Involved code has been merged into SZ v2.1
In ShiftInterval branch, we add a precomputation-based mechanism to accelerating relative-error bounded mode.
In Parallelism branch, we add a parallelism method in point-by-point processing stage to achieve an acceleration.
fork from SZ master at version : bea6aae5e6881716f778b97288f2b8222e93e3bd
Model B
shift interval edition
please add
plus_bits = 3
in config file
Xiangyu Zou, Tao Lu, Wen Xia, Xuan Wang, Weizhe Zhang, Sheng Di, Dingwen Tao and Franck Cappello, "Accelerating Relative-error Bounded Lossy Compression for HPC datasets with Precomputation-Based Mechanisms", in Proceedings of the 35th International Conference on Massive Storage Systems and Technology (MSST '19), Santa Clara, CA, USA, May. 2019.