Easy way to synthesis duplicate FPGA kernel?

[jasonlin316]

I am currently trying to use the oneAPI on DevCloud, not sure how to synthesis duplicate FPGA kernels.

Here is the case, maybe I implement a matrix multiplier, and it only cost 25% of hardware resources on The FPGA board, so I would synthesis 4 matrix multipliers to use more FPGA resources and achieve higher parallelism.

Though it seems like a simple question, I don't know how to do this using oneAPI.

I found a related tutorial: compute_units

But the tutorial is kinda complicated, all these unroller, sink and source... hard to get a clue which line/pragma is the key to implement multiple CU

[AnilErinch_A_Intel]

Hi ,

Please see the link below which explains exactly how to configure OneAPI to run multiple kernel.

https://software.intel.com/content/www/us/en/develop/documentation/oneapi-gpu-optimization-guide/top/kernels/executing-multiple-kernels-on-the-device-at-the-same-time.html

Thanks and Regards

Anil