matvis.gpu.gpu_beam_interpolation

matvis.gpu.gpu_beam_interpolation(beam: np.ndarray | gpuarray.GPUArray, daz: float, dza: float, az: np.ndarray, za: np.ndarray, gpu_func: Callable | None = None, nthreads: int = 1024, stream=None, return_on_cpu: bool = True)[source]

Interpolate beam values from a regular az/za grid using GPU.

Parameters:

  • beam – The beam values. The shape of this array should be (nbeam, nax, nfeed, nza, naz). This is the axis ordering returned by UVBeam.interp. This array can either be real or complex. Either way, the output is complex.

  • daz – The grid sizes in azimuth and zenith-angle respectively.

  • dza – The grid sizes in azimuth and zenith-angle respectively.

  • az – The azimuth and zenith-angle values of the sources to which to interpolate. These should be 1D arrays. They are not treated as a “grid”.

  • za – The azimuth and zenith-angle values of the sources to which to interpolate. These should be 1D arrays. They are not treated as a “grid”.

  • gpu_func – The callable, compiled GPU function to use. This is generated dynamically if not given, but can be given to avoid recompilation.

  • nthreads – The number of threads to use.

  • stream – An option GPU stream to write to.

  • return_on_cpu – Whether to return the result as CPU memory, or a GPUArray.

Returns:

The beam interpolated at the sources. The shape of the array is (nbeam, nax, nfeed, nsrc). The array is always complex (at single or double precision, depending on the input).

Return type:

beam_at_src