The multi-process compression ratio issue of sz2 and sz3.

[BingluCS]

Hi, when I use OpenMP to run the SZ compressor, I obeserve sz2 has a very inefficient compression ratio, while sz3 maintain the normal level. In my cusz experiment, I found the cusz doesn't implement the final lossless compresssion (zstd or zlib). Does the sz2 hvae same issue, but why sz3 achieve the function?

[disheng222]

SZ2 has largely different design compared with SZ3, especially about data prediction stage. Yes. The compression ratio should be relatively lower in openMP for SZ2. SZ2’s speed should be higher than SZ3 though.

…

Sent from my iPhone BingluCS ***@***.***>于2024年6月1日 周六下午8:13写道：

Hi, when I use OpenMP to run the SZ compressor, I obeserve sz2 has a very inefficient compression ratio, while sz3 maintain the normal level. In my cusz experiment, I found the cusz doesn't implement the final lossless compresssion (zstd or zlib). Does the sz2 hvae same issue, but why sz3 achieve the function? — Reply to this email directly, view it on GitHub <https://github.com/szcompressor/SZ/issues/117>, or unsubscribe <https://github.com/notifications/unsubscribe-auth/ACK3KSPSRFCG7GVCTAUJP6DZFG3F3AVCNFSM6AAAAABIUDKIO2VHI2DSMVQWIX3LMV43ASLTON2WKOZSGMZDSMJQHE2TINQ> . You are receiving this because you are subscribed to this thread.Message ID: ***@***.***>