when I run without valgrind of course, it runs faster so I don't need 4 minutes ... in less than 1 minute, it went from ~ 64 Mbytes to ~ 400 Mbytes --- trying with GC.collect() and GC.minimize() in main loop ... : Same result. hmm..