garyo@wdl1.UUCP (gary h ogasawara) (05/25/88)
24 May 88 We're running Knowledge Craft 3.1 (languages only) with Sun Common Lisp 2.1.1 on a Sun 3/60 w/ 20 MB which is a client. We have a very large application that must run fast. Most of the code is in Lisp (rather than Prolog or OPS which KC provides). I would like to know more about how the lisp memory management works, and whether we can adjust the memory management parameters to speed the system up. I tried a simple experiment of increasing the dynamic areas from 40 segments (64K = 1 seg) to a larger number, thinking that this would reduce the time spent garbage collecting. However, the larger the dynamic areas were the slower the times I got. I assume this is because more copying between dynamic areas must be done. There is also a reclamation ratio parameter which specifies the desired ratio of free to total dynamic storage. The default is 0.33, and adjusting that number upwards to 0.5 or 0.6 also seems to result in a slower time. So...the bottom line: what setting of parameters will result in the fastest execution? And are there any other ways that we can speed up the system? thanx, /gary h ogasawara garyo@ford-wdl1.arpa ...!sun!wdl1!garyo