Bradley Dice
Bradley Dice
To address the original topic (documenting best practices for pool allocators), I think the original post has the right questions that users will want to understand. @wence- @vyasr @harrism Do...
Related: we should document that the pool allocator does not support compute-sanitizer memchecks. More information in https://github.com/rapidsai/rmm/issues/1899.
More related issues: #1969, #2039.
I opened #1877 as a pre-requisite for this work. This is at the top of my RMM priority list for 25.06.
Currently, some parts of RMM are `constexpr` or `inline`, because it was designed as a header-only library. @vyasr Do you (or others) have opinions on whether existing `constexpr`/`inline` logic should...
The consensus from @wence- and @davidwendt seems to be that we can safely drop `constexpr` / `inline` in those files, and move as much as possible to the precompiled library....
Yes, drivers are unlikely to change in minor releases, but there are other reasons why some of the ideas mentioned above may still be relevant. Before closing this, it'd be...
@AjayThorve I think this is solved, see: - https://github.com/rapidsai/cudf/issues/14963 - https://github.com/rapidsai/docs/pull/470
@miscco I don't think RAPIDS should adopt CCCL algorithms until they are publicly exposed. Happy to discuss more offline.
@mroeschke There are a few more CI failures here like: ``` =========================== short test summary info ============================ FAILED tests/test_geoseries.py::test_memory_usage_large - assert 216789 == 216793 + where 216789 = () +...