Skip to content

Backends

NDSL has a number of "backends" which use can choose from when you execute your code. The backend can be thought of as the fuel which powers the engine of NDSL. Different fuels have pros and cons which may make them more appealing to different situations:

@someone-who-knows-this-better-than-me please write a pros/cons list (or tell me where I can find this information and I'll do it)

  • numpy
  • dace:cpu
  • dace:gpu
  • gt:cpu_ifirst
  • gt:cpu_kfirst
  • gt:gpu_ifirst
  • gt:gpu_kfirst
  • debug: This backend is used for NDSL development, and does not support GPU execution or optimized performance. This should not be used unless you are actively developing NDSL.

Looking Backwards to Move Forward

review blurb

Next, we will present a number of patterns commonly seen in weather and climate modeling, and provide examples of how they are implemented in NDSL.