Vukasin Milovanovic
Vukasin Milovanovic
my two cents: cudf's version forces a pinned bounce buffer because there is a pinned pool by default. I don't think rmm should default to a pinned allocation for the...
There might be two pieces required here: 1. A default host resource (global function that returns a `rmm::host_resource_ref`, with the default probably being a `rmm::mr::new_delete_resource`. 2. Changes to rmm::device_scalar, including...
> I was thinking we could add an optional host_resource_ref to the device_scalar constructor without needing to add a global host resource. Doesn't it need a default value to be...
To avoid this global default, we could add a constructor overload that takes both device and host `mr`s without default values. Keep the calls unambiguous. We would need to specify...
Is there an issue for the equivalent feature for `rmm::device_scalar`?
I would be okay with letting this PR move along. Should not be too hard to recreate the repro even without the element function.
/ok to test 04e5415
/ok to test 9e7c8db
> One thing I had hoped to do about three months ago was to add a small piece of documentation in the markdown: [06ae1d2#diff-6cc9f4135444fd2ed9062e7168a007f382c2c855fccc18e859834a5d1a6de278R655](https://github.com/rapidsai/cudf/commit/06ae1d2cdcede2ab6332da6b22bef88648301ebd#diff-6cc9f4135444fd2ed9062e7168a007f382c2c855fccc18e859834a5d1a6de278R655) > > * I'm not sure...
> @vuule Would you please describe the next steps before we are ready to merge? I think the code changes can be merged as they are now, I'm just waiting...