Repository navigation
release: CableDyn 0.4.1 - #106
Open
SMI-Lab-Inha wants to merge 19 commits into
Open
SMI-Lab-Inha wants to merge 19 commits into
SMI-Lab-Inha wants to merge 19 commits into
Conversation
CableDyn_UpdatePointFluidFields and CableDyn_CalcOutputDerivatives refused a deck on the coupled aggregate by naming bodies and rods only, so a deck with a finite-EI line and neither read as a wrong diagnosis. Both messages now name finite-EI lines, bodies and rods. Signed-off-by: Jae Hoon Seo <jaehoon.seo@inha.ac.kr>
…d note (#90) The LINES flag r now works in single-turbine coupled OpenFAST runs: the shell samples committed steps from RangeStart and writes <RootName>.CD.Line<L>.range.out at the end of a run that reaches TMax. The accumulators survive a restart, so OtherState gains the range state and checkpoints from earlier builds cannot be restarted. A FAST.Farm shared mooring and the C API refuse the flag by name. An initialisation note and doc/openfast.rst state how coarse the deepest interval of SeaState's depth grid is when it exceeds 5 % of the depth and a line reaches into it. Signed-off-by: Jae Hoon Seo <jaehoon.seo@inha.ac.kr>
A deck with TMax 0 and no dtM stopped with a message about a dynamic run. A static check needs no time step, so it now runs on every standalone route: the parser builds the deck on a nominal step marked as a static-check clock, which no caller can march on, and a motion record is read for its t = 0 row only. The C API keeps needing dtM on finite-EI, body and rod decks, because CableDyn_Step marches on it, and says so. Signed-off-by: Jae Hoon Seo <jaehoon.seo@inha.ac.kr>
The component wave sum took 70-80 % of a step on a long elevation record. GNU Fortran did not vectorise the sine, cosine and exponential at all. The sum stays exact: block kernels take the quadrant and exponent from the bits of the rounded value, sums run in 64 interleaved lanes added in a fixed order, components past the first 1024 are no longer evaluated twice, and the depth ratios take one division. Spectral seas, wave trains and spread seas share the gain. Results change at round-off. Signed-off-by: Jae Hoon Seo <jaehoon.seo@inha.ac.kr>
On a deck with finite-EI lines, bodies or rods the C API and Python ran in still water only: a deck current was refused, and no call reached the line nodes. - A steady deck current (current, WaterKin CurrentMod 1, MoorDyn-C Currents 1) is marched in-process, in the static start and in every step, also from host start positions. - A host can supply the ambient fluid at every node instead (ABI 1.4): CableDyn_SetHostFluid before the initialization, then CableDyn_NFluidNodes, CableDyn_GetFluidNodePositions and CableDyn_UpdateFluidFields; in Python set_host_fluid, n_fluid_nodes, fluid_node_positions and update_fluid_fields. A steady deck current then sets the static start only, as a host's current does in OpenFAST. - The field follows the standalone march and OpenFAST: sampled at t + dt at the step-start nodes and held over the step. The first field, at t = 0, sets the start acceleration, on EI = 0 lines too, and a step without a field is refused. Stepped in current and regular waves, a cable matches the standalone driver to the digits of its output (NRMSE 2e-7), with a stiffener and as-laid seabed friction too; SaveState and RestoreState carry the field. With a host field the deck is read by the coupled rules, so SEASTATE and WaveKinMod 2 policies reach the aggregate and no TMax is needed. Signed-off-by: Jae Hoon Seo <jaehoon.seo@inha.ac.kr>
Under a tapered bend stiffener the stiffness is one value per element, so every node sits on a small step the section does not have. The node outputs took the softer side, which overstated curvature by about h |EI'| / (2 EI); the element outputs carried a bias of the same size, of either sign depending on the deck. A stiffened line now keeps EI(s) at every node, one limit per side, and the larger-side rule compares those limits, so a real step stays a step. The .elements.out peaks and OutputRecovery element on sleeved elements report the section moment over EI(s), with the same soft-section fallback as the node path. Lines without a stiffener are bit-identical. Signed-off-by: Jae Hoon Seo <jaehoon.seo@inha.ac.kr>
A Linux job runs the full CTest on the Prescott OpenBLAS kernel, confirms the kernel took effect, and runs every shipped example deck, failing on a fallback note where a deck should converge cleanly. A test that passes on one kernel and fails on another is a bug. tests/check_example_fallbacks.py holds the scan and example_fallback_scan tests it. Signed-off-by: Jae Hoon Seo <jaehoon.seo@inha.ac.kr>
read_output could not read the two-table <root>.modes.out file. It now returns NaturalModes, the frequency table, with .shapes holding a ModeShapes table, and DriverResult gains modes_output and read_modes(). Signed-off-by: Jae Hoon Seo <jaehoon.seo@inha.ac.kr>
CableDyn now has init_line, init_lines and calc_output_derivatives, which bind CableDyn_InitLine, CableDyn_InitLines and CableDyn_CalcOutputDerivatives. Arguments follow the C API; shape, finiteness and dtype are checked in Python before the call and the library judges value ranges. python/tests/test_raw_lines.py compares a raw line with the same line from a deck and the Jacobians with central differences. Signed-off-by: Jae Hoon Seo <jaehoon.seo@inha.ac.kr>
… EI = 0 azimuth range (#97) CD_C_LINE_SECTION_CURVATURE and Line.section_curvature() return CurvX and CurvY at every node of a finite-EI line in one evaluation. CableDyn_GetLineValues reads position, velocity, acceleration, tension, curvature, bend moment, declination and azimuth in one pass, without a text token per node. The per-node rules moved into helpers shared with the channel evaluator, and line_values_match_channels compares the two bit for bit. An EI = 0 tension read was O(n^2) and is now linear. A line whose deck chord heads within 45 degrees of +X reported its azimuth in [0, 360) instead of (-180, 180] outside the standalone EI = 0 route, because commit_model_copy dropped azimuth_signed. Those values read 360 degrees lower. Signed-off-by: Jae Hoon Seo <jaehoon.seo@inha.ac.kr>
…and full notes (#98) A solver step retry works at any dtM/DT. The glue copies the step-start OtherState back over the predicted one, which the shell read as a rewound checkpoint: at dtM = DT it rebuilt the model on every retry, and at dtM > DT it stopped. A retry now restores the aggregate's step-start snapshot and the load predictor is mirrored back into OtherState. The committed-force cache survives a restore. CalcSteady keeps its state map through a rebuild and runs to the end of the run with more than one operating point. Range graphs are written when the host reports a completed run, in OpenFAST and FAST.Farm, and not after an error; a turbine's own instance writes <RootName>.T<n>.CD.Line<L>.range.out. Initialisation notes of a coupled line reach the console in full, and a restart no longer rewrites .CD.static.out. Parameter gains Linearize and Misc gains HostComplete, so checkpoints from earlier builds cannot be restarted. Signed-off-by: Jae Hoon Seo <jaehoon.seo@inha.ac.kr>
…d line (#99) The node outputs of a finite-EI line read the conjugate of the tangent as the moment. The element energy is EI k^2/2 per unstretched length, so that conjugate has normal length EI k/(1+e)^2, which put Curv 2e and BendMom e below their definitions. Curv, CurvX and CurvY are now the geometric curvature of the stretched line, the reciprocal of the bend radius. BendMom is the moment that balances the forces, EI k/(1+e), and StfMom the sleeve's share of it, so BendMom (1+e) = EI x Curv. Tension, positions and forces do not change. Signed-off-by: Jae Hoon Seo <jaehoon.seo@inha.ac.kr>
…ecords (#96) wave_frequencies.txt (WaveKin 7) takes a different beta on each row, as MoorDyn-C reads it, so a component file can describe a short-crested sea. Short-crested tables use their own branch of the component sum, so the single-direction path is unchanged. The new options WaveRefX and WaveRefY (m, default 0) say where a wave record or the component phases are referenced, for WaveKin 3, WaveKin 7 and WaterKin WaveKinMod 1. They extend the MoorDyn-C format. Vessel RAOs take each row's heading and the vessel position relative to the point. DeckFile now reads the files a deck names and refuses what the solver refuses, on spelling aliases and static checks too. RAO headings are judged by the bins that carry energy. A direction bin below the smallest normal weight is exactly zero natively, so which bins carry the sea depends on the spreading and the bin count alone. Signed-off-by: Jae Hoon Seo <jaehoon.seo@inha.ac.kr>
…ts (#100) CableDyn_GetInitNotes and CableDyn.init_notes return the notes of the last deck initialisation on the point route and the aggregate, for a host that does not read the standard streams. The Python binding is optional, so a library without the call still loads and only init_notes refuses. An initialisation note names the lines that a current can move along a seabed without friction, and a route that rejects frictionMu says so. Signed-off-by: Jae Hoon Seo <jaehoon.seo@inha.ac.kr>
A coupled run restarts from a checkpoint at any glue step: ChkptTime need not be a multiple of dtM, and the restarted run matches the uninterrupted one bit for bit. Nothing in the solver moves between commits, so a new OtherState field, commit_input, holds what the last commit was reached with, and a rebuild stands on it at the commit's clock. A restart on the last glue step evaluates the committed loads first. Only a time outside the mooring step is still refused. CalcSteady and -VTKLin work at dtM larger than DT. A restart before the first commit keeps a body or rod at its start. An earlier run's range graphs go as soon as the deck is parsed, so an initialisation that stops after that leaves none. The module driver keeps its orientation in double precision, so a rotating motion runs in a single-precision build. Checkpoints from earlier builds cannot be restarted. Signed-off-by: Jae Hoon Seo <jaehoon.seo@inha.ac.kr>
Records everything merged since 0.4.0 in the Unreleased section, with an Upgrading part for the outputs that change. Signed-off-by: Jae Hoon Seo <jaehoon.seo@inha.ac.kr>
) A non-zero omega = 0 row of wave_frequencies.txt (WaveKin 7), or the mean of wave_elevation.txt (WaveKin 3), is now added to the elevation alone, as MoorDyn-C does: the wave numbers stay those of WtrDpth, and the Wheeler stretching maps the water under the total elevation. CableDyn took the wave numbers at WtrDpth plus the level and measured z from it. The mean of a WaterKin WaveKinMod 1 record still raises the still-water level; the docs now say that MoorDyn-F drops it. CD_Component_Wave_Kinematics takes an optional eta_offset, a steady level in the surface the stretching maps. A new wavekin_modes case runs both files with a 0.8 m level against MoorDyn-C v2.7.2, with six WaveKin 7 components in six directions. WaveKin 7 matches to 1e-8 of its scale, where the previous rule was 7.7e-3 m off in elevation and 1e-2 in velocity. WaveKin 3 matches to the existing grid tolerances, where it was 6e-3 off in velocity. Signed-off-by: Jae Hoon Seo <jaehoon.seo@inha.ac.kr>
Set the version to 0.4.1 (2026-11-08) in the build, the Fortran banner, the OpenFAST module, the C API and its header, the Python package and CITATION.cff, and date the CHANGELOG section. The coupling-boundary page now lists the calls of ABI minor extension 4. Signed-off-by: Jae Hoon Seo <jaehoon.seo@inha.ac.kr>
Holcombe and Bergdahl rerun with the 0.4.1 GNU Release build. Both reproduce their summaries, metrics and refinement figures byte for byte. Of the files provenance.json hashes, only the two Holcombe static profiles change: their curvature and bend-moment columns, by at most 1.1e-4 relative, from the geometric curvature of a stretched line. The Bergdahl waveform comparison still stands from the 0.3.0 run. Signed-off-by: Jae Hoon Seo <jaehoon.seo@inha.ac.kr>
Codex Review Summary: ✅ CompletedThis comment shows the latest Codex review activity on this pull request.
ℹ️ About Codex in GitHubYour team has set up Codex to review pull requests in this repo. Reviews are triggered when you
Codex reacts with 👀 while any review is running, comments if it has suggestions, and reacts with 👍 once all reviews finish with no findings. |
Owner
Author
|
@codex review |
|
Codex Review: Didn't find any major issues. More of your lovely PRs please. Reviewed commit: ℹ️ About Codex in GitHubYour team has set up Codex to review pull requests in this repo. Reviews are triggered when you
If Codex has suggestions, it will comment; otherwise it will react with 👍. Codex can also answer questions or update the PR. Try commenting "@codex address that feedback". |
SMI-Lab-Inha
requested a deployment
to
testpypi
November 8, 2026 02:41 — with 
GitHub Actions
Waiting
GitHub Actions
Waiting
This branch is waiting to be deployed
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
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.
Summary
Release CableDyn 0.4.1: dev with #87 to #104, the version bump to 0.4.1 (2026-11-08), and the experiment records rerun with the 0.4.1 build. The CHANGELOG section lists every change and, under Upgrading, the outputs that change.
check_release_metadata.pypasses.Checks on this tree
twine check.The release record
validation/RELEASE_0_4_1.mdfollows with the asset, clean-machine and linearisation results. It also corrects one row of the 0.4.0 record: its six-kernel claim was five, because the Windows OpenBLAS ran Haswell for SkylakeX, and its example scan could not see the relaxed-tolerance note.