Giters

NVIDIA / MatX

An efficient C++17 GPU numerical computing library with Python-like syntax

Home Page:https://nvidia.github.io/MatX

Geek Repo:Geek Repo

Github PK Tool:Github PK Tool

[FEA] Investigate mixing transforms and operators

cliffburdick opened this issue · comments

commented

Right now to do an FFT convolution at least three separate lines must be written:

  1. FFT input
  2. element-wise multiply the output
  3. IFFT the multiplied results

This limitation is due to how we treat transforms and operators differently. This issue is to investigate mixing the two, which has many benefits, including saving temporary variables, more readable code, and more chances for optimizations in the parsing.