[UR][L0v2] Restrict events reset during command buffer enqueue

[Xewar313]

No description provided.

[EwanC]

Do we have a UR CTS test already to stress this? Something where we create an out-of-order command buffer, with every command returning a sync-point, and then only use a subset of those sync-points to as dependencies in other commands.

[pbalcer]

Do we have a UR CTS test already to stress this? Something where we create an out-of-order command buffer, with every command returning a sync-point, and then only use a subset of those sync-points to as dependencies in other commands.

Sounds like this should be a benchmark, not a CTS test.

[Xewar313]

Yea, I agree with Piotr, and I don't think that we have such benchmark yet

[EwanC]

Sounds like this should be a benchmark, not a CTS test.

Yeah that's fair, my interest is more making sure that CI is verifying that the patch is functionality correct before it merges than knowing the perf benefit. But we have lots of SYCL E2E tests that create an out-of-order command-buffer with a sync point for every node, but won't use the sync points returned for leaf nodes, so this probably is being tested.