Compiling OpenMP for GPUs with there are function calls in the parallel regions

Is having "device functions" not supported yet in OpenMP ?

It is supported with the ‘declare target’ decorator.

Arpith