PAM-A dycore work

[mrnorman]

• (Done) Reduce total kernel count in spatial operator
  • (Done)This will mainly involve pulling BC kernels into other kernels
  • (Done) Go ahead and assume periodic BCs in horizontal and solid wall in vertical for simplicity

RK

• Separate ADER from RK spatial operators
• Update RK to work with new interface requirements
• Update RK to do splitting trivially within each stage

Vertical Acoustic Subcycling

• Maybe start with multi level parallelism with the full vertical solver as a beginning stopgap. No critical decisions or new algorithmic development necessary. Just new functionality in YAKL.
• Determine which terms need vertical momentum to be acoustically upwind and which terms can have it ben advectively upwind
• Create RK and ADER options that subcycle vertical dynamics as a whole
• Create a sub-cycled version of rho, rhow, and rhotheta
  • Mass consistent, stable upwinding
  • ADER will take some figuring out to compute rho*w DTs over larger advective time step
  • RK will take figuring out as well, I think