Re: [BProc] Two problems with CM3 (NFS and MPI)

[Jack N. <jj...@pa...>]

On Fri, Dec 13, 2002 at 03:44:35PM -0700, Wilton Wong wrote:
> We're seeing the same problem.. we found a quick work around was to link
> everything statically.. Michal thinks there might be a race condition in
> mpich.. anyways link your executables with gcc --static or g77 --static 
> and see if that helps..
> 
> - Wilton
> 
> On Thu, 12 Dec 2002, Jack Neely wrote:
> 

Tried that but no joy here.  I'm running a really simple hello world
program and I can't make it run on more than 1 node.  All it does is
MPI_Init(), print out rank and size, MPI_Finalize().

[root@beo-test work]# make
/usr/mpich-p4/bin/mpicc -c -ggdb -Wall -DDEBUG  -o hello.o hello.c
/usr/mpich-p4/bin/mpicc -o hello hello.o --static
[root@beo-test work]# ldd ./hello
        not a dynamic executable
[root@beo-test work]# ./hello
p0_11539:  p4_error: init_p4_brdcst_info: my master indx bad: -1
    p4_error: latest msg from perror: Success
[root@beo-test work]# mpirun --p4 --np 1 -d ./hello
listen: n-1 32871
  0    0 11541 192.168.1.10    32773
I am processor 0 of 1
[root@beo-test work]# mpirun --p4 --np 2 -d ./hello
listen: n-1 32873
  0    0 11545 192.168.1.10    32777
  1    1 11547 192.168.1.11    32788
I am processor 0 of 2
[root@beo-test work]# mpirun --p4 --np 4 -d ./hello
listen: n-1 32875
  0    0 11551 192.168.1.10    32785
  1    1 11553 192.168.1.11    32793
  2    2 11554 192.168.1.12    32778
  3    0 11555 192.168.1.10    32787
I am processor 0 of 4
[root@beo-test work]#

I get something very similar to what was reported in the first email
running /usr/mpich-p4/examples/cpi.

Is there any other debugging information that I can provide?

Jack Neely                          

-- 
Jack Neely <sl...@qu...>
Linux Realm Kit Administration and Development 
PAMS Computer Operations at NC State University
GPG Fingerprint: 1917 5AC1 E828 9337 7AA4  EA6B 213B 765F 3B6A 5B89