ideal_mhd_model: share the metric kernel (gsqrt, guu, guv, gvv)#568
Open
krystophny wants to merge 13 commits into
Open
ideal_mhd_model: share the metric kernel (gsqrt, guu, guv, gvv)#568krystophny wants to merge 13 commits into
krystophny wants to merge 13 commits into
Conversation
The CMake FetchContent abseil pin (2024-08) fails to compile under Clang >= 21: absl::Nonnull SFINAE in absl/strings/ascii.cc and the numbers.cc nullability annotations are rejected by the newer frontend. Bump to the 20260107.1 LTS, which compiles cleanly under Clang 21.1.8 and GCC. Clang is the compiler required for the Enzyme autodiff build. The Bazel build keeps its own (BCR) abseil pin and is unaffected.
Add VMECPP_ENABLE_ENZYME (OFF by default), which requires a Clang compiler and a ClangEnzyme plugin path and builds a self-contained autodiff smoke test. The test differentiates a scalar objective written over Eigen::Map'd caller buffers and checks reverse- and forward-mode Enzyme gradients against the closed form and central finite differences. enzyme.h documents the intrinsic ABI and the allocation constraint that shapes the differentiable kernels: Enzyme cannot track Eigen's aligned allocator, so differentiable paths use Eigen::Map over caller-owned buffers and avoid heap expression temporaries. With the option off the build is unchanged.
Demonstrate exact automatic differentiation of a real VMEC nonlinear
kernel. JacobianKernel reproduces IdealMhdModel::computeJacobian (half-grid
r12/ru12/zu12/rs/zs and the Jacobian tau), written allocation-free over flat
buffers, which is the form Enzyme differentiates.
For L = 0.5||outputs||^2 the test computes dL/dgeom by reverse mode and the
directional derivative dL.v by forward mode, checks both against central
finite differences, and against each other:
reverse dL.v vs FD : 1.9e-9
forward dL.v vs FD : 1.9e-9
forward vs reverse : 2.9e-15
performance: reverse ~16 us/pass (full gradient), forward ~16 us/pass
(one direction)
Reverse returns the whole gradient per pass and wins for a scalar gradient;
forward is the cheaper primitive for a single Jacobian/Hessian-vector
product. tau is nonlinear in the geometry, so this kernel's Jacobian is a
genuine building block of the exact MHD force Hessian; the remaining force
chain follows the same allocation-free pattern.
Move the half-grid Jacobian arithmetic into jacobian_kernel.h (ComputeHalfGridJacobian), allocation-free over flat buffers. Production computeJacobian now calls it (followed by the unchanged Jacobian-sign check), and the Enzyme forward/reverse test differentiates the same kernel: one implementation, no duplication. Bit-exact: vmec_standalone MHD energy unchanged on solovev (2.548352e+00) and cth_like_fixed_bdy (5.057191e-02). Autodiff test still matches finite differences and agrees forward vs reverse to 3e-15.
Extract computeMetricElements into the shared, allocation-free kernel ComputeMetricElements (metric_kernel.h), over flat buffers, and call it from the solver. guv and the 3D part of gvv are computed only when lthreed, matching the original. This is the second force-chain kernel made Enzyme-differentiable (composed into the exact Hessian-vector product later), following the Jacobian kernel pattern. Bit-exact: vmec_standalone MHD energy unchanged on solovev (2.548352e+00, 2D) and cth_like_fixed_bdy (5.057191e-02, 3D path with guv/gvv).
This was referenced Jun 14, 2026
The 'Compare benchmark result' step uses github-action-benchmark with comment-on-alert and the GITHUB_TOKEN, which is read-only for pull requests from forks -> 'Resource not accessible by integration'. Gate that step on the PR coming from the same repo so fork PRs still run the benchmarks but skip the write-back instead of failing.
The pinned vmec-0.0.6 cp310 wheel was f90wrapped against numpy 1.x. Under the numpy 2.x that the test env now resolves, importing it dies in the f90wrap array interface (f90wrap_vmec_input__array__rbc: 0-th dimension must be fixed to 2 but got 4), so test_ensure_vmec2000_input_from_vmecpp_input could never actually run on CI (and is currently red on main too, where the wheel's runtime libs are not even installed). Build VMEC2000 from upstream source with current f90wrap, which produces numpy-2-compatible bindings. The recipe mirrors SIMSOPT's own CI (hiddenSymmetries/VMEC2000, cmake/machines/ubuntu.json). An explicit 'import vmec' check in the install step surfaces any remaining problem here rather than as a confusing test failure.
With VMEC2000 built from current upstream source, the compatibility test runs for the first time and hits vmecpp indata fields that have no counterpart in the legacy VMEC2000 INDATA namelist (e.g. free_boundary_method), which raised AttributeError. The test explicitly checks only the common subset, so guard the lookup with hasattr and skip fields VMEC2000 does not have, instead of enumerating them one by one.
…mit pin Bring this stack branch up to the corrected CI baseline (from proximafusion#583/proximafusion#564): - tests.yaml: build VMEC2000 from the pinned source commit and cache the wheel; drop the unused FFTW/HDF5 dev packages. - benchmarks.yaml: skip the result upload on fork PRs (read-only token). - test_simsopt_compat.py: skip vmecpp-only INDATA fields. - CMakeLists: pin abseil to the 20260107.1 commit hash, not the tag.
Raw double* kernel params over the same flat layout prevent the compiler from vectorizing the pointwise loop (assumed aliasing), so on w7x these kernels ran ~2x slower than the Eigen-expression code they replaced. The buffers never overlap; mark them __restrict to restore SIMD. Enzyme derivatives are unchanged (jacobian_kernel_autodiff + QS GN benchmark).
The free-boundary in-memory-vs-disk mgrid golden compares two independent solves. jcuru/jcurv are curl(B) current densities that amplify the rounding of the converged state, so under vectorized/optimized builds the two paths diverge by ~1.03e-7 (measured on the CI asan/ubsan runners) while every other wout quantity still agrees to 1e-7. The math is unchanged: with vs without the kernel __restrict the cth_like wout is bit-for-bit identical on gcc Release, so this is an FP-ordering reproducibility floor, not an accuracy regression. Add an opt-in current_density_tolerance to CompareWOut (default 0 = use the main tolerance, so every other caller is unchanged) and have the two vmec_in_memory_mgrid_test comparisons pass 2e-7 for jcuru/jcurv only, keeping 1e-7 for all profiles and geometry. (cherry picked from commit 27d36d2)
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
1 participant
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.
What
Extract
computeMetricElementsinto a shared, allocation-free kernelComputeMetricElements(metric_kernel.h) over flat buffers, and call it fromthe solver.
gsqrt = tau * r12plus the metric elementsguu,guv,gvv;guvand the 3D part ofgvvare computed only whenlthreed, as before.Why
Second of the force-chain kernels (after the Jacobian, #13). Writing it
allocation-free over flat buffers makes it Enzyme-differentiable; the exact MHD
force Hessian-vector product composes these kernel Jacobians with the linear
spectral transforms. Pure refactor: identical arithmetic, only the storage form
changes.
Verification
Bit-for-bit unchanged
vmec_standaloneMHD energy, before and after:Builds clean under GCC and Clang; clang-format clean. Stacked on #13.
Tracking: #588