Matthew Martineau

Some minor CMake changes to try to fix some issues with building on systems without MPI and OpenMP CXX.

DRAFT: Remove runtime type selection

3

Managing runtime selection of the supported types in AmgX makes the code more complex but doesn't seem to benefit users significantly from our discussions so far. This is a draft...

DRAFT: Replacing memory manager with cudaMallocAsync API

1

Currently proof of concept demonstrating that cudaMallocAsync can match performance of the existing custom memory allocator. This requires the threshold of the memory pool to be set (currently maxing out...

Thrust namespace fix

4

Thrust now provides a workaround for issues caused by multiple libraries / applications using thrust by wrapping the thrust namespace (https://github.com/NVIDIA/thrust/releases/tag/1.14.0) This patch wraps all thrust calls in the amgx...

I would personally like to remove the consolidation feature. This includes fine and coarse level gluing. In the future I think we should revisit the coarse level gluing, but the...

This is an in progress change that will reimplement many of the sparse operators and improve the structure of the code. Significant attention is focused on aggressive coarsening optimisations.

Re-instated the eigensolver functionality as first class. Moved the configuration files to the examples directory.

Adopting the REUSE standard for licensing

1

Replaced all license headers with SPDX equivalents and added a main LICENSE file in the LICENSES/ directory