Bob Freeman

I need to compress around 35-40Gb of data every weekend. There are roughly 150 files and each file contains records of variable length - between 21 and 52 bytes. Each file is for one day. Each record starts with 1 byte type and 7 byte of unix datestamp with usecs, hence first 3 bytes are going to be static for each record type (there are 4 types) throughout each file. Types are coming in random order. Distribution of types: Len(b) %% 21 38.3 48 34.9 27 22.9 52 3.9 Compression takes a lot of time...

Thank you! -mmt=4 works - I have now 100% use of all 3 CPUs

I am using this on CentOS 9: 7-Zip (a) [64] 16.02 : Copyright (c) 1999-2016 Igor Pavlov : 2016-05-21 p7zip Version 16.02 (locale=en_GB.utf8,Utf16=on,HugeFiles=on,64 bits,3 CPUs AMD EPYC Processor (830F10),ASM,AES-NI) command: 7za a -mx=9 -mmt=3 .. but average CPU usage is 61%.. There is no discernable disk IO queues and 80% free RAM. You say LZMA uses 3 threads with mx=9 Why cant I achieve higher CPU usage?

I am using this on CentOS: 7-Zip (a) [64] 16.02 : Copyright (c) 1999-2016 Igor Pavlov : 2016-05-21 p7zip Version 16.02 (locale=en_GB.utf8,Utf16=on,HugeFiles=on,64 bits,3 CPUs AMD EPYC Processor (830F10),ASM,AES-NI) command: 7za a -mx=9 -mmt=3 .. but average CPU usage is 61%.. There is no discernable disk IO queues and 80% free RAM. You say LZMA uses 3 threads with mx=9 Why cant I achieve higher CPU usage?