Eve library

Phase 129 completion evidence

Tracker

Revision 1

On this page

Historical task source. The Eve task board holds current task state.

Verified on 2026-07-14 in /mnt/volume300/workspace/oshun-lt-2.

Completion audit#

  • TODOS/phase-129.md contains 77 completed tasks and no unchecked task. The implementation is distributed across the seven published Phase 129 libraries: multiscale, phase-field, crystal-plasticity, dislocation-dynamics, damage-mechanics, process-simulation, and multiscale-workflows.
  • A source scan found no actionable TODO/FIXME markers, unimplemented-error stubs, or disabled tests in the seven libraries. Their 15 TypeScript test files contain 186 declared test cases.
  • All seven Nx test targets pass after replacing their deprecated @nx/vite:test executors and invalid shared config paths with working @nx/vitest:test targets and package-local configs.
  • All seven strict TypeScript checks, ESLint targets, and direct package builds pass sequentially. Their library TypeScript configs opt out of composite mode so workspace source aliases can be checked without false cross-project TS6307 file-list errors.
  • Dependency edges connect the Phase 129 libraries to their Phase 117, 120, 122, 123, 125, 126, and 128 inputs. Runnable targets cover native and WASM Rust builds, Python bridge tests, integration tests, benchmark suites, and bundled phase-field, CPFEM, and process-simulation examples.

Scientific and solver corrections#

  • Added conservative 2:1-balanced adaptive quadtree phase-field operators, physics-indicator refinement, conservative field transfer, and CFL control.
  • Added P1 triangular finite-element assembly on arbitrary geometry, automatic energy-derived residual/Jacobian evaluation, Newton iteration, and sparse conjugate-gradient solves.
  • Added a coupled multicomponent CALPHAD adapter and semi-implicit spectral Cahn-Hilliard solver with full mobility and gradient-energy matrices. Ternary regression tests verify coupled chemical potentials, Hessian symmetry, and independent conservation of every species field.
  • Corrected the analytic interface energy for f(phi) = W phi^2 (1-phi)^2 from sqrt(2 kappa W)/3 to sqrt(2 kappa W)/6, with an exact quadrature regression.
  • Added global tetrahedral/hexahedral CPFEM mesh assembly, essential boundary conditions, sparse direct/iterative solvers, and Newton-Raphson nonlinear element callbacks.
  • Added uncertainty-weighted joint inverse calibration across tension, compression, shear, nanoindentation, micropillar, diffraction-strain, and DIC field datasets, including per-modality residual diagnostics.
  • Added topology-aware 3-D DDD line discretization, Peach-Koehler and elastic interaction forces, overdamped mobility, line tension, segment annihilation, shared-node junction formation, and thermally activated screw cross-slip.
  • Added spectral operator learning with exact mean conservation and compatible hard bounds, plus a neural surrogate whose output bounds and coordinate-wise monotonicity are enforced by construction. Held-out operator and between-sample constraint tests complement the existing Gaussian-process coverage.

Native and Python interoperability#

  • Rust workspace: periodic PDE kernels and a validated CSR/Jacobi-PCG sparse solver expose safe Rust and C ABIs. Six Rust unit tests and doc tests pass; cargo fmt --check, Clippy with -D warnings, native release, and wasm32-unknown-unknown release builds all pass.
  • Python bridge: the portable reference backend and real ctypes native dispatch both pass. The portable run passes four tests with the native parity test intentionally unavailable; the native target passes all five tests, including native/reference numerical parity. Ruff check/format and strict mypy pass.

Benchmarks, integration, and examples#

  • The serialized benchmark targets pass 32 phase-field, 27 crystal-plasticity, 17 process-simulation, and five cross-library canonical benchmark cases.
  • The Phase 129.8.2.4 process-structure-property-performance regression passes with provenance and deterministic replay.
  • All three example workflows bundle and execute. The phase-field example reduces energy from 1.9242743551 to 1.8913125042; the CPFEM example uses 12 FCC slip systems and produces nonzero stress/plastic strain; the process example produces a finite melt-pool depth and keyholing classification.
  • No Phase 129 package contains a mounted TSX/JSX application surface. The workbench integration supplied by this phase is typed dashboard/viewer data, EBSD import, and notebook templates, so no browser route exists on which Playwright coverage would be applicable.

Resource-safe execution#

The host had 15 GiB RAM, 31 GiB swap, 44 GiB free disk, and approximately 13 GiB available memory before final verification. Nx lint, tests, builds, benchmarks, integration, examples, Rust, and Python gates were run sequentially with Phase 129 scope and --parallel=1 where applicable. No gate approached the 2 GiB available-memory stop threshold.