NVDLA/sw compilation on Xavier

Robotics & Edge Computing Jetson & Embedded Systems Jetson AGX Xavier
1

Hi.
Trying to use DLA on Xavier. In order to do this I need to build KMD external kernel module from https://github.com/nvdla/sw.
While building I get error:

/home/nvidia/workspace/NVDLA/sw/kmd/port/linux/nvdla_core_callbacks.c:232:8: error: too few arguments to function ‘dma_buf_begin_cpu_access’
  ret = dma_buf_begin_cpu_access(buf, DMA_BIDIRECTIONAL);

which is due to the fact that signature of dma_buf_begin_cpu_access function from dma-buf.h in kernel 4.9 needs 4 arguments, while only 2 are provided here, according to kernel 4.13 version.
How can I build that kernel module?

2

Hi pkolomiets, the DLA engines on Jetson AGX Xavier should be used through TensorRT. Please refer to Chapter 6 of the TensorRT Developer Guide for more information: [url]https://docs.nvidia.com/deeplearning/sdk/tensorrt-developer-guide/index.html#dla_topic[/url]

3

Hello,

Is there any other way to use the DLA engines without TensorRT?

1 Like