How to configure Hierarchical QoS offloading in connectX-6 NIC?

As it is stated in the DPDK website, offloading TX scheduling to hardware is supported.

I have a simple DPDK application for TX traffic scheduling (shaping) which for now works in software. I would like to use the hardware offloading to perform this function. Any help is appreciated.

Could you shed more light on exactly what need to be offloaded/configured? Is this something that covered bo QoS testing -