[Feature]: update OpenCL-kernel

[kitsudaiki]

Feature

Description

The last recent changes were only tested with CPU and CUDA-kernel. The OpenCL-kernel compatible with the current data-structures. This has to be fixed by porting the CUDA-kernel to OpenCL.

Acceptance Criteria

1. The OpenCL-kernel works the same way with the same accuracy like the CUDA-kernel
2. There should be an automatic detection, if the available gpu is an nvidia-gpu or not
3. The OpenCL-kernel is used instead of the CUDA-kernel, when no nvidia-gpu is available
4. For the purpose of testing, it is possible to override the automatic detection and enforce the OpenCL-kernel even on nvidia-gpus

Additional Information

In Version 0.2.0 there was already a working version. Check this one out

Blocked by

• 61 because cmake privides a better option to compile text-files into the code

How to test

• Test on node with OpenCL- and CUDA-kernel and if possible test on a node with AMD and a node with NVIDIA gpu

[kitsudaiki]

At the moment this is only a nice-to-have feature and not necessary for now, so it was moved into the backlog