cuIO
E890462
cuIO is the RAPIDS GPU-accelerated input/output library that enables fast reading and writing of data formats like CSV, Parquet, and ORC directly on NVIDIA GPUs.
Statements (45)
| Predicate | Object |
|---|---|
| instanceOf |
GPU-accelerated input/output library
ⓘ
RAPIDS library component ⓘ |
| benefit |
high-throughput data ingestion
ⓘ
lower end-to-end latency for data loading ⓘ reduced CPU-GPU data transfer overhead ⓘ |
| componentOf | RAPIDS data processing stack NERFINISHED ⓘ |
| designedFor | GPU-accelerated data I/O ⓘ |
| developedBy | NVIDIA NERFINISHED ⓘ |
| documentation | https://docs.rapids.ai/api/cudf/stable ⓘ |
| feature |
direct GPU memory data loading
ⓘ
parallel parsing on GPU ⓘ support for various compression codecs depending on format ⓘ zero-copy or reduced-copy data paths where possible ⓘ |
| goal |
enable end-to-end GPU data pipelines
ⓘ
minimize CPU bottlenecks in data I/O ⓘ |
| integratesWith |
RAPIDS Accelerator for Apache Spark
NERFINISHED
ⓘ
cuDF NERFINISHED ⓘ |
| license | Apache License 2.0 (via RAPIDS) NERFINISHED ⓘ |
| optimizedFor | columnar data formats ⓘ |
| partOf |
RAPIDS
NERFINISHED
ⓘ
RAPIDS cuDF NERFINISHED ⓘ |
| programmingLanguage |
C++
ⓘ
CUDA NERFINISHED ⓘ |
| repository | https://github.com/rapidsai/cudf ⓘ |
| runsOn | NVIDIA GPUs NERFINISHED ⓘ |
| supportsEnvironment |
cloud GPU instances
ⓘ
on-premise GPU servers ⓘ |
| supportsFormat |
CSV
ⓘ
JSON (in some RAPIDS releases) ⓘ ORC ⓘ Parquet NERFINISHED ⓘ |
| supportsIntegration |
Apache Spark via RAPIDS Accelerator
ⓘ
Dask-cuDF NERFINISHED ⓘ Python data science workflows using cuDF ⓘ |
| supportsLanguageBinding |
C++
NERFINISHED
ⓘ
Python NERFINISHED ⓘ |
| supportsOperation |
compression and decompression
ⓘ
read ⓘ write ⓘ |
| targetUser |
ML engineers
ⓘ
data engineers ⓘ data scientists ⓘ |
| useCase |
ETL pipelines on GPUs
ⓘ
data science workloads on GPUs ⓘ machine learning data preprocessing ⓘ |
Referenced by (1)
Full triples — surface form annotated when it differs from this entity's canonical label.