It's good to see you again!
On Thu, Mar 8, 2012 at 05:15, Ketan Maheshwari
<ketancma...@gmail.com> wrote:
> [steamroller:devide-re-v12.2.7-lin64]$ ./dre devide
> [steamroller:devide-re-v12.2.7-lin64]$
>
> As you see, I get the prompt back after the DeVIDE gui starts *momentarily*
> and quits.
Dang, it doesn't get very far.
Try some of the following, and please post output to this list:
./dre versions (this should spit out the versions of all included packages)
./dre ipython (should give you an ipython shell, then try:)
>>> import vtk
>>> import itk
>>> import wx
What would also be great, is if you could do an strace:
strace ./dre devide
See you,
Charl
--
You received this message because you are subscribed to the Google Groups "devide-users" group.
To post to this group, send email to devide...@googlegroups.com.
To unsubscribe from this group, send email to devide-users...@googlegroups.com.
For more options, visit this group at http://groups.google.com/group/devide-users?hl=en.
haha of course I do!
> [steamroller:devide-re-v12.2.7-lin64]$ ./dre ipython
> WARNING: Readline services not available on this platform.
> WARNING: The auto-indent feature requires the readline library
That's a bit strange. On a normal Linux machine, readline should be a
standard component. I'd look into this as well.
> Python software:
> Python 2.7.2 final-0
> numpy 1.6.1
> matplotlib 1.1.0
> wxPython 2.8.11.0
> VTK 5.8.0
> ITK 3.20.0
> gdcm 2.0.17
This means that the DRE can at least import all its own software.
> Strace output is here: http://www.mcs.anl.gov/~ketan/strace.out
The only thing that stood out from the strace, is that something went
wrong with a child process. dre spawns a python that spawns yet
another python. I'm guessing that some security measure on your system
does not like this. Do you have any of the security extensions
insntalled? You often find these on RedHat systems. If so, you could
try adding an exception for the DeVIDE system.
Another simple test you could try is:
./dre devide/devide.py
> Here is why we want to use DeVIDE:
>
> We do not want to use DeVIDE for visualization purposes. we are building a
> network data movement toolkit/kernel. We want to adopt DeVIDE's box and pipe
> model to use pipes to describe network characteristics: protocols:
> ftp/tcp/udp/scp etc., data and control. We want to use boxes to describe the
> executables: /bin/sleep, /bin/cat etc. and the location where to run those
> binaries in terms of ip addresses. Later, we want to increase the
> sophistication to branch-out branch-in etc.
That's a neat idea. At the moment, the DeVIDE scheduler analyses the
network and executes it in topological order. In your case, it sounds
like you would actually prefer it to be able to execute different
branches of the tree in parallel, as many modules will have to block
waiting for input. The scheduler does not do that yet.
With regard to the GIL: Python can switch between threads, but only
in-between opcodes. It can indeed never execute opcodes truly in
parallel. However, what is often the case with blocking or IO code,
the GIL can be given back to Python whilst the long-running blocking /
IO code does its thing. I've modified the VTK that ships with DeVIDE
in this way, so that you can in fact, from multiple Python threads
execute multiple truly parallel-running VTK threads.
Have fun!
Charl
That's a neat idea. At the moment, the DeVIDE scheduler analyses the
> Here is why we want to use DeVIDE:
>
> We do not want to use DeVIDE for visualization purposes. we are building a
> network data movement toolkit/kernel. We want to adopt DeVIDE's box and pipe
> model to use pipes to describe network characteristics: protocols:
> ftp/tcp/udp/scp etc., data and control. We want to use boxes to describe the
> executables: /bin/sleep, /bin/cat etc. and the location where to run those
> binaries in terms of ip addresses. Later, we want to increase the
> sophistication to branch-out branch-in etc.
network and executes it in topological order. In your case, it sounds
like you would actually prefer it to be able to execute different
branches of the tree in parallel, as many modules will have to block
waiting for input. The scheduler does not do that yet.
With regard to the GIL: Python can switch between threads, but only
in-between opcodes. It can indeed never execute opcodes truly in
parallel. However, what is often the case with blocking or IO code,
the GIL can be given back to Python whilst the long-running blocking /
IO code does its thing. I've modified the VTK that ships with DeVIDE
in this way, so that you can in fact, from multiple Python threads
execute multiple truly parallel-running VTK threads.
Have fun!
Charl
--
You received this message because you are subscribed to the Google Groups "devide-users" group.
To post to this group, send email to devide...@googlegroups.com.
To unsubscribe from this group, send email to devide-users...@googlegroups.com.
For more options, visit this group at http://groups.google.com/group/devide-users?hl=en.
You should definitely try it on a completely different machine /
installation. DeVIDE doesn't run on OSX natively, but you could try
parallels / virtualbox with a Linux installation.