nvdbaranec
nvdbaranec
@gpucibot merge
Just realized the CI failure was because of actual compile errors in the benchmarks. Fixed.
Followups are still in flight.
Actually, upon further review, this mystery "list" element is in the parquet file itself (it's one of the odd ways in which the spec allows you to specify list columns)....
Right, that's the question: are these details something you'd expect the end user to know or care about, or would they just expect "A.B.C"? Maybe this is a what-would-Pandas-do question.
> @nvdbaranec I'd like your opinion on how to handle the sizes. Do you want me to decode the delta lengths to get exact sizes or do you think the...
Sounds like doing it the accurate way is the way to go. We could live with approx if it was a conservative/overestimate approx, but if it can be too small...
I am more of a fan of having a global function that sets an rmm allocator to be used for pinned allocations. For now we could limit it to cuIO...