Release 2.1.8.2

A tiny fix:
remove spurious printf for 2D double decompression.
(Previous released version was added a debugging printf in the official code by mistake)

Release 2.1.8.1

Support fortran compile in CMake

Release 2.1.8.0

Add transpose function to SZ (supporting compression along the highest dimension in case of high smoothness of the data in this dimension). To switch it on, you need to modify sz.config.

In addition, sz 2.1.8 supports protecting value range (making sure the value range are not changed at all). You need to modify defines.h and change #define PROTECT_VALUE_RANGE 0 to be 1, and then recompile. PROTECT_VALUE_RANGE is set to 0 by default, because the protect_value_range function costs about 10% time overhead on decompression. It has zero overhead on compression.

Release 2.1.7.0

Compression and decompression speed are both improved prominently. The performance improvement can be observed clearly especially on KNL machines such as ANL Theta supercomputer. Also support double-precision for openmp version.

Release 2.1.6.0

Fix a bug (error couldn't be bounded) happening when setting point-wise relative error bound for 3D double-precision data.

Release 2.1.5.3

Revise CMakeLists.txt to properly detect openmp and remove useless flags

Release 2.1.5.2

Revise CMakeLists.txt such that the header files will be copied to the installation dir during the compilation/installation with cmake.

Release 2.1.5.1

Fix compilation bug for Fortran version
Add the zc version of sz.c (i.e., sz_zc.c) for supporting visualizing decompressed data
Revise Readme

Release 2.1.5.0

Fix some bugs in temporal compression and support random-access variable selection for compression/decompression.

Release of SZ 2.1.4.2

Revise a bug in 2.1.4.2:
The bug happened only when the data size is too small (smaller than a datablock such as 6x6x6) for 2D and 3D cases.