All pull requests

Pull requests

• Graded multi-device pole-media fields are judged by the accumulated float32 rounding law, not a per-step 9-ULP bar (weekly S2 failure, #1402)

  lane:distributedlane:nu-meshmerge-train
  #1899·rfx-agent[bot] opened Oct 19, 2026Bot·

  ·
  1

• Graded multi-device runs with a waveguide port or a TFSF source fall back to one device on both run() and forward() (#1373 census A5)

  lane:absorberlane:ci-infralane:distributedlane:nu-mesh
  #1898·rfx-agent[bot] opened Oct 19, 2026Bot·

  ·

• Lumped RLC elements no longer add a whole-grid copy per element (#1823)

  lane:ci-infralane:coax-mixed-portlane:performancemerge-train
  #1897·rfx-agent[bot] opened Oct 19, 2026Bot·

  ·
  1

  1

• compute_waveguide_s_matrix and the lumped/wire S scan run the same preflight as run() (#1373)

  lane:ci-infralane:distributedlane:examples-docslane:waveguide-port
  #1895·rfx-agent[bot] opened Oct 19, 2026Bot·

  ·

• Graded run() returns lumped-port S, formed as on uniform meshes (#1373)

  lane:absorberlane:ci-infralane:coax-mixed-portlane:distributedlane:examples-docslane:nu-mesh
  #1892·rfx-agent[bot] opened Oct 19, 2026Bot·

  ·

• Support page refusal tables checked cell by cell against the path table (docs-consistency, stacked on #1634)

  lane:ci-infra
  #1716·rfx-agent[bot] opened Oct 13, 2026Bot·

  ·

• Multi-device runs: the 2.0 page filled (#1171)

  lane:examples-docs
  #1701·rfx-agent[bot] opened Oct 13, 2026Bot·

  ·

• run() and forward(): the 2.0 page filled; api/simulation until_decay line corrected (#1171)

  lane:examples-docs
  #1700·rfx-agent[bot] opened Oct 13, 2026Bot·

  ·

• A float32 graded profile is refused on the concrete path but admitted when traced

  lane:nu-mesh
  #1654·rfx-agent[bot] opened Oct 12, 2026Bot·

  ·

• Far-field (NTFF) and H-plane objectives in forward(gradient="adjoint"): 2.1x a forward run at 5e5 cells, 1.7x at 3.7e7 where jax.grad runs out of memory (#1424)

  lane:performance
  #1439·rfx-agent[bot] opened Oct 2, 2026Bot·

  ·
  3

• #498: fix the openEMS comparator's missing Kurokawa factor, and close the wider-box branch at zero compute

  #1059·bk-squared opened Sep 15, 2026Owner·

  ·
  4