Sorry, you need to enable JavaScript to visit this website.

ndzip: A High-Throughput Parallel Lossless Compressor for Scientific Data

Primary tabs

Citation Author(s):
Fabian Knorr, Peter Thoman, Thomas Fahringer
Submitted by:
Fabian Knorr
Last updated:
3 March 2021 - 1:34am
Document Type:
Presentation Slides
Document Year:
2021
Event:
Presenters Name:
Fabian Knorr

Abstract 

Abstract: 

Exchanging large amounts of floating-point data is common in distributed scientific computing applications. Data compression, when fast enough, can speed up such workloads by reducing the time spent waiting for data transfers. We propose ndzip, a high-throughput, lossless compression algorithm for multi-dimensional univariate regular grids of single- and double-precision floating point data. Tailored towards efficient implementation on modern SIMD-capable multicore processors, it compresses and decompresses data at speeds close to main memory bandwidth, significantly outperforming existing schemes. We evaluate this novel method using a representative set of scientific data, demonstrating a competitive trade-off between compression effectiveness and throughput.

up
0 users have voted:

Dataset Files

Presentation Slides

(49)