Mark Harris comments

Results 341 comments of


                                            Mark Harris

trafficstars

[FEA] Consider creating a pinned host memory resource that can leverage all of the allocation strategies

This is still relevant, but now depends on https://github.com/NVIDIA/libcudacxx/pull/105

[FEA] Consider creating a pinned host memory resource that can leverage all of the allocation strategies

If you proceed ahead of `cuda::memory_resource` this will just create more for us to refactor.

[FEA] Consider creating a pinned host memory resource that can leverage all of the allocation strategies

I thought we were close before Christmas. But we keep running into more design issues.

[skip gpuci] Add GitHub Actions workflow for PRs

Should we push this to 22.10? Still a draft as we are entering code freeze.

Enable constructing DeviceBuffers using a non-default memory resource and correctly manage memory resource lifetime in DeviceBuffer

@shwina can we bump this to next release?

[REVIEW] Add NVTX ranges to RMM

Still waiting on NVTX headers release.

[FEA] Priority support in cuda_stream_pool

@seunghwak have you prototyped this with raw CUDA streams to verify that you get the benefits you expect?

[FEA] Priority support in cuda_stream_pool

Can you share some specifics (speedups) here to help motivate?

[FEA] Python interface for `arena_memory_resource`

Two other points: 1. I believe arena_memory_resource uses separate read and write locks which may enable more concurrency between host threads. We can try something similar in `pool_memory_resource`. 2. Just...

[FEA] Add a argument or flag to control whether the constructor and operator= should be async

I think the intended design is that everything is async and stream-ordered. A user is free to add a synchronous wrapper if needed, but in general in RAPIDS we need...