v1.4.1

This release updates to stark-backend v1.2.1 for CUDA performance gains in the backend. Other changes include:

Added

• (CLI) Add configurable segmentation arguments to cargo openvm prove commands.

Changed

• (Toolchain) Update cargo openvm build to use Rust nightly version nightly-2025-08-02.
• (Primitives Library) Fix in openvm-algebra-complex-macros to ensure const byte arrays have proper memory alignment.
• (Executor) Modified VirtualMachine::build_metered_ctx to take the program (&VmExe<Val<E::SC>>) as an argument.
• (Prover/Executor) Changed default segmentation limits and updated segmentation logic.

What's Changed

• chore: mention "keygen before commit" in the docs by @Golovanov399 in #2111
• ci(vocs): update pnpm lock file by @jonathanpwang in #2112
• ci(vocs): hard copy all symlink dirs by @jonathanpwang in #2113
• ci(ami): add RunsOn image for cuda 12.9 by @jonathanpwang in #2115
• fix(ci/docs): recommend CUDA toolkit 12.9 by @jonathanpwang in #2116
• ci: build docs with cuda feature by @jonathanpwang in #2117
• ci: add release workflow to upload agg,halo2 pk to s3 by @jonathanpwang in #2114
• chore(audit): ignore cargo audit false positive by @jonathanpwang in #2120
• docs: add whitepaper link to vocs by @Qumeric in #2119
• ci: skip debug during CI air test by @stephenh-axiom-xyz in #2118
• fix: reduce threads per block for FRI tracegen launcher by @stephenh-axiom-xyz in #2124
• ci: script + workflow to run clang-tidy during CUDA lints by @stephenh-axiom-xyz in #2122
• docs(vocs): use versioned rustdocs redirect by @jonathanpwang in #2125
• perf: pass state variables by value in tco exec handlers by @shuklaayush in #2121
• chore(cuda): use newer ami by @gaxiom in #2131
• ci(cuda-13): clean git cache by @jonathanpwang in #2132
• perf: only do exit check for instructions that can exit by @shuklaayush in #2126
• ci(runs-on): set CUDA_ARCH env var in pre-install by @jonathanpwang in #2133
• chore(cuda): update cuda-backend with Virtual Memory Manager and NTT refactor by @gaxiom in #2127
• fix: add std feature to openvm-prof serde by @stephenh-axiom-xyz in #2135
• fix: take program air height into account in metered execution by @shuklaayush in #2137
• chore(cuda): update poseidon2 kernel to use memory manager by @gaxiom in #2108
• fix: ensure segments are below thresholds by @shuklaayush in #2138
• fix: log more information for segments by @shuklaayush in #2143
• feat: E2 supports suspension & VmExecState supports clone by @nyunyunyunyu in #2140
• chore: Rename execute_metered_until_suspension to execute_metered_until_suspend by @nyunyunyunyu in #2144
• chore: update install link by @trevor-trinh in #2146
• chore: bump stark-backend to v1.2.1-rc.2 by @jonathanpwang in #2148
• chore(cuda): BabyBear unified + CUDA_DEBUG + opener in natural order by @gaxiom in #2149
• fix: Over-align public constants in moduli_init by @Avaneesh-axiom in #2150
• chore: update openvm build nightly version by @jonathanpwang in #2152
• fix: execution from non-zero instret by @jonathanpwang in #2155
• chore: add sdk app prove without verify by @luffykai in #2164
• chore(cuda): keccakf memcpy handling of padding by @jonathanpwang in #2167
• feat(cuda): VPMM v3 and async app prover by @gaxiom in #2165
• chore: Remove send + sync bound on executor by @Schaeff in #2156
• feat(cuda): use VPMM v3.1 by @gaxiom in #2169
• docs(transpiler): clarify itof function definition by @jonathanpwang in #2170
• audit: v1.4.1 report by @jonathanpwang in #2172
• chore(docs): update link to example repo by @jonathanpwang in #2176
• fix(cuda): sync memory subtree streams before drop by @jonathanpwang in #2175
• fix: Expose CUDA primitives by @qwang98 in #2179
• feat(cli): add SegmentationArgs to prove command by @jonathanpwang in #2178
• release: v1.4.1 by @jonathanpwang in #2171

New Contributors

• @trevor-trinh made their first contribution in #2146
• @Schaeff made their first contribution in #2156
• @qwang98 made their first contribution in #2179

Full Changelog: v1.4.0...v1.4.1