[FEA] Support LZ4 compression in Parquet readers and writers #14495
Labels
0 - Backlog
In queue waiting for assignment
cuIO
cuIO issue
feature request
New feature or request
libcudf
Affects libcudf (C++/CUDA) code.
Milestone
Is your feature request related to a problem? Please describe.
LZ4 is a general-purpose compression algorithm that is known to be quite efficient on GPUs. This codec is expected to produce faster decompression speeds than Snappy on both CPU and GPU systems. For this reason, LZ4-compressed Parquet files have become more prevalent in recent years. We should support LZ4 compression in cudf.
Describe alternatives you've considered
Use Pandas to read in LZ4-compressed files, and convert to cudf after IO.
The text was updated successfully, but these errors were encountered: