Charm++ parallel programming system

[Abhishek Gupta <gupta59 AT illinois.edu>]

Hi Geoff,

I think you need to start charmd.exe , which is under net-win32/bin on all the nodes as listed in the .nodelist file (in your case just one).  Charmrun is trying to start node programs by contacting daemons (charmd not ssh in this case) on compute nodes.

Is you want use ssh, you can use net-cygwin build of charm instead of net-win32.

Thanks,
Abhishek

On Wed, Sep 5, 2012 at 7:23 AM, Geoff D. Hilyard <gdhilyard AT regaldecision.com> wrote:

Hi all.

 

I’m going thru the tutorials, trying to test our Charm++.

Basically.  I’m trying to run a program ( charmtestexe ) using the command “./charmrun ./charmtestexe +p1”.  P1 being a “Remote” computer, which is really the localhost.  The error message comes up with “Error ‘N’ starting remote node program on DCM8QHH1 – could not contact daemon—is it running?”

 

Backround first:  Not a Unix/Linux guy.  I have a Windows XP SP3 machine, and have downloaded and installed Cygwin.  I have started SSH(D), and have setup password-less logins with SSH.  I can “ssh localhost date” without a password, and it returns the date.  I have made a charmtestexe, and have compiled it.  It’s basically the array Hello World example, where a Chare calls the next Chare.

 

I downloaded Charm++ 6.2 binaries ( Can’t get the 6.4 from the website ) for net-win32.  I put them under my D Drive.  Cygwin is under my C Drive ( Which could be 1 issue ).  My CharmTestExe is also compiled on the D Drive. 

 

Firewall is turned off.  And finally staying off.

 

I put a AddCharm.sh in my profile.d ( cygwin\etc\profile.d ) directory, with the following:

PATH=/cygdrive/d/ModelAttempts/CharmPP/charm6.2/net-win32/bin:${PATH}

PATH=/cygdrive/d/ModelAttempts/CharmPP/CharmTest/CharmTest:${PATH}

export PATH

 

The first path is where I have downloaded the Charm++ program, and the binaries live.  The second is where my test program resides.  I was thinking that it couldn’t find the exe to execute remotely.  This did not help.  If I “echo $PATH”, I do have both paths listed, and separated by a ‘:’.  However, I do not need charmrun or charmtestexe in my home directory to run the commands with this in.

 

My Nodelist file is as follows (DCM8QHH1 is my local machine, with 2 CPUs, and I’m using ssh to connect ):

group main

host DCM8QHH1 ++cpus 2 ++shell ssh

 

I have also copied charmrun.exe and charmtestexe to my $Home directory ( cygwin\home\gdhilyard, with gdhilyard being my user name on the domain ).  I was thinking, again, it couldn’t find the program.  This didn’t change anything either.  ( I currently do NOT have them there )

 

If I add the ++verbose, I get the following ( How do you copy and paste from the cygwin terminal? ):

$Charmrun charmtestexe +p1 ++verbose

Charmrun> charmrun started…

Charmrun> using ./nodelist as nodesfile

Charmrun> adding client 0:  “DCM8QHH1”, IP:192.168.10.19

Charmrun> Starting node program 0 on “DCM8QHH1’ as c:\cygwin\home\gdhilyard\charmtestexe

Error ‘N’ starting remote node program on DCM8QHH1 –

could not contact daemon—is it running?

 

If I run do the ++local ($ charmrun charmtestexe +p2 ++verbose ++local), it works as expected.  With two processors used ( If I have +p1, only one is used. )

 

So.  Where have I gone wrong?  I’d like to get this up and running.  We are going to move our pedestrian / vehicle computer simulation to a parallel language, and so we are testing a few out.  Charm++ looks the most promising, but I’m getting hung up just getting the examples to run.

 

Thanks in advance

 

Geoff Hilyard

Lead Software Developer

Regal Decision Systems, Inc.

_______________________________________________
charm mailing list
charm AT cs.uiuc.edu
http://lists.cs.uiuc.edu/mailman/listinfo/charm

_______________________________________________
ppl mailing list
ppl AT cs.uiuc.edu
http://lists.cs.uiuc.edu/mailman/listinfo/ppl