Non-blocking variants in mp.f90, courtesy of Thorsten Kurth.
This work has been extracted from the Merge Request opened by Thorsten Kurth for q-e-gpu
.
His message: added non-blocking variants in mp.f90, mainly for circular shift left and right (only c2d is tested, but implemented for r2d, i2, i1, i0) and also testall
and waitall
which later can be used for any non blocking comm call.
TODO: add tests in UtilXlib unit test-suite.