Fabian Hoppe

Since this is related to #341, which is important and stil open, this issue should be left open. (Reviewed within #1109)

Since `.npz`-format is not an HPC-typical format, it does not make sense from my point of view, to save `DNDarray`'s that really need distribution in this format. On the other...

This is about an interesting functionality that is still missing. Therefore, reviewed and kept within #1109

Could serve as a first issue for someone who already wants to start with a not so straightforward example.

Error seems to be as follows: ``` # apply torch convolution operator > signal_filtered = fc.conv2d(signal, weight) E RuntimeError: GET was unable to find an engine to execute this computation...

just a comment: once #1097 has been completed, one might add higher-dimensional convolution built on FFT (which is efficient at least for large kernels)

For 1,2,3 MPI-processes on GPU I get the error: ``` ====================================================================== ERROR: test_convolve2d (heat.core.tests.test_signal.TestSignal) ---------------------------------------------------------------------- Traceback (most recent call last): File "****/heat/heat/core/tests/test_signal.py", line 165, in test_convolve2d conv = ht.convolve2d(dis_signal, kernel_odd,...

I will try to complete review as soon as possible

Since this PR does not contain high-level routines ready to be used by non-experts and it does not seem to be active, I close this PR (as discussed in the...

Meanwhile I have implemented a basic idea of `DXarray` in the corresponding branch. In the near future I plan to go on with: * `DXarray.to_xarray()`: convert distributed `DXarray` to a...