I was able to quantize resnet50 (tensorflow version) with decent tool and also could generate .elf with dnnc.
But now there is an issue when I try to run .elf file generated from CPU ELF object regarding Hybrid Execution. It doesn't matter if I compile it from SDSoC or DNNDK tool on the board, in both cases I get the error: Failed to open /dev/xlnk
It is clear that xlnk is the SDSoC kernel driver that helps with memory management, accelerator control, and data movement among other things; and this error occurs when the xlnk driver is not configured in the petalinux project or it is not included the xlnk fragment in the devicetree. But just for demo I took the already exising image from here: xilinx-zcu104-prod-dpu1.4-desktop-buster-2019-04-23.img.zip and it is obvious that xlnk is not configured in the petalinux.
Now the question is, how I can add the xlnk to the already existing image?
The next question is, why this problems occurs even when the code is compiled with the DNNDK tool on the board?