LLM translations are rather fucked up. Will need to rewrite from scratch to ensure validity. Sofar for speed gains.

Execution times are cut in half, but outliers with low execution speeds muddle the results of benchmarking but would also mess with the use of Fortran in optimization (gaining little to nothing).

Need to implement proper memory management. There is too much redundancy passing data back and forth. Check interfaces and data types by passing stuffing the output matrix directly.