#138 outputs conversion failing when organ is not created but requested as output #165
Conversation
…ssues and connector data
…mesteps handled, but bad meshing with refvalue/refvectors cause overwrites if we use multiscale mappings, and unexplored issues if we avoid them. Also, many-node to many-node issues if multiscale mapping. Some TODO comments, structs, commented out code and notes are outdated. More exploration and cleanup required, but current state is much more interesting than the previous commit.
|
Note : My local branch for multi-timestep prototyping is pointing to this branch, which, uh, isn't normal. In the process, I deleted another related pull request which had a comment you can find here : #164 I thought I'd fixed it earlier. Will look into patching up the branch names later. Current prototype state : Multiple timesteps handled, but bad meshing with refvalue/refvectors cause overwrites if we use multiscale mappings, and unexplored issues if we avoid them. If we make timestep mappings not use the scale mapping codepaths, then we need to remove the timestep-mapped variables from vars_need_init and add an extra stage when the dependency graph is created to create additional links between timestep-mapped softdependency nodes and then see what other issues might arise. I think they would also need to be removed from the scale mapping if some of them are multi-scale, which might be fiddly. I haven't experimented with that yet. More exploration and cleanup required, but current state is much more interesting than the previous commit. I also expect many of the existing tests to work properly with the new timestep-handling branches enabled, though I might have some unintentional breaking API changes. XPalm tests report a version incompatibilty, so they aren't actually running. |
Benchmark Results (Julia v1)Time benchmarks
Memory benchmarks
|
…imestep-mapped variables from their source, to avoid overwriting source (inputs to non-default-timestep models are changed by the accumulation function so can't be a simple Ref to source)
…oose ends to investigate
No description provided.