Dieter Weber

Regarding the sorting of array indices: CuPy can do this about two orders of magnitude faster than the CPU even with transfer to and from the GPU. It just requires...

Results from some profiling. Test code: ```python %load_ext line_profiler import numpy as np from libertem.api import Context from libertem.udf.base import NoOpUDF from libertem.executor.inline import InlineJobExecutor from libertem.udf.base import UDFRunner, UDFTask...

The distributed case should be tested with a real DataSet since a `MemoryDataSet` creates a lot of overhead. With a very small memory data set the additional overhead compared to...

Ok, using the file names for very large datasets sounds like a good idea since we can be reasonably sure that they were generated by the Merlin software or generally...

Also not sure if we keep this for 0.6 or bump to 0.7 since it doesn't feel like a show stopper.

> maybe adding a warning in 0.6 already, at least for use from the Python API. Maybe we can add some info to the documentation about this, in the MIB...

Discussion with @sk1p on behavior (WIP): Two options to implement the navigation dimension: 1. If the ROI partitions don't overlap and cover the entire navigation space, the interface should work...

> Maybe the `DataSet` could provide a hook for coordinate transformation, which is then used by `UDFMeta.coordinates` - so the default implementation works like the implementation from #793 by @AnandBaburajan,...

FYI, in the Ptycho 4.0 call we just discussed this item. Since Simeon is picking up work on the CUDA/Alpaka side of this, we should put it on the short...

As a follow-up, it seems that DOIs from Zenodo don't appear on Crossref at all. [Searching for a Zenodo DOI doesn't show a result](https://search.crossref.org/?from_ui=yes&q=10.5281%2Fzenodo.5083032). It appears that only Crossref members...