Some binary file formats use half precision complex numbers (float16 for the real and imaginary parts)

Since numpy already supports half precision floats wouldn't it be possible to support the complex counterpart?