peram_gen by Bastian Knippschild, modified by Bartosz Kostrzewa.

Implementation of stochastic LapH method for the generation of perambulators,
specifically tuned for the use of four NVIDIA GPUs through a special version
of tmLQCD with the QUDA interface with direct passthrough (quda_invert_direct).

This version of peram_gen is to be linked against a serial version of tmLQCD
which itself should be linked against QUDA. It works on machines with four GPUs.
The four GPUs are used in isolation, each inverting one of the four spin diluted
random sources.