Good afternoon. The task is to rewrite an algorithm for linear two-dimensional spline interpolation from the PC (Python code) into the FPGA based on input data from 128 sensors. Has anyone encountered/done something like this? And what do you think would be better/simplest:
- Try to write everything in VHDL /SystemVerilog.
- Try to write the algorithm in HLS.
- Use 2 FPGAs - one with sufficient input LVDS pins, for example, big package Artix, to receive input data and the second, Zync, to implement the algorithm on Python.
Thanks for the comments and recommendations.
Hi Leo,
As always, it depends! What is the data rate? Do you need to perform real-time calculations, or can you record smaller chunks of data and process them offline without worrying too much about processing time?
I recommend sticking with what you know best. If you’re confident in VHDL, then go with VHDL. If you’re more comfortable with HLS, then use HLS.
However, I would suggest considering a third option (using 2 FPGAs). If you choose this route (with one FPGA handling lots of I/Os), I recommend using a standard PC or NUC instead of a Zynq. You could then transfer the data via USB, serial, or Ethernet for processing in Python or Jupyter Notebook, avoiding the need to configure Linux on the Zynq—unless you already have that setup ready.
Dominik