May 18, 2023 by Samuel Li

VAST staff presents the latest research in lossy data compression at IPDPS

VAST project scientist Samuel Li presented his latest research on lossy data compression at the 37th IEEE International Parallel & Distributed Processing Symposium, or IPDPS, in St. Petersburg, Florida. In this presentation, Sam described high-level designs of SPERR, a wavelet-based scientific data compressor, and also presented results comparing SPERR against other leading scientific data compressors (ZFP, SZ, TTHRESH, and MGARD). These results showcase the superior compression efficiency achieved by SPERR. The corona emission visualized above illustrates the effect of SPERR compression: after compressing 32-bit floating-point values to 1.4 bit-per-point (a reduction factor of 22X), there’s no difference in the rendering.

IPDPS is the flagship conference sponsored by the Technical Community on Parallel Processing (TCPP) of IEEE Computer Society. The slide deck of this presentation is publicly available here, and the published paper containing more technical details is accessible here. Finally, SPERR is open-source software (Github repo).