[FEA] Add support to extract the exponents of tensor with float / complex values
Philip0116 opened this issue · comments
Similar to __device__ float frexpf ( float x, int* nptr ) method that can applied to a tensor, which returns the exponents of a tensor with float values.
It's even better if matx can support this feature for complex values.
This will help me evaluate the dynamic range of my data and process necessary normalization on my data.
Fixed in #609