Create engine and inference with DLAs on Xavier

Autonomous Machines Jetson & Embedded Systems Jetson AGX Xavier
1

I tried to create engine with DLA on Xavier.
But I don’t know how to create. And I would like to use one or two DLAs on Xavier.
(Xavier has two DLA cores)
And I had already known the way to inference with engine which does “NOT” dla cores.
Is there any difference between “with” and “without” DLA ?

Thank you.

3

Hi,

DLA can be used by following function.

https://docs.nvidia.com/deeplearning/sdk/tensorrt-api/c_api/classnvinfer1_1_1_i_runtime.html#a5725940f92ecd5f8c414ced5184a7d08

runtime->setDLACore(0);    // using DLA0

In our sample trtexec, DLA can be used by using –useDLACore=0 or –useDLACore=1.

 $ /usr/src/tensorrt/bin/trtexec ... --useDLACore=0

Thanks.

4

Thank you,AastaLLL.

I try your method.
If I do not set DLA, the all inference is done on GPU. Isn’t it ?

5

Hi,

Yes, the default device is GPU.
Thanks.

6

Hi,

Thanks AastaLLL. I’ll try it.