Data Types and Precision

Omega supports all standard data types and uses some additional defined types to guarantee a specific level of precision. In particular, we define I4, I8, R4 and R8 types for 4-byte (32-bit) and 8-byte (64-bit) integer and floating point variables. Note that these exist in the Omega namespace so use the scoped form OMEGA::I4, etc. In addition, we define a Real data type that is, by default, double precision (8 bytes/64-bit) but if the code is built with a -DOMEGA_SINGLE_PRECISION (see insert link to build system) preprocessor flag, the default Real becomes single precision (4-byte/32-bit). For floating point variables, developers should use this Real type instead of the specific R4 or R8 forms unless the specific form is required (eg converting to single precision before output or if a particular algorithm is known to require double precision). This allows us to easily convert all reals to single precision to explore performance or accuracy in single precision mode. In some cases creating literal values of type Real is necessary to avoid unwanted promotions. For that purpose a user-defined literal _Real is provided. As an example, we can compute the inverse area of a cell using only the Real type as follows:

    Real InvAreaCell = 1._Real / AreaCell(ICell);

Arrays and Kokkos

The C++ language does not have native support for multi-dimensional arrays as part of the language standard, though there are a number of implementations as part of the Standard Template Library and elsewhere. Omega uses the Kokkos framework for defining and allocating arrays on both the CPU host and any accelerator device that may be present. Because the syntax for defining such arrays is somewhat long, we instead define a number of alias array types of the form ArrayNDTT where N is the dimension of the array and TT is the data type (I4, I8, R4, R8 or Real) corresponding to the types described above. The dimension refers to the number of ranks in the array and not the physical dimension. Although Kokkos supports Fortran ordering, we will use C ordering for array indices. Within Omega the default location for an Array should be on the device with a similar type HostArrayNDTT defined for arrays needed on the host. As an example, we can define and allocate a device and host array using:

   Array3dReal Temperature("Temperature",nTimeLevels, nCells, nVertLevels);
   HostArray3dReal TemperatureHost("Temperature",nTimeLevels, nCells, nVertLevels);

Alternatively, you can use the copy functions to create a host copy from the device or vice versa.

   auto TemperatureHost = OMEGA::createHostMirrorCopy(Temperature);

Finally, the arrays can be deallocated explicity using the class deallocate method, eg Temperature.deallocate(); or if they are local to a routine, they will be automatically deallocated when they fall out of scope on exit. More details on Kokkos arrays are available in the Kokkos documentation.