Suchir Kavi

We've taken to maintaining a second struct for each Message we decode with prost. Typically the only difference is that it removes most Options. We decode with prost, then immediately...

my primary theory is that for each poll, all futures are iterated over regardless of whether they have been woken or not. iirc FuturesUnordered does some more complicated stuff there...

> > my primary theory is that for each poll, all futures are iterated over regardless of whether they have been woken or not. iirc FuturesUnordered does some more complicated...

> > > my primary theory is that for each poll, all futures are iterated over regardless of whether they have been woken or not. iirc FuturesUnordered does some more...

> Thanks so much for submitting this! - Despite not doing the best in these benchmarks yet, I think they're valuable because they allow us to learn where we have...

@yoshuawuyts this may have fallen through the cracks. I believe I've addressed your initial concerns.

Hey @mkhadatare I'm seeing the exact same issue on one system, but not on another. I am using `nppiDistanceTransformPBA_8u32f_C1R_Ctx`. System with issue: RTX 4070 CUDA/NPP version: 11.8 System without issue:...

This is internally tracked at 4832970 btw.

This issue has also been posted to the Nvidia Developer Forums: https://forums.developer.nvidia.com/t/xid-31-error-when-two-cudagraph-captured-executioncontexts-are-executed-concurrently/302553/1

https://github.com/NVIDIA/TensorRT/issues/3633 sounds very similar. @zerollzeng @oxana-nvidia any chance we could get confirmation on this being the same issue? And if so, is there any news on a fix?