I spoke too soon...the mpi log shows that some processes are yet
being starved of data:
MPI: libxmpi.so 'SGI MPT 1.23 03/28/09 11:45:59'
MPI: libmpi.so 'SGI MPT 1.23 03/28/09 11:43:39'
MPI Environmental Settings
MPI: MPI_DSM_DISTRIBUTE (default: not set) : 1
MPI: The default size of the mapped stack area is 780 MBytes. The
current
stack limit (unlimited) is greater than this size. To specify
a new
size (in bytes) of the mapped stack, set the MPI_MAPPED_STACK_SIZE
environment variable.
Host: r2i2n2, CPUs: 16, TotPhysMem 3048401 pages, PhysMemPerCPU
190525 pages
Memmap mmap size: 4621264 pages (18928697344 bytes) mapped ranks 8
Memmap_init complete. shm base=0x2b129e733000, sym_static=1
RANK:0 sbrk base=0x 64f000, pagesize=0x1000 (4096)
RANK:0 static/heap base=0x 506000, top=0x 2ed43000,
len=190525 pages
RANK:0 stack base=0x 7fff1d7b2000, top=0x 7fff4bfef000,
len=190525 pages
RANK:0 mpibuffer base=0x 0, top=0x 0,
len= 0 pages
RANK:0 symheap base=0x 2b126e72b000, top=0x 2b129e72b000,
len=196608 pages
RANK:1 sbrk base=0x 64f000, pagesize=0x1000 (4096)
RANK:1 static/heap base=0x 506000, top=0x 2ed43000,
len=190525 pages
RANK:1 stack base=0x 7fff1d7b2000, top=0x 7fff4bfef000,
len=190525 pages
RANK:1 mpibuffer base=0x 0, top=0x 0,
len= 0 pages
RANK:1 symheap base=0x 2b126e72b000, top=0x 2b129e72b000,
len=196608 pages
RANK:2 sbrk base=0x 64f000, pagesize=0x1000 (4096)
RANK:2 static/heap base=0x 506000, top=0x 2ed43000,
len=190525 pages
RANK:2 stack base=0x 7fff1d7b2000, top=0x 7fff4bfef000,
len=190525 pages
RANK:2 mpibuffer base=0x 0, top=0x 0,
len= 0 pages
RANK:2 symheap base=0x 2b126e72b000, top=0x 2b129e72b000,
len=196608 pages
RANK:3 sbrk base=0x 64f000, pagesize=0x1000 (4096)
RANK:3 static/heap base=0x 506000, top=0x 2ed43000,
len=190525 pages
RANK:3 stack base=0x 7fff1d7b2000, top=0x 7fff4bfef000,
len=190525 pages
RANK:3 mpibuffer base=0x 0, top=0x 0,
len= 0 pages
RANK:3 symheap base=0x 2b126e72b000, top=0x 2b129e72b000,
len=196608 pages
RANK:4 sbrk base=0x 64f000, pagesize=0x1000 (4096)
RANK:4 static/heap base=0x 506000, top=0x 2ed43000,
len=190525 pages
RANK:4 stack base=0x 7fff1d7b2000, top=0x 7fff4bfef000,
len=190525 pages
RANK:4 mpibuffer base=0x 0, top=0x 0,
len= 0 pages
RANK:4 symheap base=0x 2b126e72b000, top=0x 2b129e72b000,
len=196608 pages
RANK:5 sbrk base=0x 64f000, pagesize=0x1000 (4096)
RANK:5 static/heap base=0x 506000, top=0x 2ed43000,
len=190525 pages
RANK:5 stack base=0x 7fff1d7b2000, top=0x 7fff4bfef000,
len=190525 pages
RANK:5 mpibuffer base=0x 0, top=0x 0,
len= 0 pages
RANK:5 symheap base=0x 2b126e72b000, top=0x 2b129e72b000,
len=196608 pages
RANK:6 sbrk base=0x 64f000, pagesize=0x1000 (4096)
RANK:6 static/heap base=0x 506000, top=0x 2ed43000,
len=190525 pages
RANK:6 stack base=0x 7fff1d7b2000, top=0x 7fff4bfef000,
len=190525 pages
RANK:6 mpibuffer base=0x 0, top=0x 0,
len= 0 pages
RANK:6 symheap base=0x 2b126e72b000, top=0x 2b129e72b000,
len=196608 pages
RANK:7 sbrk base=0x 64f000, pagesize=0x1000 (4096)
RANK:7 static/heap base=0x 506000, top=0x 2ed43000,
len=190525 pages
RANK:7 stack base=0x 7fff1d7b2000, top=0x 7fff4bfef000,
len=190525 pages
RANK:7 mpibuffer base=0x 0, top=0x 0,
len= 0 pages
RANK:7 symheap base=0x 2b126e72b000, top=0x 2b129e72b000,
len=196608 pages
Connected to client
Process id: 2 >> ERROR: In
/home/pratikm/source/ParaView/ParaView-3.10.1/VTK/Graphics/vtkTensorGlyph.cxx,
line 131
vtkTensorGlyph2 (0x4706a70): No data to glyph!
Process id: 3 >> ERROR: In
/home/pratikm/source/ParaView/ParaView-3.10.1/VTK/Graphics/vtkTensorGlyph.cxx,
line 131
vtkTensorGlyph2 (0x47029a0): No data to glyph!
pratik
On Thursday 28 April 2011 10:30 PM, pratik wrote:
But hey here's something..... i check use minimal memory and it
WORKED!
i wonder how that happened? anyway, i'm glad that it worked :)
Thanks Sven and Utkarsh for all the help !
pratik
On Thursday 28 April 2011 10:27 PM, pratik wrote:
Just tried it...but getting same result :( .
First loaded data, then applied D3 filter and then applied tensor
glyph..... am i doing something wrong here?
pratik
On Thursday 28 April 2011 10:20 PM, pratik wrote:
Thanks Sven! I'll try that out and post the results.
Btw. that would be a nice thing to add to the wiki, if it isn't
there already :)
pratik
On Thursday 28 April 2011 10:17 PM, Sven Buijssen wrote:
Hi,
Being the one who uploaded the plugin (having added a mere
frosting to the
servermanager XML that got posted here in 2009) I'd like to add
the following:
The plugin is merely a convenient way to use vtkTensorGlyph.cxx
from within
ParaView, but does not add additional features. As such it does
not provide any
magic to support multi-core/multiple pvserver processes.
When using the input file tensors8.vtk shipped with the plugin
in a multi-core
environment, you have to use the D3 filter first to distribute
the data. You
will notice that then the glyphs are already wrong for -np 2
Sven
pratik wrote, On 04/28/11 18:41:
I am quite sure now that the issue is do do with the some of
the processes
spawned by mpi not being able to do their jobs; i already
posted the verbose
output of the run previously, now i will also attach an image
that clearly shows
that parts of the glyphs are not being rendered. Can any one
the paraview
developers working with mpi please have a look at why this
seems to be happening?
I am now looking through the code of the plugin to see if that
is where the
problem is......If i do find out I'll definitely post it here.
any help on this would be really helpful.
pratik
On Thursday 28 April 2011 06:37 PM, pratik wrote:
Great Idea. Also, thanks very much for helping me Utkarsh.
analysis:
1) works on client perfectly
2) started pvserver on head node with -np 1 (for mpi):- no
problem
3) works on head node UPTO -np 3 .... -np 4 becomes fractured.
Now, the other filters that i have used have worked
perfectly....this may be
something to do with the plugin, as you suggested.
pratik
On Thursday 28 April 2011 06:19 PM, Utkarsh Ayachit wrote:
Ok first things first. This is community contributed plugin
so I have
no idea what it does or what it doesn't. To identify the
issue, I
always tend to start simple. Run without connecting to the
server at
all. In builtin mode, does the plugin work? Next try with 1
pvserver
processes without MPI. Next 2 pvserver processes and so on.
Utkarsh
On Thu, Apr 28, 2011 at 8:46 AM,
pratik<pratik.mal...@gmail.com> wrote:
iran mpirun with the verbose option on the server and got
this as output;
MPI: libxmpi.so 'SGI MPT 1.23 03/28/09 11:45:59'
MPI: libmpi.so 'SGI MPT 1.23 03/28/09 11:43:39'
MPI Environmental Settings
MPI: MPI_DSM_DISTRIBUTE (default: not set) : 1
MPI: The default size of the mapped stack area is 780
MBytes. The current
stack limit (unlimited) is greater than this size. To
specify a new
size (in bytes) of the mapped stack, set the
MPI_MAPPED_STACK_SIZE
environment variable.
Host: r2i1n0, CPUs: 16, TotPhysMem 3048401 pages,
PhysMemPerCPU 190525 pages
Memmap mmap size: 4621264 pages (18928697344 bytes) mapped
ranks 8
Memmap_init complete. shm base=0x2b70b120b000, sym_static=1
RANK:0 sbrk base=0x 64f000, pagesize=0x1000
(4096)
RANK:0 static/heap base=0x 506000, top=0x
2ed43000,
len=190525 pages
RANK:0 stack base=0x 7fff0acd9000, top=0x
7fff39516000,
len=190525 pages
RANK:0 mpibuffer base=0x 0,
top=0x 0, len=
0 pages
RANK:0 symheap base=0x 2b7081203000, top=0x
2b70b1203000,
len=196608 pages
RANK:1 sbrk base=0x 64f000, pagesize=0x1000
(4096)
RANK:1 static/heap base=0x 506000, top=0x
2ed43000,
len=190525 pages
RANK:1 stack base=0x 7fff0acd9000, top=0x
7fff39516000,
len=190525 pages
RANK:1 mpibuffer base=0x 0,
top=0x 0, len=
0 pages
RANK:1 symheap base=0x 2b7081203000, top=0x
2b70b1203000,
len=196608 pages
RANK:2 sbrk base=0x 64f000, pagesize=0x1000
(4096)
RANK:2 static/heap base=0x 506000, top=0x
2ed43000,
len=190525 pages
RANK:2 stack base=0x 7fff0acd9000, top=0x
7fff39516000,
len=190525 pages
RANK:2 mpibuffer base=0x 0,
top=0x 0, len=
0 pages
RANK:2 symheap base=0x 2b7081203000, top=0x
2b70b1203000,
len=196608 pages
RANK:3 sbrk base=0x 64f000, pagesize=0x1000
(4096)
RANK:3 static/heap base=0x 506000, top=0x
2ed43000,
len=190525 pages
RANK:3 stack base=0x 7fff0acd9000, top=0x
7fff39516000,
len=190525 pages
RANK:3 mpibuffer base=0x 0,
top=0x 0, len=
0 pages
RANK:3 symheap base=0x 2b7081203000, top=0x
2b70b1203000,
len=196608 pages
RANK:4 sbrk base=0x 64f000, pagesize=0x1000
(4096)
RANK:4 static/heap base=0x 506000, top=0x
2ed43000,
len=190525 pages
RANK:4 stack base=0x 7fff0acd9000, top=0x
7fff39516000,
len=190525 pages
RANK:4 mpibuffer base=0x 0,
top=0x 0, len=
0 pages
RANK:4 symheap base=0x 2b7081203000, top=0x
2b70b1203000,
len=196608 pages
RANK:5 sbrk base=0x 64f000, pagesize=0x1000
(4096)
RANK:5 static/heap base=0x 506000, top=0x
2ed43000,
len=190525 pages
RANK:5 stack base=0x 7fff0acd9000, top=0x
7fff39516000,
len=190525 pages
RANK:5 mpibuffer base=0x 0,
top=0x 0, len=
0 pages
RANK:5 symheap base=0x 2b7081203000, top=0x
2b70b1203000,
len=196608 pages
RANK:6 sbrk base=0x 64f000, pagesize=0x1000
(4096)
RANK:6 static/heap base=0x 506000, top=0x
2ed43000,
len=190525 pages
RANK:6 stack base=0x 7fff0acd9000, top=0x
7fff39516000,
len=190525 pages
RANK:6 mpibuffer base=0x 0,
top=0x 0, len=
0 pages
RANK:6 symheap base=0x 2b7081203000, top=0x
2b70b1203000,
len=196608 pages
RANK:7 sbrk base=0x 64f000, pagesize=0x1000
(4096)
RANK:7 static/heap base=0x 506000, top=0x
2ed43000,
len=190525 pages
RANK:7 stack base=0x 7fff0acd9000, top=0x
7fff39516000,
len=190525 pages
RANK:7 mpibuffer base=0x 0,
top=0x 0, len=
0 pages
RANK:7 symheap base=0x 2b7081203000, top=0x
2b70b1203000,
len=196608 pages
Connected to client
Process id: 3>> ERROR: In
/home/pratikm/source/ParaView/ParaView-3.10.1/VTK/Graphics/vtkTensorGlyph.cxx,
line 131
vtkTensorGlyph2 (0x4656cd0): No data to glyph!
Process id: 7>> ERROR: In
/home/pratikm/source/ParaView/ParaView-3.10.1/VTK/Graphics/vtkTensorGlyph.cxx,
line 131
vtkTensorGlyph2 (0x4651510): No data to glyph!
Client connection closed.
How can there be no data to glyph? Is it some communication
problem which is
causing this? I think that the two processes mentioned here
may not be doing
their part of the processing.... what do you think?
pratik
On Thursday 28 April 2011 06:09 PM, pratik wrote:
Yes...it is still broken :( ;(
have you ever had the situation when paraview at both the
ends were
compiled with different compilers? Is it an X- related
problem?
pratik
On Thursday 28 April 2011 06:01 PM, Utkarsh Ayachit wrote:
Not sure, I was able to play with this plugin just fine.
Try this:
Go to "Edit | Settings" dialog. Open up the "Render View"
pages and go
to the "Server" Page. There, uncheck the remote-render
threshold. Are
things still broken?
Utkarsh
On Thu, Apr 28, 2011 at 8:21 AM,
pratik<pratik.mal...@gmail.com> wrote:
sure...here it is.
This is a test file from the plugin folder called
"tensors8.vtk"
The glyphs are fractured...that is the problem.
I tried it only on client and it works fine.
pratik
On Thursday 28 April 2011 05:41 PM, Utkarsh Ayachit wrote:
Can you post a image?
Utkarsh
On Thu, Apr 28, 2011 at 8:08 AM,
pratik<pratik.mal...@gmail.com>
wrote:
I did just as you told and the two pluigns loaded properly
however...the plugin is yet not working correctly; the
tensors are
appearing
fractured...What may be the source of error now?
pratik
On Wednesday 27 April 2011 05:54 PM, Utkarsh Ayachit
wrote:
Yes and then load the right plugin on the right
process from the
"Manage PLugins" dialog. From the error, it looks
like you are
trying
to load the icc built plugin on the gcc built client.
Utkarsh
On Wed, Apr 27, 2011 at 6:59 AM,
pratik<pratik.mal...@gmail.com>
wrote:
Hi Utkarsh,
I compiled the plugin on BOTH server and my laptop,
using the
*respective*
configurations(i.e server plugin using pv server
settings(icc
compiler),
client plugin using client pv(gcc compiler)
settings). Like you had
asked
earlier, i did compile pv on the cluster through
source (which made
the
paraviewconfig..cmake file available ) and used it to
compile the
plugin,
on
the server.
Is this what you are asking for?
pratik
On Wednesday 27 April 2011 04:18 PM, Utkarsh Ayachit
wrote:
You need plugins compiled with both compilers and
load the icc
plugin
on
server while load the gcc plugin on client. You
cannot interchange
them
or
load the same one on both sides.
Utkarsh
Sent from my iPad
On Apr 27, 2011, at 12:03 AM,
pratik<pratik.mal...@gmail.com>
wrote:
I compiled PV on my laptop using gcc but on cluster
using icc.
When i
try
to use the TensorglyphFilter plugin, the following
error shows:
Load Error, Mismatch in version:
Paraview Signature: paraviewplugin|GNU|3.10
Plugin signature: paraviewplugin|Intel|3.10
I also tried compiling both plugins using gcc/g++
compiler, but
the
same
error persists. What may be the problem?
pratik
_______________________________________________
Powered by www.kitware.com
Visit other Kitware open-source projects at
http://www.kitware.com/opensource/opensource.html
Please keep messages on-topic and check the
ParaView Wiki at:
http://paraview.org/Wiki/ParaView
Follow this link to subscribe/unsubscribe:
http://www.paraview.org/mailman/listinfo/paraview
_______________________________________________
Powered by www.kitware.com
Visit other Kitware open-source projects at
http://www.kitware.com/opensource/opensource.html
Please keep messages on-topic and check the ParaView Wiki at:
http://paraview.org/Wiki/ParaView
Follow this link to subscribe/unsubscribe:
http://www.paraview.org/mailman/listinfo/paraview