Query regarding GPU cache L1 and L2 for Xavier

Is there a GPU cache implementation available publicly.
Is there a way to inject bit-flip in any cache address that is being used in current computations and see if ECC could correct that bit flip?

Hi lavinams13,

We’re checking this requirement with our internal team, we will do the update after confirmed.

ECC for GPU Cache is not enabled in Default HW.
Customer can burn the fuse to enable it.