feat: make ShaVmAir's degree 3#2422
Draft
Golovanov399 wants to merge 268 commits intodevelop-v2.0.0-betafrom
Draft
feat: make ShaVmAir's degree 3#2422Golovanov399 wants to merge 268 commits intodevelop-v2.0.0-betafrom
Golovanov399 wants to merge 268 commits intodevelop-v2.0.0-betafrom
Conversation
- replace `SegmentationLimits` with `SegmentationConfig` in `SystemConfig` - add a `interaction_cell_weight` parameter to `SegmentationConfig` that specifies how much cells does an interaction contribute at each row
Resolves INT-5611
For large metrics, mermaid is too much text. Also switched to outputting detailed metrics in a separate markdown file. In the benchmark CI, we still cat the detailed metrics back into the main markdown file. The svg chart is uploaded to public s3 similar to flamegraphs so they can be viewed from the markdown. For later: - I feel like we can switch to having the detailed metrics be stored in a sqlite file. That way it can be downloaded and processed more easily for complex metrics. For now I just split it into a separate markdown for simplicity.
closes INT-5391
segment_ctx.rs: - DEFAULT_MAX_CELLS → DEFAULT_MAX_MEMORY = 15gb - max_cells → max_memory in SegmentationLimits - set_max_cells → set_max_memory ctx.rs: - with_max_cells → with_max_memory metered_cost.rs: - Updated import to use DEFAULT_MAX_MEMORY cli/src/commands/prove.rs: - Updated import to DEFAULT_MAX_MEMORY - segment_max_cells → segment_max_memory - with_max_cells → with_max_memory benchmarks/prove/src/bin/async_regex.rs: - segment_max_cells → segment_max_memory - set_max_cells → set_max_memory benchmarks/prove/src/util.rs: - segment_max_cells → segment_max_memory - set_max_cells → set_max_memory --------- Co-authored-by: Copilot <175728472+Copilot@users.noreply.github.com>
Due to the difference in error types, switched to using `should_panic` for now. We should go through later and switch everything back to the precise error types. closes INT-5904
Resolves INT-5850.
Also shows delta for "Parallel Proof Time (N provers)" column
Resolves INT-5851.
Resolves INT-5853 and INT-5906.
…n] (#2366) Resolves INT-5852.
…and ECC (#2372) - **Replace expensive BigUint computation during preflight with fast native field arithmetic** (halo2curves/blstrs) for all known field types (K256, P256, BN254, BLS12-381) and ECC curve operations. The trace filler already re-executes with BigUint for constraint generation, so preflight only needs to compute outputs for memory writes. - **Cache modulus constants** with `once_cell::Lazy<BigUint>` to eliminate repeated hex string parsing in `get_field_type()`/`get_fp2_field_type()` and `get_curve_type()` (previously called on every instruction). - **Cache `FieldType`/`CurveType` on executor structs** at construction time, eliminating per-instruction BigUint comparisons in preflight. - **Remove `DynArray` heap allocations** in preflight by using stack-allocated typed arrays directly from adapter read/write, with `as_flattened()` for zero-cost conversions. - **Add `adapter()` accessor** to `FieldExpressionExecutor` for use by custom `PreflightExecutor` implementations. SETUP operations and unknown field types fall back to `run_field_expression_precomputed`. - [x] `cargo nextest run -p openvm-algebra-circuit` — all 18 non-pre-existing-failure tests pass (8 modular addsub/muldiv, 2 is_equal positive, 8 fp2_chip) - [x] `cargo nextest run -p openvm-ecc-circuit` — all 8 tests pass (3 add_ne, 5 double including nonzero_a) - [x] `cargo clippy -p openvm-algebra-circuit -p openvm-ecc-circuit --all-targets` — no new warnings 🤖 Generated with [Claude Code](https://claude.com/claude-code) --------- Co-authored-by: Claude Opus 4.5 <noreply@anthropic.com>
keccak (`p3_inner_tracegen`) - force `__noinline__` - get -92% stack size BigUintGPU::mod_div - force `__noinline__` - get -85% stack size sha256 (first and second pass) - -76% & -90%: - `generate_block_trace`, `generate_missing_cells` → `__noinline__` - `generate_carry_ae`, `generate_intermed_4`, `generate_intermed_12` → Compute on-the-fly The goal: reduce memory peak (and get close to mem tracker report) Was +2 GB -> reduced to +0.9 GB Should be tested on various blocks (was tested on 21M)
Co-authored-by: Jonathan Wang <31040440+jonathanpwang@users.noreply.github.com>
This PR tunes the leaf aggregation parameters by changing `n_stack` in `default_leaf_params` from 18 to 19. **Configuration:** - `l_skip = 2` (unchanged) - `n_stack = 19` (changed from 18) - **Total: l_skip + n_stack = 21** This PR is part of a series testing different values for l_skip + n_stack. Related to #419 🤖 Generated with [Claude Code](https://claude.ai/code) --------- Co-authored-by: claude[bot] <41898282+claude[bot]@users.noreply.github.com> Co-authored-by: Zach Langley <zlangley@users.noreply.github.com> Co-authored-by: Jonathan Wang <jonathanpwang@users.noreply.github.com>
include was using internal thrust file docker build failed
…486) ## Summary - Replace rayon iteration with `std::thread::scope` for preflight in the verifier sub-circuit - With only 3-4 proofs max (`MAX_NUM_CHILDREN_LEAF=4`, `MAX_NUM_CHILDREN_INTERNAL=3`), this avoids Rayon's thread pool overhead (wake-up, work stealing, synchronization) while still getting parallelism with minimal overhead from direct thread spawning - CPU module tracegen still uses `par_iter` for parallelism across modules ## Test plan - [x] Existing tests pass - [ ] Verify aggregation performance is unchanged or improved 🤖 Generated with [Claude Code](https://claude.com/claude-code) Co-authored-by: Claude Opus 4.5 <noreply@anthropic.com>
…#489) Summary metrics for `run_preflight` have disappeared, adding them back
Adds a second internal-recursive layer before the compression layer. We found that on occasion the last internal-recursive layer may have > 131000 Poseidon2 trace row columns, which contributes significantly to the `Proof` size. We mitigate this risk by adding an additional layer. --------- Co-authored-by: Jonathan Wang <31040440+jonathanpwang@users.noreply.github.com>
Before merging, I will force push to rebase `openvm:develop-v2` and
`stark-backend:develop-v2` with the current `develop-v2-new` branches.
Updated reth-bench to use `openvm-eth.git` which updates reth to v1.10
so there are slight guest program changes (improvements).
Closes INT-6109
## Claude Summary
Upgrade Plonky3 from git revision to version 0.4.1 and update
stark-backend to `develop-v2-new` branch.
This PR applies all necessary API renames and fixes to make `crates/`
compile with the new Plonky3 0.4.1 release.
## Changes
### Dependency Updates
- Plonky3: git rev `539bbc8` → version `0.4.1` (crates.io)
- stark-backend: branch `develop-v2` → `develop-v2-new`
- Removed `nightly-features` feature flag from sdk
### Plonky3 API Renames
**Trait renames:**
- `FieldAlgebra` → `PrimeCharacteristicRing`
- `FieldExtensionAlgebra` → `BasedVectorSpace`
**Associated type renames:**
- `PrimeCharacteristicRing::F` →
`PrimeCharacteristicRing::PrimeSubfield`
**Method renames:**
| Old | New |
|-----|-----|
| `from_canonical_u8/u32/usize` | `from_u8/u32/usize` |
| `from_wrapped_u32/u64` | `from_u32/u64` |
| `from_base_slice` | `from_basis_coefficients_slice` |
| `from_base_iter` | `from_basis_coefficients_iter` |
| `from_base_fn` | `from_basis_coefficients_fn` |
| `as_base_slice` | `as_basis_coefficients_slice` |
| `sample_ext_element` | `sample_algebra_element` |
| `from_f` | `from` |
| `Bn254Fr` | `Bn254` |
### API Signature Changes
**Methods now return `Option`:**
- `from_basis_coefficients_slice` → added `.unwrap()` calls
- `from_basis_coefficients_iter` → added `.unwrap()` calls
- `row_slice(n)` → added `.expect("window should have two elements")`
calls
- `ith_basis_element(n)` → added `.expect("basis element index out of
bounds")` calls
**New conversion method:**
- Use `from_prime_subfield()` instead of `from()` when converting from
`PrimeSubfield` type
### Additional Trait Bounds
- Added `InjectiveMonomial<BABY_BEAR_POSEIDON2_SBOX_DEGREE>` bound where
`Poseidon2SubChip` is used
## Test Plan
- [x] `cargo build` compiles all default crates
- [x] `cargo build --features cuda` compiles with CUDA support
- [ ] Run test suite
---------
Co-authored-by: Claude Opus 4.5 <noreply@anthropic.com>
This resolves INT-6035. All changes are about using the flattened opening claims properly.
Merge openvm-org/stark-backend#246 first and then update stark-backend branch before merging this PR. closes INT-5862 INT-5824 --------- Co-authored-by: Claude Opus 4.6 <noreply@anthropic.com> Co-authored-by: Zach Langley <zach@axiom.xyz>
Update paths and code usage for sdk-v2, recursion-v2, continuations-v2
Co-authored-by: Jonathan Wang <31040440+jonathanpwang@users.noreply.github.com>
- Comment out v1 crates (continuations, verify_stark) from workspace - Remove openvm-native-compiler dev-dep from openvm-circuit, replace v1 native opcodes with rv32im equivalents in program tests - Fix sdk-v2 cuda feature missing dep:openvm-cuda-backend - Migrate extensions/ecc/tests from openvm-sdk to sdk-v2 - Fix FriParameters -> SystemParams in pairing tests - Fix air_test_impl type inference in ruint tests
jpw-axiom
force-pushed
the
develop-v2.0.0-beta
branch
from
1952632 to
c9f04db
Compare
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
9 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.
This will resolve INT-6189