I am having some problems running on a SMP Linux system. The system I have
is a Gateway Dual 300Mhz Pentium II system with 512MB of RAM. I have
installed Redhat Linux 5.2 and updated the kernel to 2.2.4. I have enabled
SMP and have added Tim Hockin's PSET patch. My program is doing the
following activites: I have a Motif GUI that controls a real-time program
which is controlling a 5-axis hydraulic flight table. The flight table is
being updated at 1000Hz. The real-time program reads the rdtsc and executes
the code necessary to update the table, then spins on the rdtsc until it
reaches 1 millisecond, then repeats the cycle. This goes on for 15 seconds.
The communication to the flight table is via a PCI reflective memory card.
I wrote a module character driver that allows the user to open, mmap and
close the device to get access to the memory. The actual time the program
takes to do its I/O and computation is less than 100 microseconds.
Now for the question / problem:
If I set my real-time program to be SCHED_FIFO at max priority, the X
windows GUI and mouse on the computer cease to function. This happens
whether or not I am making sysmp calls (the Tim Hockin stuff). Shouldn't
the scheduler be smart enough to place my on one of the processors and still
allow the X stuff to work? I know this is how other SMP operating systems
work. It seems as if the scheduler is not SMP aware and is allowing my job
to completely take over the entire system. Even if I use the sysmp calls
and isolate one CPU and place my code on that CPU, I get the same results.
Tim Hockin is looking into this as well. I also wanted to email it here
because I think it is an SMP issue and not just a PSET issue.
Thanks,
Anthony Mayhall
[EMAIL PROTECTED]
[EMAIL PROTECTED]
Boeing Phantom Works
US Army Missile RDEC - Advanced Simulation Center
-
Linux SMP list: FIRST see FAQ at http://www.irisa.fr/prive/mentre/smp-faq/
To Unsubscribe: send "unsubscribe linux-smp" to [EMAIL PROTECTED]
