LZ4_MIN_PERCENT

#define LZ4_MIN_PERCENT   90

Minimal expected compression.

We will hand lz4_compress_limitedOutput() a buffer which is this percentage of the input buffer size. If compression to this size is not possible, lz4_compress_limitedOutput() will return an error and we will just write out the clear-text data. Usually, small tensors are highly compressible, but large tensor blocks of std::complex<double>, as they occur in large-scale DMRG calculations, are hardly compressible. Not compressing them gives us an advantage on decompression, as that is just another std::memcpy().