
Array.extract_append |
6.379 s | 6.467 s | +1.4% | 22.227 s | 22.199 s | -0.1% | 72.260 | 72.350 | +0.1% | 53.03 GiB | 52.99 GiB | -0.1% | 11.02 MiB | 11.02 MiB | +0.0% | 58.6 ms | 56.6 ms | -3.5% 🟢 | 98.00B | 98.00B | +0.0% |
_private.Init.Data.Range.Polymorphic.SInt.0.Int64.instRxcHasSize_eq |
3.516 s | 3.568 s | +1.5% | 13.822 s | 13.891 s | +0.5% | 135.080 | 134.400 | -0.5% | 34.99 GiB | 35.03 GiB | +0.1% | 11.05 MiB | 11.05 MiB | +0.0% | 56.3 ms | 56.9 ms | +1.1% | 56.49B | 56.49B | +0.0% |
Std.HashMap |
4.066 s | 3.984 s | -2.0% | 15.121 s | 14.994 s | -0.8% | 135.040 | 136.190 | +0.9% | 37.37 GiB | 37.35 GiB | -0.1% | 11.06 MiB | 11.06 MiB | +0.0% | 61.1 ms | 61.7 ms | +1.0% | 62.93B | 62.93B | +0.0% |
String.append |
429.0 ms | 449.7 ms | +4.8% |
2.139 s | 2.113 s | -1.2% | 152.910 | 154.730 | +1.2% | 4.96 GiB | 4.95 GiB | -0.1% | 9.78 MiB | 9.78 MiB | +0.0% | 50.9 ms | 50.3 ms | -1.2% | 3.45B | 3.45B | +0.0% |
Nat.add_comm |
262.1 ms | 263.5 ms | +0.5% | 973.2 ms | 983.4 ms | +1.0% | 47.260 | 46.780 | -1.0% | 4.67 GiB | 4.50 GiB | -3.7% 🟢 | 8.95 MiB | 8.95 MiB | +0.0% | 42.4 ms | 42.9 ms | +1.2% | 314.33M | 314.33M | +0.0% |
FRI verifier on FRI (7 constants)
| constant | execute-time (main) | execute-time (PR) | Δ% | prove-time (main) | prove-time (PR) | Δ% | throughput (const/s) (main) | throughput (const/s) (PR) | Δ% | peak-ram (main) | peak-ram (PR) | Δ% | proof-size (main) | proof-size (PR) | Δ% | verify-time (main) | verify-time (PR) | Δ% | fft-cost (main) | fft-cost (PR) | Δ% |
|---|---|---|---|---|---|---|---|---|---|---|---|---|---|---|---|---|---|---|---|---|---|
ByteArray.utf8DecodeChar?_utf8EncodeChar_append |
4.840 s | 4.826 s | -0.3% | 30.484 s | 29.998 s | -1.6% | 91.030 | 92.500 | +1.6% | 102.13 GiB | 101.68 GiB | -0.4% | 4.00 MiB | 4.00 MiB | +0.0% | 25.0 ms | 33.1 ms | +32.5% (1.32× slower) |
209.69B | 209.69B | +0.0% |
Char.ofOrdinal_le_of_le |
4.898 s | 4.945 s | +1.0% | 29.907 s | 29.900 s | -0.0% | 92.390 | 92.410 | +0.0% | 101.72 GiB | 101.65 GiB | -0.1% | 4.00 MiB | 4.00 MiB | +0.0% | 27.6 ms | 24.7 ms | -10.7% (1.12× faster) 🟢 | 208.66B | 208.66B | +0.0% |
Array.extract_append |
4.631 s | 4.642 s | +0.2% | 27.465 s | 27.613 s | +0.5% | 58.480 | 58.160 | -0.5% | 91.76 GiB | 91.81 GiB | +0.1% | 4.00 MiB | 4.00 MiB | +0.0% | 23.1 ms | 23.3 ms | +0.9% | 198.51B | 198.51B | +0.0% |
_private.Init.Data.Range.Polymorphic.SInt.0.Int64.instRxcHasSize_eq |
4.809 s | 4.889 s | +1.7% | 29.379 s | 30.735 s | +4.6% |
63.550 | 60.750 | -4.4% |
99.53 GiB | 99.49 GiB | -0.0% | 4.02 MiB | 4.02 MiB | +0.0% | 38.6 ms | 22.4 ms | -41.8% (1.72× faster) 🟢 | 203.56B | 203.56B | +0.0% |
Std.HashMap |
4.883 s | 4.786 s | -2.0% | 29.086 s | 28.963 s | -0.4% | 70.210 | 70.500 | +0.4% | 97.27 GiB | 97.34 GiB | +0.1% | 4.00 MiB | 4.00 MiB | +0.0% | 37.4 ms | 34.7 ms | -7.3% (1.08× faster) 🟢 | 208.71B | 208.71B | +0.0% |
String.append |
3.875 s | 3.858 s | -0.4% | 26.097 s | 26.392 s | +1.1% | 12.530 | 12.390 | -1.1% | 89.25 GiB | 89.35 GiB | +0.1% | 4.00 MiB | 4.00 MiB | +0.0% | 23.4 ms | 23.4 ms | +0.0% | 169.61B | 169.61B | +0.0% |
Nat.add_comm |
3.115 s | 3.092 s | -0.7% | 17.280 s | 17.329 s | +0.3% | 2.660 | 2.650 | -0.4% | 58.67 GiB | 58.66 GiB | -0.0% | 4.01 MiB | 4.01 MiB | +0.0% | 21.2 ms | 21.8 ms | +2.9% | 129.00B | 129.00B | +0.0% |
Aggregate flat join (7 constants)
| constant | execute-time (main) | execute-time (PR) | Δ% | prove-time (main) | prove-time (PR) | Δ% | peak-ram (main) | peak-ram (PR) | Δ% | proof-size (main) | proof-size (PR) | Δ% | verify-time (main) | verify-time (PR) | Δ% | fft-cost (main) | fft-cost (PR) | Δ% |
|---|---|---|---|---|---|---|---|---|---|---|---|---|---|---|---|---|---|---|
ByteArray.utf8DecodeChar?_utf8EncodeChar_append |
n/a | n/a | n/a | n/a | n/a | n/a | n/a | n/a | n/a | n/a | n/a | n/a | n/a | n/a | n/a | n/a | n/a | n/a |
Char.ofOrdinal_le_of_le |
n/a | n/a | n/a | n/a | n/a | n/a | n/a | n/a | n/a | n/a | n/a | n/a | n/a | n/a | n/a | n/a | n/a | n/a |
Array.extract_append |
n/a | n/a | n/a | n/a | n/a | n/a | n/a | n/a | n/a | n/a | n/a | n/a | n/a | n/a | n/a | n/a | n/a | n/a |
_private.Init.Data.Range.Polymorphic.SInt.0.Int64.instRxcHasSize_eq |
n/a | n/a | n/a | n/a | n/a | n/a | n/a | n/a | n/a | n/a | n/a | n/a | n/a | n/a | n/a | n/a | n/a | n/a |
Std.HashMap |
n/a | n/a | n/a | n/a | n/a | n/a | n/a | n/a | n/a | n/a | n/a | n/a | n/a | n/a | n/a | n/a | n/a | n/a |
String.append |
n/a | n/a | n/a | n/a | n/a | n/a | n/a | n/a | n/a | n/a | n/a | n/a | n/a | n/a | n/a | n/a | n/a | n/a |
Nat.add_comm |
n/a | n/a | n/a | n/a | n/a | n/a | n/a | n/a | n/a | n/a | n/a | n/a | n/a | n/a | n/a | n/a | n/a | n/a |
Pipeline total (7 constants)
| constant | total-time (main) | total-time (PR) | Δ% | pipeline-throughput (const/s) (main) | pipeline-throughput (const/s) (PR) | Δ% | pipeline-peak-ram (main) | pipeline-peak-ram (PR) | Δ% |
|---|---|---|---|---|---|---|---|---|---|
ByteArray.utf8DecodeChar?_utf8EncodeChar_append |
59.788 s | 59.375 s | -0.7% | 46.410 | 46.740 | +0.7% | 102.13 GiB | 101.68 GiB | -0.4% |
Char.ofOrdinal_le_of_le |
54.900 s | 55.749 s | +1.5% | 50.330 | 49.560 | -1.5% | 101.72 GiB | 101.65 GiB | -0.1% |
Array.extract_append |
49.691 s | 49.812 s | +0.2% | 32.320 | 32.240 | -0.2% | 91.76 GiB | 91.81 GiB | +0.1% |
_private.Init.Data.Range.Polymorphic.SInt.0.Int64.instRxcHasSize_eq |
43.201 s | 44.626 s | +3.3% |
43.220 | 41.840 | -3.2% |
99.53 GiB | 99.49 GiB | -0.0% |
Std.HashMap |
44.207 s | 43.957 s | -0.6% | 46.190 | 46.450 | +0.6% | 97.27 GiB | 97.34 GiB | +0.1% |
String.append |
28.236 s | 28.506 s | +1.0% | 11.580 | 11.470 | -0.9% | 89.25 GiB | 89.35 GiB | +0.1% |
Nat.add_comm |
18.254 s | 18.313 s | +0.3% | 2.520 | 2.510 | -0.4% | 58.67 GiB | 58.66 GiB | -0.0% |
Point Aiur at the multi-stark revision that balances large mixed-height commitments across host CPUs and CUDA according to device capacity. The backend carries hybrid placement through lookup, quotient, openings, and FRI while preserving the existing Goldilocks/BLAKE3 protocol and CPU verification path. No Ix source or build configuration changes are required: the existing opt-in cuda feature chain selects the new backend, while default builds remain CUDA-independent. Cargo.lock records the matching dependency graph. Validated with release clippy across the non-CUDA CI feature set, a release all-targets CUDA check, and IX_CUDA=1 lake build bench-typecheck, including the final Rust static-library link.
arthurpaulino
force-pushed
the
ap/gpu
branch
from
September 4, 2026 20:02
1de9691 to
280da3f
Compare
johnchandlerburnham
approved these changes
Sep 4, 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.
Point Aiur at the multi-stark revision that balances large mixed-height commitments across host CPUs and CUDA according to device capacity. The backend carries hybrid placement through lookup, quotient, openings, and FRI while preserving the existing Goldilocks/BLAKE3 protocol and CPU verification path.
No Ix source or build configuration changes are required: the existing opt-in cuda feature chain selects the new backend, while default builds remain CUDA-independent. Cargo.lock records the matching dependency graph.
Validated with release clippy across the non-CUDA CI feature set, a release all-targets CUDA check, and IX_CUDA=1 lake build bench-typecheck, including the final Rust static-library link.