Squeak and parallel computation
Bert Freudenberg
bert at isg.cs.uni-magdeburg.de
Tue Oct 8 15:18:52 UTC 2002
On Tue, 8 Oct 2002, Ian Piumarta wrote:
> On Mon, 7 Oct 2002, Aaron J Reichow wrote:
> > On Tue, 8 Oct 2002, Ian Piumarta wrote:
> > > if your VM runs on one processor and your X server on another then
> > > you should see an improvement for graphics-bound code.
>
> I can get the ratio to 5:1 Squeak:X by scrolling a large window of text.
>
> Every time Squeak does screen (or keyboard/mouse) i/o it's fighting with X
> server for resources. One of the "improving performance" hints in the X11
> FAQ is to run clients on a different machine, which is the same thing as
> "one of your other processors" modulo the network. (This doesn't work
> well for Squeak since it stores the display as a client-side image and the
> network bandwith goes through the roof when you try to run it remotely.
> OTOH, running it on one of your other processors should yield the same
> advantages as running remotely but without any of the disadvantages of
> network saturation -- particularly when using shared memory to send the
> image to the server.)
>
> But I'm only guessing. Maybe somebody with a SMP machine can testify from
> experience.
I'm running on a dual-processor PIII 600MHz here. Scrolling a text window
utilizes one CPU to almost 100%, X is using the second processor:
bert.balloon ~ > squeak -xshm -display unix:0 &
bert.balloon ~ > top
PID USER PRI NI SIZE RSS SHARE LC STAT %CPU %MEM TIME COMMAND
2711 bert 11 0 12820 12M 4268 1 R 99.8 5.0 1:11 squeak
1441 root 9 0 147M 19M 9132 0 S 16.2 7.6 6:25 X
Of cource, doing it without shared memory increases the load for X:
bert.balloon ~ > squeak &
bert.balloon ~ > top
PID USER PRI NI SIZE RSS SHARE LC STAT %CPU %MEM TIME COMMAND
2772 bert 14 0 13504 13M 1076 0 R 90.5 5.3 1:15 squeak
1441 root 10 0 144M 16M 6080 1 R 33.6 6.4 7:17 X
-- Bert
More information about the Squeak-dev
mailing list
|