GPU version of interpreter and miner

[ckrause]

Create a GPU version of the LODA interpreter and miner. It should be in a new repository loda-gpu. Use OpenCL? In the first step, we should be able to run a really simple LODA program (w/o loops) on a GPU without any parallelization.

[ckrause]

We would first need to migrate away from the BOINC wrapper app to a native BOINC app. That alone is a major effort. After that we would need to extend loda-cpp with separate code and new binaries specifically for BOINC/GPU. Contrary to the original plan, it would be within loda-cpp, not in a separate repo.

• Native BOINC apps: https://boinc.berkeley.edu/trac/wiki/BasicApi
• Building GPU apps: https://boinc.berkeley.edu/trac/wiki/GPUApp

[ckrause]

GPU-related ML code is considered in https://github.com/loda-lang/loda-python