Adding gauge shift

[sbacchio]

This is a (partial) implementation of shifting for gauge field. The CPU kernel is missing, but in light of the unification of cpuGaugeField and cudaGaugeField, I would avoid implementing it.

[sbacchio]

@maddyscientist thank you for the review. I applied the suggested changes and run clang format via git-clang-format (hopefully the right way).

Only the unit testing would be missing. Could you please point me to one of the current tests to extend / base it on?

[maddyscientist]

@maddyscientist thank you for the review. I applied the suggested changes and run clang format via git-clang-format (hopefully the right way).

Looks good I think.

Only the unit testing would be missing. Could you please point me to one of the current tests to extend / base it on?

I guess the easiest one to hack from as a base is plaq_test. It's not quite what you want since this proposed test would be done using native QUDA types rather through the interface.

The bigger question is what the test would consist of. Some ideas, OTOH:

• check that the plaquette is invariant after shifting
• check that performing a closed loop of shifts leads to a bitwise identical guage field to the initial field

Anything else come to mind?