v0.2.0
FastIsostasy v0.2.0
Breaking change
SolverOptions
was introduced to specify options and replaces a number of keyword arguments normally provided directly toFastIsoProblem
.- Running interactive sea level requires a mask to prevent changes of the load column close to the boundary of the domain.
- To make sure that the elastic and geoid response are saved in the output for a step-wise forcing, one needs to provide an additional time step. For instance:
t = [-1e-8, 0, 50_000]
for a forcing that starts at zero and a simulation that ends at 50 kyr. - Anything related to time-variable viscosity was removed, since we have no idea of the evolution of viscosity over time.
Improvements
- Support of fully interactive sea level with absolute ice thickness, bathymetry... etc.
- Data loaders for most of the data required to run a simulation.
- Generic boundary conditions which minimise memory allocation on GPU and CPU.
InplaceConvolution
internally allows to reduce the memory allocation (which was critical for high-resolution runs).- Type instabilities of
dudt_isostasy
were fixed, code is faster than ever! - Piecewise linear approximation of ocean surface falls back to piecewise constant approximation if does not converge. Robust and sensible output (however impossible to benchmark so far...)
Merged pull requests:
- inplace fft and column anoms saves mem (#71) (@JanJereczek)
- Couple adaptive ocean (#82) (@JanJereczek)
- adapt docs to v0.2 (#86) (@JanJereczek)
- Final modif of plots for test 4 before submission (#94) (@JanJereczek)
Closed issues: