Vukasin Milovanovic
Vukasin Milovanovic
CC @GregoryKimball @nvdbaranec
Parquet reader benchmarks (partial) show clear signal compared to pageable input: ``` | io_type | compression_type | cardinality | run_length | Samples | CPU Time | Noise | GPU Time...
> @vuule is this PR waiting on anything? Just noticed it from #12926. Once I saw the CI failing with this change I got really discouraged to mess with this....
> This is... very weird. You seem to have gotten extraordinarily unlucky because all of the failures that I can see appear to be either unrelated issues that have since...
There it is `FAILED tests/test_orc.py::test_orc_reader_apache_negative_timestamp - AssertionError: DataFrame.iloc[:, 0] (column name="v") are different`
@bdice should I modify the PR to try $TZDIR (if set) before the system dir?
> This looks fine. I didn't see any test or performance regressions or improvements with this patch. AFAIK the new path is not (yet) use by default.
> On GH200, the `brc_pipeline` example seems to be calling `cudaHostRegister` on the entire file for each chunk. That unexpected. I'll check the code and update here.