enqueueV2 return error

[lix19937]

Description

enqueueV2 return error ， What situation will cause this ?

Environment

TensorRT Version: TRT8510

NVIDIA GPU:ORIN-X

NVIDIA Driver Version:

CUDA Version:11.4

[zerollzeng]

What error did you see and can you reproduce it with trtexec?

[lix19937]

@zerollzeng I use mutli-thread to run plan (each thread run a different plan), when a plan (on dla device)

[12/24/2023-14:48:13] [I] Starting inference
[12/24/2023-14:48:13] [E] Error[1]: [cudlaUtils.cpp::submit::95] Error Code 1: DLA (Failed to submit program to DLA engine.)
[12/24/2023-14:48:13] [E] Error occurred during inference

and gpu util is 32% keep unchanged !

[lix19937]

I think drive os 6060 maybe have some bug of DLA

[zerollzeng]

could you please try latest 6080/6090?

[lix19937]

could you please try latest 6080/6090?

Our OTA version is drive os 6060

[zerollzeng]

I remember we have some improvement on this since 6.0.8.0(DLA 3.14), and if you need a patch on 6.0.6 then you need to request help from your company's Tier-1 or other supporters.

[lix19937]

ok, thanks.