Are there any issues with calling enqueueV3 on multiple Streams with a single ExecutionContext?

AI & Data Science Deep Learning (Training & Inference) TensorRT
1

enqueue and enqueueV2 include the following warning in their documentation:

Calling enqueueV2() in from the same IExecutionContext object with different CUDA streams concurrently results in undefined behavior. To perform inference concurrently in multiple streams, use one execution context per stream

enqueueV3’s documentation does not. Should it? Is there any locking or other performance concerns I should be aware of?

2

The documentation for enqueueV3 doesn’t explicitly mention the limitation, but it’s likely still applicable. There hasn’t been any official confirmation of a change in behavior, and it’s safer to assume it works with name-based methods similarly to enqueueV2.