Re: [Paraview] how to compile it again
Hi Gang, I noticed a previous post of yours on the mailing list where you said you wanted to include headers like QPoint and QTime. Are these the headers you are trying to add to pvserver.cxx? Normally, Qt headers like this are not added to files in ParaView3/Servers/Executables. The pvserver executable is not linked with Qt libraries, only the client GUI executable is (files in ParaView3/Qt and ParaView3/Applications/Client) Pat 2008/9/10 Moreland, Kenneth [EMAIL PROTECTED] The header files you are trying to include are apparently not in the include path. You can add paths for the compile to search for include files using the INCLUDE_DIRECTORIES CMake command in the ParaView3/Servers/Executables/CMakeLists.txt file. -Ken On 9/9/08 7:07 PM, 吴亚刚 [EMAIL PROTECTED] wrote: hi , everyone, I was a little confused about some re-compile problems on Paraview. I have added some include files into the original file pvserver.cxx for some custom usage. (Using VS.Net2005) When compile it , it said that the include file could not be found. And If I use the absolute directory, then the lib file would not be found. Of course that was not a good solutio. How to simplely compile it ? thanks. ___ ParaView mailing list ParaView@paraview.org http://www.paraview.org/mailman/listinfo/paraview ___ ParaView mailing list ParaView@paraview.org http://www.paraview.org/mailman/listinfo/paraview
Re: [Paraview] hdf5 in paraview segfaults on sparc
On Wed, Sep 10, 2008 at 7:42 PM, Berk Geveci [EMAIL PROTECTED] wrote: Which version of hdf5 is Debian using? Maybe we should upgrade the one in ParaView... 1.6.6-4 I don't know which version Paraview uses, but the problem is that even the newest hdf5 from upstream one has the bug, see the bug report in my original email for details. Ondrej ___ ParaView mailing list ParaView@paraview.org http://www.paraview.org/mailman/listinfo/paraview
[Paraview] vtkScalars used in Paraview?
Hi! Isn´t the vtk class vtkScalars used in Paraview? I didn´t find the .h or .cxx file for it. How can I add it, i.e. where do I place the source and which cMakeLists files do I have to alter to get it compiled with the rest? thx, Natalie _ Express yourself instantly with MSN Messenger! Download today it's FREE! http://messenger.msn.click-url.com/go/onm00200471ave/direct/01/___ ParaView mailing list ParaView@paraview.org http://www.paraview.org/mailman/listinfo/paraview
[Paraview] filterneu
_
Express yourself instantly with MSN Messenger! Download today it's FREE!
http://messenger.msn.click-url.com/go/onm00200471ave/direct/01/#include vtkHorizontalAverage.h
#include vtkDataSet.h
#include vtkStructuredGrid.h
#include vtkStructuredPoints.h
#include vtkRectilinearGrid.h
#include vtkDataArray.h
#include vtkObjectFactory.h
#include vtkPointData.h
#include vtkFloatArray.h
vtkStandardNewMacro(vtkHorizontalAverage);
vtkHorizontalAverage::vtkHorizontalAverage()
{
this - Normalize = 0;
}
void vtkHorizontalAverage::PrintSelf(ostream os, vtkIndent indent)
{
this-Superclass::PrintSelf(os,indent);
os indent Normalize:
(this-Normalize ? this-Normalize : 0) endl;
}
void vtkHorizontalAverage::Execute()
{vtkDataSet *input = this - GetInput();
vtkStructuredGrid *output = this -GetOutput();
int DataSetType;
DataSetType = input - GetDataObjectType();
int dims[3],i,j,k;
// pointer for casting
vtkRectilinearGrid *rect_input = vtkRectilinearGrid::New();
vtkStructuredGrid *structgrid_input = vtkStructuredGrid::New();
vtkPointData *iPtr,*oPtr;
double *inPtr, *outPtr;
vtkDataArray *scaldat;
vtkDataArray *new_scaldat;
vtkFloatArray *scalars = vtkFloatArray::New();
vtkFloatArray *new_scalars = vtkFloatArray::New();
int numPts = input - GetNumberOfPoints();
output - Allocate(numPts);
if((DataSetType == VTK_RECTILINEAR_GRID) (rect_input = vtkRectilinearGrid::SafeDownCast(input)))
{
rect_input - GetDimensions(dims);
//get the scalar attribute Data
iPtr = rect_input-GetPointData();
scaldat = iPtr - GetScalars();
oPtr = output-GetPointData();
new_scaldat = oPtr - GetScalars();
scalars = vtkFloatArray::SafeDownCast(scaldat);
new_scalars = vtkFloatArray::SafeDownCast(new_scaldat);
if(!scaldat !new_scaldat) vtkErrorMacro(Downcast to vtkFloatArray failed);
// inPtr = rect_input-GetPoint(0);
// outPtr = output-GetPoint(0);
}
else if((DataSetType == VTK_STRUCTURED_GRID ) (structgrid_input = vtkStructuredGrid::SafeDownCast(input)))
{vtkPointData *iPtr = structgrid_input-GetPointData();
vtkPointData *oPtr = output-GetPointData();
iPtr = structgrid_input-GetPointData();
scaldat = iPtr - GetScalars();
oPtr = output-GetPointData();
new_scaldat = oPtr - GetScalars();
scalars = vtkFloatArray::SafeDownCast(scaldat);
new_scalars = vtkFloatArray::SafeDownCast(new_scaldat);
if(!scaldat !new_scaldat) vtkErrorMacro(Downcast to vtkFloatArray failed);
structgrid_input-GetDimensions(dims);
}
else vtkErrorMacro(unsupported Dataset Format);
double *horizontal = new double[dims[0]];
double *horizontal_sigma = new double[dims[0]];
int index;
// compute the horizontal average
for (i=0; idims[0]; i++) {
horizontal[i] = 0.0;
for (j=0; jdims[1]; j++) {
for (k=0; kdims[2]; k++) {
index = k*(dims[1]*dims[0]) + j*dims[0] + i;
horizontal[i] += scalars - GetComponent(index, 1); //horizontal[i] += inPtr[index];
}
}
horizontal[i] = horizontal[i]/(dims[1]*dims[2]);
}
//subtract horizontal average
for (i=0; idims[0]; i++) {
for (j=0; jdims[1]; j++) {
for (k=0; kdims[2]; k++) {
index = k*(dims[1]*dims[0]) + j*dims[0] + i;
new_scalars - InsertNextTuple ((scalars - GetScalar(index)) - horizontal[i]);
}
}
}
if (this-GetNormalize()) {
// compute sigma
for (i=0; idims[0]; i++) {
horizontal_sigma[i] = 0.0;
for (j=0; jdims[1]; j++) {
for (k=0; kdims[2]; k++) {
index = k*(dims[1]*dims[0]) + j*dims[0] + i;
horizontal_sigma[i] += (scalars - GetComponent(index, 1))*(GetComponent(index, 1));
}
}
horizontal_sigma[i] /= (dims[1]*dims[2]-1);
horizontal_sigma[i] = sqrt(horizontal_sigma[i]);
}
// now normalize
for (i=0; idims[0]; i++) {
for (j=0; jdims[1]; j++) {
for (k=0; kdims[2]; k++) {
index = k*(dims[1]*dims[0]) + j*dims[0] + i;
//outPtr[index] /= horizontal_sigma[i];
new_scalars - InsertNextTuple((new_scalars - GetScalar(index)) /= horizontal_sigma[i])
}
}
}
}
delete [] horizontal;
delete [] horizontal_sigma;
//now set output
output - CopyStructure(input);
output - SetScalars (new_scalars);
iPtr-Delete();
oPtr-Delete();
delete [] inPtr;
delete [] outPtr;
}
#ifndef __vtkHorizontalAverage_h
#define __vtkHorizontalAverage_h
#include vtkDataSetToStructuredGridFilter.h
class VTK_EXPORT vtkHorizontalAverage : public vtkDataSetToStructuredGridFilter
{ public:
static vtkHorizontalAverage *New();
// vtkStandardNewMacro(vtkHorizontalAverage);
vtkTypeMacro(vtkHorizontalAverage,vtkDataSetToStructuredGridFilter);
vtkSetMacro(Normalize,int);
vtkGetMacro(Normalize,int);
vtkBooleanMacro(Normalize, int);
void PrintSelf(ostream, vtkIndent);
protected: vtkHorizontalAverage();
~vtkHorizontalAverage(){};
int Normalize;
void operator=(const vtkHorizontalAverage) {};
void Execute();
};
Re: [Paraview] vtkScalars used in Paraview?
vtkScalars? Even VTK doesn't have such a class. Are you sure you have the correct name for the class you are looking for? Utkarsh Natalie Happenhofer wrote: Hi! Isn´t the vtk class vtkScalars used in Paraview? I didn´t find the .h or .cxx file for it. How can I add it, i.e. where do I place the source and which cMakeLists files do I have to alter to get it compiled with the rest? thx, Natalie Express yourself instantly with MSN Messenger! MSN Messenger http://clk.atdmt.com/AVE/go/onm00200471ave/direct/01/ ___ ParaView mailing list ParaView@paraview.org http://www.paraview.org/mailman/listinfo/paraview ___ ParaView mailing list ParaView@paraview.org http://www.paraview.org/mailman/listinfo/paraview
Re: [Paraview] vtkScalars used in Paraview?
You probably have a code that depends on old VTK. Look at: http://www.vtk.org/Wiki/VTK_FAQ#Changes_in_VTK_between_3.2_and_4.0 -berk On Wed, Sep 10, 2008 at 4:05 PM, Natalie Happenhofer [EMAIL PROTECTED] wrote: Hi! Isn´t the vtk class vtkScalars used in Paraview? I didn´t find the .h or .cxx file for it. How can I add it, i.e. where do I place the source and which cMakeLists files do I have to alter to get it compiled with the rest? thx, Natalie Express yourself instantly with MSN Messenger! MSN Messenger ___ ParaView mailing list ParaView@paraview.org http://www.paraview.org/mailman/listinfo/paraview ___ ParaView mailing list ParaView@paraview.org http://www.paraview.org/mailman/listinfo/paraview
Re: [Paraview] slow rendering speed
Hi Ken. Thanks for reply again.I have tried this command, but the error is still the same: DISPLAY is not accessible on the server side. In the wiki, there are three things to do with X Connections: 1. Have xdm run on each cluster node at startup. Although xdm is almost always run at startup on workstation installations, it is not as commonplace to be run on cluster nodes. Talk to your system administrators for help in setting this up. 2. Disable all security on the X server. That is, allow any process to open a window on the x server without having to log in. Again, talk to your system administrators for help. 3. Make sure each process is run with the DISPLAY environment variable set to localhost:0 (or just :0). I have talked to system administrator about the first two things. The guy said that both are ok on the cluster, but he isn't 100 percent sure about the second thing. I try to run the glxgeasr program but it doesn't work. The error I got is couldn't open display '(null)' Could the problem result from the first two configurations? What do you think? Thanks, Aaron On Wed, Sep 10, 2008 at 1:37 PM, Moreland, Kenneth [EMAIL PROTECTED]wrote: It sounds like you are using the env program wrong. The env program runs a command, specified on its command line, with a modified environment, but does not modify the environment of the calling program. To use env, you need to use it within your mpirun command: mpirun -np 4 /usr/bin/env DISPLAY=localhost:0 ./pvserver -Ken On 9/10/08 11:32 AM, Biao She [EMAIL PROTECTED] wrote: Thanks your help again, Ken. You figure out my problem. I indeed connect to localhost:10.0. I am a newbie of Linux and ssh, I don't know how to connect to localhost:0.0 instead of 10.0. After I ssh to the cluster, the DISPLAY automatically set to localhost:10.0. I use /usr/bin/env DISPLAY=localhost:0 command to set it back. Then, when I try to connect to the paraview server, I get an error like this DISPLAY is not accessible on the server side. I know that if I start pvserver from the cluster directly, the DISPLAY is localhost:0, but the cluster doesn't have any keyboard and monitor. How can I connect to localhost:0 remotely? Thanks! Aaron On Wed, Sep 10, 2008 at 7:09 AM, Moreland, Kenneth [EMAIL PROTECTED] wrote: First, let me just say that you should not expect the parallel rendering of such a small data set to be faster than just rendering it locally. The parallel rendering has a constant overhead involving the readback of pixels, the transfer of data, and the computation of compositing operations. That said, based on the error it looks like your DISPLAY environment is set wrong. You should be connecting to localhost:0.0, not localhost:10.0. You are probably forwarding all of your rendering commands back to the desktop you logged in from though ssh forwarding. That, of course, will really slow things down because all the rendering commands are being shoved through the network and all are being processed on a single machine. Make sure that you are actually rendering to the local GPU. See the ParaView Wiki for more advice. http://www.paraview.org/Wiki/Setting_up_a_ParaView_Server -Ken On 9/9/08 4:24 PM, Biao She [EMAIL PROTECTED] wrote: Hi there. I use the client/server mode of paraview to render a small data(about 4M), but the speed is very slow. Below is how I do it: I first ssh to the cluster and start the pvserver by this command: mpirun -np 4 ./pvserver. The cluster is SGI Altix 4000. The operating system on the cluster is SUSe linux, 64 bits system. Then, I launch the paraview client on my local windows computer. After the client connected to the server, there are two types of errors: 1. Xlib: extension XFree86-DRI missing on display localhost:10.0 2. unaligned access to 0x6197a1bc, ip=0x23568e0 Paraview works regardless of these errors, but the problem is that the rendering speed is way too slow. I am wondering what could cause the slow down? Maybe the paraview doesn't use GPU on the clusters at all? How could I tell if the paraveiw use GPU or not? Or the unaligned access slow down the speed? These errors keep on repeating whenever I change the viewplane. I have stalled by this problem for days, any idea would be highly appreciated. Thanks! Aaron ___ ParaView mailing list ParaView@paraview.org http://www.paraview.org/mailman/listinfo/paraview
Re: [Paraview] slow rendering speed
Hi Aaron,
Here are some things to try. Ssh to the cluster and run glxgears. You
should not see a glxgears window appear on your screen (if you do, it is
using X forwarding). The command should appear to hang, and you will kill
it with control-c. If glxgears fails, try different values of DISPLAY until
it works, for example:
export DISPLAY=:0.0
some shells set environment variables like this-
setenv DISPLAY :0.0
Once glxgears works on the main node, try it with mpirun. Some versions of
mpirun have a command line option -x, this will export a local environment
variable to all the mpi processes:
mpirun -np 4 -x DISPLAY glxgears
or
mpirun -np 4 -x DISPLAY=:0.0 glxgears
If glxgears doesn't work for you, here is a simple debugging program I have
used in the past-
/* main.c */
#include mpi.h
#include stdio.h
#include stdlib.h
int main (int argc, char *argv[])
{
int rank, size;
MPI_Init (argc, argv); /* starts MPI */
MPI_Comm_rank (MPI_COMM_WORLD, rank);/* get current process id */
MPI_Comm_size (MPI_COMM_WORLD, size);/* get number of processes
*/
printf( I am process %d of %d. DISPLAY=%s\n, rank, size,
getenv(DISPLAY) );
MPI_Finalize();
return 0;
}
Compile it and run it:
$ mpicc -o main main.c
$ mpirun -np 4 ./main
I am process 0 of 4. DISPLAY=:0.0
I am process 1 of 4. DISPLAY=:0.0
I am process 2 of 4. DISPLAY=:0.0
I am process 3 of 4. DISPLAY=:0.0
Pat
On Wed, Sep 10, 2008 at 5:02 PM, Biao She [EMAIL PROTECTED] wrote:
Hi Ken. Thanks for reply again.I have tried this command, but the error is
still the same: DISPLAY is not accessible on the server side.
In the wiki, there are three things to do with X Connections:
1. Have xdm run on each cluster node at startup. Although xdm is almost
always run at startup on workstation installations, it is not as
commonplace
to be run on cluster nodes. Talk to your system administrators for help in
setting this up.
2. Disable all security on the X server. That is, allow any process to
open a window on the x server without having to log in. Again, talk to your
system administrators for help.
3. Make sure each process is run with the DISPLAY environment variable
set to localhost:0 (or just :0).
I have talked to system administrator about the first two things. The guy
said that both are ok on the cluster, but he isn't 100 percent sure about
the second thing. I try to run the glxgeasr program but it doesn't work. The
error I got is couldn't open display '(null)'
Could the problem result from the first two configurations? What do you
think?
Thanks,
Aaron
On Wed, Sep 10, 2008 at 1:37 PM, Moreland, Kenneth [EMAIL PROTECTED]wrote:
It sounds like you are using the env program wrong. The env program runs
a
command, specified on its command line, with a modified environment, but
does not modify the environment of the calling program. To use env, you
need to use it within your mpirun command:
mpirun -np 4 /usr/bin/env DISPLAY=localhost:0 ./pvserver
-Ken
On 9/10/08 11:32 AM, Biao She [EMAIL PROTECTED] wrote:
Thanks your help again, Ken.
You figure out my problem. I indeed connect to localhost:10.0. I am a
newbie
of Linux and ssh, I don't know how to connect to localhost:0.0 instead
of
10.0. After I ssh to the cluster, the DISPLAY automatically set to
localhost:10.0. I use /usr/bin/env DISPLAY=localhost:0 command to set
it
back. Then, when I try to connect to the paraview server, I get an error
like
this DISPLAY is not accessible on the server side.
I know that if I start pvserver from the cluster directly, the DISPLAY
is
localhost:0, but the cluster doesn't have any keyboard and monitor.
How can I connect to localhost:0 remotely?
Thanks!
Aaron
On Wed, Sep 10, 2008 at 7:09 AM, Moreland, Kenneth [EMAIL PROTECTED]
wrote:
First, let me just say that you should not expect the parallel
rendering of
such a small data set to be faster than just rendering it locally. The
parallel rendering has a constant overhead involving the readback of
pixels,
the transfer of data, and the computation of compositing operations.
That said, based on the error it looks like your DISPLAY environment is
set
wrong. You should be connecting to localhost:0.0, not localhost:10.0.
You
are probably forwarding all of your rendering commands back to the
desktop
you logged in from though ssh forwarding. That, of course, will really
slow
things down because all the rendering commands are being shoved through
the
network and all are being processed on a single machine.
Make sure that you are actually rendering to the local GPU. See the
ParaView Wiki for more advice.
http://www.paraview.org/Wiki/Setting_up_a_ParaView_Server
-Ken
On 9/9/08 4:24 PM, Biao She [EMAIL PROTECTED] wrote:
Hi there.
I use the client/server mode of paraview to render a small data(about
4M),
but
the speed is very slow.
