Datastructure Features

[vetschn]

This should close #27 and related issues.

[vincent-maillou]

As we are supporting to_dense() I think we should also support to_sparse() here are my reasons why:

• For the Inv method of the GF solvers. For the gpu support, we may want to densify the entire matrix on the GPU instead of the CPU (And I know that this dense solver is a bit of a debug thing, but I feel like if we keep it, it should be supported and implemented correctly)
• For debug/compatibility purposes: let say someone have developed a bit of code that operate on a sparse system matrix. Then we could plug and test very easily its idea in QuaTrEx
• I feel like it can happen very quickly that someone want to look at what the system matrix or the self energies looks like on a system that cannot be exported with to_dense().

[vetschn]

I managed to mostly handle the higher stack dimensions. However, there is a very weird bug in the distributed transposition somewhere that I haven't been able to pin down yet. Most of the time the .dtranspose() works beautifully, but for some seemingly random combinations of stack_shape and comm.size the data on certain ranks is wrong.

[vetschn]

The GPU aware implementation is working as well now, i.e. all the datastructure test cases pass both with cupy and numpy backend. The data corruption problem mentioned above is not fixed yet and i need to add a couple more tests.

[vetschn]

Everything should work now, last steps are increasing test coverage.

One thing to note: Currently, as we have no GPU-aware MPI, .dtranspose() does not work with the cupy backend. Calling it when using cupy will raise segmentation faults (!)

Edit: I made it so that dsbsparse checks if the MPI backend is GPU-aware.

[codecov-commenter]

:warning: Please install the to ensure uploads and comments are reliably processed by Codecov.

Codecov Report

Attention: Patch coverage is 81.21547% with 68 lines in your changes missing coverage. Please review.

Please upload report for BASE (dev@67de316). Learn more about missing BASE report.

Files with missing lines	Patch %	Lines
src/qttools/datastructures/dsbsparse.py	76.61%	29 Missing :warning:
src/qttools/datastructures/dsbcoo.py	83.96%	17 Missing :warning:
src/qttools/utils/gpu_utils.py	60.00%	16 Missing :warning:
src/qttools/datastructures/dsbcsr.py	93.75%	3 Missing :warning:
src/qttools/utils/mpi_utils.py	81.81%	2 Missing :warning:
src/qttools/greens_function_solver/inv.py	50.00%	1 Missing :warning:

Additional details and impacted files

```diff @@ Coverage Diff @@ ## dev #29 +/- ## ====================================== Coverage ? 69.53% ====================================== Files ? 17 Lines ? 650 Branches ? 0 ====================================== Hits ? 452 Misses ? 198 Partials ? 0 ```

:umbrella: View full report in Codecov by Sentry.
:loudspeaker: Have feedback on the report? Share it here.

[github-actions[bot]]

Package	Line Rate	Health
.	100%	✔
datastructures	81%	✔
greens_function_solver	22%	❌
obc	92%	✔
utils	64%	➖
Summary	70% (452 / 650)	➖