[mpich-discuss] MPICH2 for ARM

Fri Nov 18 07:10:56 CST 2011

Hi

So with the MPICH2 version 1.3.2 it works fine when I start the 
applications with -launcher fork. This is a big step forward and my next 
step is do start to controllers and let them communicate together. Until 
now it is possible to ping one controller from the other but i can not 
start Mpi via ssh. I installed openssh with make menuconfig in the 
buildroot butI always get the error “connection refused”. Do I need to 
include ssh in the kernel? I think it is a problem with the 
configuration of ssh but i have to lock. If you have any advice please 
write back. Thanks

Harald

On 15.11.2011 00:44, Pavan Balaji wrote:
>
> Well, if you want to use ssh, you obviously need to have it installed.
>
> -- Pavan
>
> On 11/14/2011 07:35 AM, Harald Schuster wrote:
>> Ok I used now the mpich2-1.3.2 version with hydra and start the program
>> with the parameter "-launcher fork" and it works. I don't get a
>> segmentation fault any more and thats fine. But I have one question
>> about the fork launcher. Is it possible to start a programm on different
>> nodes with this launcher or must i use the ssh launcher for this?
>> Because with the ssh launcher I got an error but I think this error
>> depends on the configuration of ssh on my node?!
>>
>> [mpiexec at da830] HYDU_create_process
>> (/mnt/ti_am1707_root/home/default/mpich2-1.3.2/src/pm/hydra/utils/launch/launch.c:69): 
>>
>> execvp error on file ssh (No such file or directory)
>>
>> best regards
>>
>> Harald
>>
>> On 14.11.2011 12:59, Harald Schuster wrote:
>>> Hi
>>>
>>> When I use the version mpich2-1.3.2 with the process manager hydra and
>>> I run he "mpirun /bin/hostname" command i got the following error:
>>>
>>> [mpiexec at da830] HYDU_create_process
>>> (/mnt/ti_am1707_root/home/default/mpich2-1.3.2/src/pm/hydra/utils/launch/launch.c:69): 
>>>
>>> execvp error on file ssh (No such file or directory)
>>>
>>> Do I need ssh when I use hydra and ch3:sock?!
>>>
>>> Harald
>>>
>>> On 11.11.2011 22:51, Pavan Balaji wrote:
>>>>
>>>> Does the following command work:
>>>>
>>>> % mpirun /bin/hostname
>>>>
>>>> If this gets a segmentation fault, the problem might be with mpirun.
>>>> In this case, you can try to debug it using:
>>>>
>>>> % gdb mpirun
>>>> (gdb) r /bin/hostname
>>>>
>>>> -- Pavan
>>>>
>>>> On 11/11/2011 03:31 AM, Harald Schuster wrote:
>>>>> Hi
>>>>>
>>>>> I now tried the cpi example from the mpich2 sources and I also got a
>>>>> segmentation fault. I used the following configuration:
>>>>>
>>>>> CFLAGS="-static" CC=arm-linux-gcc ./../configure
>>>>> --prefix=/mnt/ti_am1707_root/usr/mpich2-install --with-pm=hydra
>>>>> --disable-f77 --disable-fc --disable-cxx --enable-g=dbg,mem,log
>>>>> --with-device=ch3:sock --host=arm-linux
>>>>>
>>>>> How is it possible to trace the execution of the program?!
>>>>>
>>>>>
>>>>>
>>>>> On 10.11.2011 16:33, Harald Schuster wrote:
>>>>>> I added the config.log to this mail but i changed the pm from 
>>>>>> hydra to
>>>>>> gforker only for testing and know the helloWorld example runs but 
>>>>>> when
>>>>>> I tried a example with a broadcast the programm stops without any
>>>>>> error but it not finishes. The gforker and hydra uses sockets or?
>>>>>>
>>>>>> On 10.11.2011 16:12, Jeff Hammond wrote:
>>>>>>> I think it will help if you post config.log.
>>>>>>>
>>>>>>> Can you run the test that fails within GDB or Valgrind? I do not
>>>>>>> think that mpirun itself is segfaulting, but rather that it 
>>>>>>> activates
>>>>>>> the code path that does. Maybe this much was clear to you already.
>>>>>>>
>>>>>>> Jeff
>>>>>>>
>>>>>>> On Thu, Nov 10, 2011 at 3:33 AM, Harald Schuster
>>>>>>> <schuster at technikum-wien.at> wrote:
>>>>>>>> Hi
>>>>>>>>
>>>>>>>> I use MPICH2 on my normal Linux PC and it works fine. My next goal
>>>>>>>> is to run
>>>>>>>> MPICH2 on a arm926 (am1707) which is simulated in SimSoc and 
>>>>>>>> where a
>>>>>>>> normal
>>>>>>>> linux OS is running. The linux is compiled with the buildroot
>>>>>>>> 2010.05 and I
>>>>>>>> use the arm-linux-gcc 4.3.4. I tried to make the the mpich2 
>>>>>>>> with the
>>>>>>>> arm-linux-gcc and it works. I used the following configuration:
>>>>>>>>
>>>>>>>> CFLAGS="-static" CC=arm-linux-gcc ./../configure
>>>>>>>> --prefix=/mnt/ti_am1707_root/usr/mpich2-install --with-pm=hydra
>>>>>>>> --disable-f77 --disable-fc --disable-cxx --disable-sharedlibs
>>>>>>>> --host=arm-linux --with-device=ch3:sock
>>>>>>>>
>>>>>>>> Also the make install process works fine. Then I compiled a small
>>>>>>>> helloWorld
>>>>>>>> example with the mpicc for arm and it also works. When I run the
>>>>>>>> helloWorld
>>>>>>>> example on the simulated processor I always got an segmentation
>>>>>>>> fault. For
>>>>>>>> starting the program I use the following command:
>>>>>>>>
>>>>>>>> mpirun -np 1 helloWorld
>>>>>>>>
>>>>>>>> When I start the helloWorld program without the mpirun command 
>>>>>>>> I got
>>>>>>>> the
>>>>>>>> right output. So I think the segmentation fault is caused by the
>>>>>>>> mpirun
>>>>>>>> command but I don't know where and what I can do to solve the
>>>>>>>> problem.
>>>>>>>>
>>>>>>>> Best regards
>>>>>>>> Harald
>>>>>>>> _______________________________________________
>>>>>>>> mpich-discuss mailing list mpich-discuss at mcs.anl.gov
>>>>>>>> To manage subscription options or unsubscribe:
>>>>>>>> https://lists.mcs.anl.gov/mailman/listinfo/mpich-discuss
>>>>>>>>
>>>>>>>
>>>>>>>
>>>>>>
>>>>>>
>>>>>>
>>>>>> _______________________________________________
>>>>>> mpich-discuss mailing listmpich-discuss at mcs.anl.gov
>>>>>> To manage subscription options or unsubscribe:
>>>>>> https://lists.mcs.anl.gov/mailman/listinfo/mpich-discuss
>>>>>
>>>>>
>>>>>
>>>>>
>>>>> _______________________________________________
>>>>> mpich-discuss mailing list mpich-discuss at mcs.anl.gov
>>>>> To manage subscription options or unsubscribe:
>>>>> https://lists.mcs.anl.gov/mailman/listinfo/mpich-discuss
>>>>
>>>
>>> _______________________________________________
>>> mpich-discuss mailing list mpich-discuss at mcs.anl.gov
>>> To manage subscription options or unsubscribe:
>>> https://lists.mcs.anl.gov/mailman/listinfo/mpich-discuss
>>
>> _______________________________________________
>> mpich-discuss mailing list mpich-discuss at mcs.anl.gov
>> To manage subscription options or unsubscribe:
>> https://lists.mcs.anl.gov/mailman/listinfo/mpich-discuss
>

-- 
---------------------------------------
Harald Schuster, MSc
University of Applied Sciences
Technikum Wien
Institut für Embedded Systems
Höchstädtplatz 5
A-1200 Wien
T: +43 1 333 40 77-309
E: schuster at technikum-wien.at
H: http://embsys.technikum-wien.at
---------------------------------------