Vitis HLS Implementation of a Line Intersection
A basic problem in many physical simulations is to determine how close points in 3D are to simple shapes. This problem arises, for example, in robotics for collision detection. In this lab, we consider a simple version of this problem where we build a Vitis HLS IP to compute the distance of 3D points to a line segment.
In going through this lab, you will learn to:
- Describe command and response jobs to an IP with Data Schemas
- Build a python Golden model for a loop
- Implement a pipelined loop of the same computation in Vitis HLS
- Design memory layouts and nested loops for optimal pipelining
- Synthesize the design
- Compare expected results with the Golden model for functional validation
- Parse C synthesis reports to determine the latency and initiation interval (II) of the design
Files
All files for the lab can be found in this repo in:
hwdesign/labs/intersect
In this directory, you will find:
intersect/
|-- run_hls.tcl // TCL script for simulation and C synthesis
|-- run_tests.py // Submission program
|-- line_inter_tb.cpp // Vitis HLS testbench
`-- partial/
|-- line_inter.ipynb // main jupyter notebook
`-- line_inter.cpp // Vitis kernel
Copy all the files in the intersect/partial directory to intersect.
Go to building the python model