nvdbaranec

Results 28 comments of nvdbaranec

Parquet reader list microkernel

One thing I've been thinking about is maybe splitting this file into two or three pieces. - One cu file containing the core loops for each of the major kernels...

[BUG] TDIGEST_MERGE group by aggregation scales very badly

I don't even think this needs to be a segmented sort(thrust doesn't have one in any case. The sort-by-key functionality is a bit of a misnomer). But, I think if...

[FEA] cudf::column needs a set_stream() function.

I think that's totally fine. The stuff Mike is working on is just experimental stuff as is, and doesn't actually require this from `column` itself.

Optimizations for tdigest generation.

/merge

[BUG] Shuffle performance scales badly with number of columns and number of partitions

Any reason to keep this open, now that kud0 is in place?

[BUG] GpuGetJsonObject does not expand escaped characters

You are correct.

[FEA] Add a custom decoder to convert Parquet dictionary pages into `cuco::static_set` objects

There's some issues with this API. We need considerably more than just the decompressed page data itself. We need quite a bit of information to start decoding page data. See...

[FEA] Add a custom decoder to convert Parquet dictionary pages into `cuco::static_set` objects

@mhaseeb123