Hi,
On Fri May 09, 2014 at 18:08:49 +0900, Irvanda Kurniadi wrote:
I want to ask you about my problem in running mpi-program over L4Linux. I'm usinng ubuntu 12.04 kernel with gnu libc-2.15. I put the library from /lib/i386-linux-gnu/ to /lib in my custom ramdisk. Then I put the library that is compiled in the kernel's host to the /usr/lib in my custom ramdisk. I also put my mpi hello program and my own mpi program which are compiled in the kernel's host (I mean in ubuntu 12.04 with libc-2.15). I'm using MPICH3 for the mpi library and I use gforker process manager. When I run the mpi Hello program, it can run properly (I invoke "mpiexec -n 4 ./mpi_hello and there are 4 output from different 4 different processor, it means the program run properly). Then I run my own mpi application which also include the tiff, jpeg, amd libstep libraries that I put altogether in /usr/lib/ in my custom ramdisk. But I got this error message when I invoke "mpiexec -n 1 ./mpi_program image.tif" or "mpiexec -n 2 ./mpi_program image.tif" command (shown below).
l4x-evict_tasks: 1089 callbacks suppressed l4x-evict: Found no process to free l4x-evict: Found no process to free l4x-evict: Found no process to free INFO: rcu_sched detected stalls on CPUs/tasks: { 1} (detected by 0, t=348877 jiffies, g=4294967184, c=4294967183, q=(5) sending NMI to all cpus:
and I can't interrupt the running process.
I presume this problem happens because the all of process have been finalized but the l4linux kernel still think that there is a remaining process that should be freed. But I'm not sure about this. If you know about this, plis let me know how to fix it and if you can give me an explanation about this I will be really grateful.
I think this an out of resources issue. Is your MPI program generating many processes or similar?
Adam