Vyas Ramasubramani
Vyas Ramasubramani
@robertmaynard can you take a look here? I've lost track of what decisions we've made where about keeping or removing this particular ABI mangling. I know it's caused headaches with...
Technically rapids-cython has never promised that target names would be generated in any particular way, but in practice the name generation has always been stable and we rely on it...
Replaced by #479.
Hi there, I'm one of the kvikio developers. There's a lot in flux as we improve our tooling here, and you're right to note that things are being moved into...
The main value-add that kvikio provides on top of lower-level cuFile APIs is that it detects GDS availability at runtime and provides transparent fallback to POSIX-based I/O when GDS is...
> Sorry, I missed your last comment here. I think your proposal sounds good. Something like: > > 1. Either prime (or not) the async allocator > > 2. Do...
Re:CCCL, see https://github.com/rapidsai/rmm/issues/1307#issuecomment-2683247073.
> Ah, this is a nasty gotcha. `pread` doesn't hold a reference to `buf`, so you allocate `buf` in your `pread_one` function but it is dropped by the time we...
I took another quick look at this and it appears that the issue is that the determination of when the operation is done has a bug. The underlying C++ `RemoteHandle::pread`...
Clarifying the above: even if you store the rf and buf objects created in the `pread_one` call into global variables or something else that persists them, you still see a...