Chris Ryan (ryan++at++glut.angel.com)
Fri, 31 Jan 1997 09:36:32 -0800 (PST)
The workaround, later blessed by sgi rep, was the following C/C++ program
run in the background. Don't laugh, it worked for us and sgi hotline
finally ended up recommending that this is the "fix" until 6.X addresses
kernel/scheduler bugs.
extern "C" int sginap(long);
main(int, char**) {
while (1) {
sginap(0);
}
}
If you read the man pages on sginap(), you can see why this might work if
in fact the problem is a scheduler bug. It definitely changes your
gr_osview output so that you don't have an idle CPU. For us, draw times
recovered miraculously.
On Fri, 31 Jan 1997, BOCCARA Michael wrote:
>
>
> Hi,
>
> I have a problem of graphics performance on an Onyx RE2 when I run a
> Performer application.
> I have already opened a call on the SGI Hotline of Paris, but I would like to
> have an advice from one of you pfusers, or a personnal answer from a person
> of the Performer technical staff to help me.
>
> We have 2 Onyx RE2 with each 1 RM4, and 2 CPUs. We are working (still) with
> IRIX 5.3, and Performer 2.0.1
>
> When I am running any Performer application, even perfly, both machines don't
> give the same graphics performance. The drawing time is at least twice longer
> on one Onyx than on the other.
> I guess there is a hardware problem on the bad machine, or a problem of
> configuration.
>
> But I have a strange phenomenon, that could give you a clue to help me
> searching in the good direction. I have looked at the system activity with
> gr_osview while running my Performer application.
> In the good Onyx, the CPU activity is almost 90 percents occupied by the
> "user" part (the blue one), and the system is automatically alternating
> between the 2 CPUs. I guess this is quite normal.
> In the bad Onyx, the CPU activity is half shared by the "sys" part (the red
> one), and there is no alternating between the two CPUs ; only one CPU is
> working during one simulation. It seems like the system is fully running when
> it deals with graphics display, but I 'm not a hardware specialist at all.
>
> With what I told you, does any have an idea on where is the problem ? Which
> test I should do to get a finer approach of the bug ? Which trick I could try
> to turn around the bug ?
>
> Anyway I am waiting for the HotLine answer, but thanks in advance to the
> one(s) who will make me go a little forward.
>
> Regards,
> Mike
>
> Michael Boccara - ! Aerospatiale
> Software engineer ! Centre commun de
> email : ! recherche Louis
> Bleriot
> michael.boccara(a)siege.aerospatiale.fr ! 12 rue Pasteur
> Tel : (+33) (0)1 46 97 32 40 ! 92150 Suresnes
> Fax : (+33) (0)1 46 97 32 59 ! France
> ===================================List Archives, FAQ, FTP: http://www.sgi.com/Technology/Performer/
> Submissions: info-performer++at++sgi.com
> Admin. requests: info-performer-request++at++sgi.com
>
=======================================================================
List Archives, FAQ, FTP: http://www.sgi.com/Technology/Performer/
Submissions: info-performer++at++sgi.com
Admin. requests: info-performer-request++at++sgi.com
This archive was generated by hypermail 2.0b2 on Mon Aug 10 1998 - 17:54:30 PDT