Hi,
I don’t know what is DLA.
In tensorRT, is it right that i can use dla engines by using setDLACore() function?
And then, Can I get more performance than do not use?
I don’t know exactly what is DLA, and how to use it.
Thanks.
Hi,
I don’t know what is DLA.
In tensorRT, is it right that i can use dla engines by using setDLACore() function?
And then, Can I get more performance than do not use?
I don’t know exactly what is DLA, and how to use it.
Thanks.
Hi,
DLA is new deep learning hardware can leverage GPU workload.
You can enable DLA with this function:
// Build
builder->setFp16Mode(true);
builder->setDefaultDeviceType( nvinfer1::DeviceType::kDLA );
builder->setDLACore(0);
// Inference
infer->setDLACore(0);
Here are some introduction and document for your reference: http://nvdla.org/
Thanks.
Hi,
can i use first tensorrt engine in gpu and second tensorrt engine in dla at the same time?
Thanks.
Hi,
It’s required to launch the engine in a single process for the same model.
But you can run the different model in DLAs and GPU at the same time.
Thanks.