On Sat, Mar 27, 2004 at 04:06:17PM -0800, Peter G wrote: > hi all, read thru the archives, esp thru thread: "PAE > testing results ..." from circa May 03. > > I am getting a diffrent problem on a uniProc machine. > with 5.2.1 and a stock PAE build. > > machine is stable and does not crash, but the big mem > user job we want to submit to it, all these jobs die > when they start using about 400MB of mem, swap doesn't > activate. the user submitted job just core dumps. > With previous 4.5 FBSD machines with the following > kernel options compiled in (before PAE days) on same > machine w/same 1.5GB of physical mem and plenty of > swap, we'd get 1.2GB of usable mem for user job and > rest would be kernel space. Job would run but not > satisfactory, thought modern PAE would be a better > bet. > > The kernel options we had used w/ 4.5 back then were: > options MAXDSIZ="(1664*1024*1024)" > options DFLDSIZ="(1664*1024*1024)"
PAE is for addressing >4GB of memory; since you only have about 1.5GB, that's not applicable. You'll need to include more details about your system and kernel configuration, including error messages showing what is going wrong. Kris
pgp00000.pgp
Description: PGP signature
