04-23-2014 08:29 AM
04-23-2014 10:59 PM
You can use Xilinx HLS with targeting sysgen design . Use the attached code for LU decomposition
LU Decomposition is a well known algorithm for matrix decomposition that is further used for matrix inversion for example.
This design integrates actually 2 different implementations showing how the code must be modified in order to optimize for Hardware implementation.
BasicLU: contains the VHLS project using the standard algorithm.
SimpleLU: contains the VHLS project optimized for Hardware
The PowerPoint file explains precisely why we should modify the code and explore the various solutions.
04-28-2014 06:52 AM
I have implemented matrix Inversion (2x2) in Sysgen very similar to your aproach:
My implementation is very ressource and timing orientated. As you can see i onl yuse one divider (reciprocal 1/X), which I implemented as a lookup table. You can even save more ressources by using Time Domain Multiplexing methods.
I would not use this method however for matrixes above 3x3. Even for 3x3 this gets quite heavy. Instead I would got the LU decomposition route.
05-01-2014 06:29 AM
05-01-2014 06:33 AM