Releases: NCAR/SPERR
v0.8.5
v0.8.4
v0.8.3
This release contains a series of performance optimizations, mostly enabled by memory improvements and SIMD instructions. The amount of performance improvement is up to 10% in 3D cases.
The SIMD instructions used are from the AVX2 family. Compilation with AVX2 is enabled by default. To turn it off (e.g., on ARM CPUs), one can use the -DENABLE_AVX2=OFF option during CMake.
v0.8.2
v0.8.1
This release contains mostly miscellaneous changes, including a small patch to CLI11 that removes a warning during CMake. However, there's also a small change on how the levels of wavelet transform is to be performed given an array length, some compressed data cannot be correctly decompressed using this version of code anymore.
v0.8.0
v0.7.2
v0.7.1
v0.7.0
This is a major release relative to v0.6.x series, thus the compressed bitstreams are not compatible. This release focuses on runtime performance improvements, with the most significant one being a novel implementation based on integer operations. Compared to v0.6.x series, this release is 20% to 50% faster.
A new functionality in this release is *Progressive Access" on 3D SPERR bitstreams. See this page for more discussion.