scott@mcs-server.gac.edu (Scott Hess) (04/28/91)
In article <7e1G6bu.1@cs.psu.edu> melling@cs.psu.edu (Michael D Mellinger) writes: In article <SCOTT.91Apr27094820@texnext.gac.edu> scott@texnext.gac.edu (Scott Hess) writes: 8MB RAM is still going to be a plus. The main reason people should have 16M on a NeXT is because DPS takes up alot of RAM! Right now, I've not got much running on my NextStation (Stuart, Edit, Background, IconBounce), so there shouldn't be too many windows, right? The RSIZE of windowserver from ps aux reports 2.74M, the VSIZE 20.0M. So more RAM would be a plus. Besides, RAM is getting cheaper as we speak . . . The output of ps has to be wrong. Look at the values for csh on the NeXT. I agree that it is misleading, but I've found for processes using a lot of data, it's fairly good. Awhile back I was testing some new memory management routines for Stuart, and the basic idea was that they were supposed to use somwhere around 1/2 of the VM the old stuff used. I kept finding that the difference was more like 4/5 of the usage - but I knew that my overhead was less, and combining that with malloc overhead should make the win be greater (the new stuff doesn't use malloc in places the old did). Anyhow, what I found was that running Stuart and not generating any output left me with a VSIZE about 2x the size of the executable. I'm not really sure what to make of this - Stuart is a NextStep app, so I don't really have a lot of control over what the AppKit is doing with memory, but in this case it shouldn't be much - the menus, and a window should be fairly small. But, to get to the point, once I'd accounted for the apparently bloated starting size, the memory usage reported by VSIZE was almost exactly what I had expected. So, apparently VSIZE has at least some meaning in the real world! Later, -- scott hess scott@gac.edu Independent NeXT Developer GAC Undergrad <I still speak for nobody> "Simply press Control-right-Shift while click-dragging the mouse . . ." "I smoke the nose Lucifer . . . Banana, banana."
melling@cs.psu.edu (Michael D Mellinger) (04/28/91)
In article <SCOTT.91Apr27094820@texnext.gac.edu> scott@texnext.gac.edu (Scott Hess) writes:
I'd still want a disk drive. Makes it much easier to do software
updates - remember that NextStep1.0 couldn't do -NXHost to a
2.0 system, so those would be needed even in the windowserver.
8MB RAM is still going to be a plus. The main reason people
should have 16M on a NeXT is because DPS takes up alot of RAM!
Right now, I've not got much running on my NextStation (Stuart,
Edit, Background, IconBounce), so there shouldn't be too
many windows, right? The RSIZE of windowserver from ps aux
reports 2.74M, the VSIZE 20.0M. So more RAM would be a plus.
Besides, RAM is getting cheaper as we speak . . .
The output of ps has to be wrong. Look at the values for csh on the
NeXT.
USER PID %CPU %MEM VSIZE RSIZE TT STAT TIME COMMAND
melling 3204 0.3 2.3 1.34M 384K p8 S 0:01 -csh (csh)
-rwxr-xr-x 1 root wheel 106496 Oct 22 1990 /bin/csh*
Now compare this with tcsh a SPARC station running Sun OS 4.1.
USER PID %CPU %MEM SZ RSS TT STAT START TIME COMMAND
melling 1840 0.0 2.4 128 264 p1 S 06:27 0:01 -tcsh (tcsh5.18)
-rwxrwxr-x 1 flee staff 212992 Jul 17 1990 tcsh5.18*
sunws8<!>% size tcsh5.18
text data bss dec hex
188416 24576 17008 230000 38270
sunws8<!>% file tcsh5.18
tcsh5.18: sparc demand paged dynamically linked executable
Now tcsh contains a lot more features than NeXT's version of csh(CMU
shell), and it doesn't take up nearly as much memory on the Sun.
Notice that it is also using shared libaries.
So what gives?
Eric Scott told me that using /usr/etc/vmmprint would give me the
answer, but I haven't really looked at it yet(I noticed that there
isn't a man page for it).
-Mike