Commits on 15 Sep 2026
full model: remove the rigid-body modes of the free plate after every batch
Jeremy Magland committed
d608f95docs: measured GPU timings (RTX 6000 Ada) for the eigensolver and the full model
Jeremy Magland committed
20ab5aafull model: on CUDA, capture the steps of one audio sample into a CUDA graph replayed on static buffers
Jeremy Magland committed
49f6606full model: the compile check no longer consumes a step; run() fails instead of spinning when no progress is made
Jeremy Magland committed
ff3df5dfull model: mark CUDA-graph steps so the compiled kernel runs with reduce-overhead
Jeremy Magland committed
5a24e5ccymsim: Python port of cymbalsim with a modal linear model and a PyTorch von Karman model
Jeremy Magland committed
3688cea