Michael Schellenberger Costa
Michael Schellenberger Costa
NOTE to reviewers: I have pushed the concept emulation into its own branch and targeted this PR to that branch. Consequentlly, we can review only the `memory_resource` part of the...
@griwes I pushed a small change to the interface of `{async_}resource_ref` after your review. We now also accept `allocate(size)` and `allocate_async(size, stream_ref)` and default to `alignof(max_align_t)` for the alignment in...
yeah every default we choose will be the wrong one. That said, from my perspective there are 2 pretty fine solutions for users: 1. Provide the desired alignment. 2. Ignore...
@wmaxey Is this the right way to test host only functionality or do I need to put them into the `heterogeneous` subfolder?
Notably, this does not include any of the ranges work. We can only add that once we added ranges support.
> It seems like there are more changes to the upstream implementation than I'm used to seeing for porting a header for libc++. Could you update the PR description with...
> > Yeah, `` underwent a _ton_ of last minute changes. > > Ah, so this is because our upstream copy of libc++ is a bit old. That makes more...
> Should we consider keeping a copy of the original libcxx tests in `libcxx/test/std/...`? > > I think it's an important piece of coverage. I did not test the old...
run tests
> @miscco so what's your current thinking about the steps we'd take to transition rapids libraries? > > Ultimately we want to get rid of the `device_memory_resource` base class all...