(0036670)
Cory Quammen (developer)
2016-08-04 09:51

Haocheng,

Do you think your recent patch addressed this issue?

https://gitlab.kitware.com/vtk/vtk/merge_requests/1587 [^]

(0036674)
Haocheng Liu (developer)
2016-08-04 10:19

Hi Cory, FMHO I think my patch addressed this issue since I balanced the input points for different parallel parts. Testing this condition in paraview would be a good idea to validate my patch.

May I ask is the paraview source code on gitlab based on the newest vtk branch? If so, I can test it on my local paraview build :)

(0036676)
Cory Quammen (developer)
2016-08-04 10:57

Yes, I believe VTK has been updated to include your commit:

commit 56d7c50aaa9cb15d18cb7c2808503015bc6e1e7d
Author: Haocheng Liu <haocheng.liu@kitware.com>
Date: Thu Jun 23 16:51:28 2016 -0400

    Bug 0016089: Fix vtkmaskPoints multi-process missing condition

    In old verion, if proportionalMaximumNumberOfPoints is enabled with one
    process, localMaxPts would be set to MaximumNumberOfPoints which obsolete
    the upper line. After adding the process number checking condition, the if
    statement in GetLocalSampleSize function is no longer needed. And we also
    need to compare maximumNumberOfPoints with the total for garbage memory
    safety reason.

(0036677)
Haocheng Liu (developer)
2016-08-04 14:28

Hi Cory, I'm trying to find a pvtu test file for this patch, but somehow i can neither find one from paraview data nor internet... May I ask do you have any idea about where could i find such a test file?

(0036678)
Cory Quammen (developer)
2016-08-04 14:51

Haocheng, I attached an example pvtp file with associated vtp files that should work for testing.

To test, run ParaView, remote server with mpirun -np 4
In client, connect to server
Load attached data file
Add Glyph filter, set Glyph Mode to Uniform Spatial Distribution, Maximum Number of Sample Points to 5.
Color by vtkProcessId

You should see an even distribution of points across the different process ids.

It seems that the Maximum Number of Sample Points is applied to each processor, so you wind up with 4 * 5 = 20 glyphs. That's a separate bug, though.

(0036683)
Haocheng Liu (developer)
2016-08-04 15:20

Thanks for the guidance Cory! Currently I'm working on rewriting all the public and private module dependency in vtk and it's almost there:) If it's not urgent, hopefully I would test the file tomorrow.

(0036685)
Cory Quammen (developer)
2016-08-04 15:26

Nope, not urgent. I just noticed the possible relationship between this bug and your work while going through the bug tracker today.

(0036689)
Haocheng Liu (developer)
2016-08-08 17:30

Hi Cory, from my test result it seems that my patch didn't addressed this error. my patch fixed the problem that vtk would assign extreme large number of points when running in parallel, however it seems that the points passed to vtkMaskPoints are not uniform distributed. I suspect that there is a bug in random model of vtkMaskPoints rather than the parallel part.
A screenshot is also attached. There are more blue and green arrows than red ones.

(0036690)
Cory Quammen (developer)
2016-08-09 09:56

Thanks for checking, Haocheng!

(0038193)
Kitware Robot (administrator)
2016-08-12 09:59

Resolving issue as `moved`.

This issue tracker is no longer used. Further discussion of this issue may take place in the current ParaView Issues page linked in the banner at the top of this page.

Issue History
Date Modified	Username	Field	Change
2012-04-16 04:22	Tommy C	New Issue
2012-04-16 04:22	Tommy C	File Added: prv_glyph_screenshots.zip
2016-08-04 09:51	Cory Quammen	Note Added: 0036670
2016-08-04 09:52	Cory Quammen	Relationship added	related to 0016089
2016-08-04 10:19	Haocheng Liu	Note Added: 0036674
2016-08-04 10:57	Cory Quammen	Note Added: 0036676
2016-08-04 14:28	Haocheng Liu	Note Added: 0036677
2016-08-04 14:45	Cory Quammen	File Added: pvtp_example.zip
2016-08-04 14:51	Cory Quammen	Note Added: 0036678
2016-08-04 15:20	Haocheng Liu	Note Added: 0036683
2016-08-04 15:26	Cory Quammen	Note Added: 0036685
2016-08-08 17:30	Haocheng Liu	Note Added: 0036689
2016-08-08 17:32	Haocheng Liu	File Added: uneven result screenshot.png.zip
2016-08-09 09:56	Cory Quammen	Note Added: 0036690
2016-08-12 09:59	Kitware Robot	Note Added: 0038193
2016-08-12 09:59	Kitware Robot	Status	backlog => closed
2016-08-12 09:59	Kitware Robot	Resolution	open => moved
2016-08-12 09:59	Kitware Robot	Assigned To	=> Kitware Robot

Notes
(0036670) Cory Quammen (developer) 2016-08-04 09:51	Haocheng, Do you think your recent patch addressed this issue? https://gitlab.kitware.com/vtk/vtk/merge_requests/1587 [^]

(0036674) Haocheng Liu (developer) 2016-08-04 10:19	Hi Cory, FMHO I think my patch addressed this issue since I balanced the input points for different parallel parts. Testing this condition in paraview would be a good idea to validate my patch. May I ask is the paraview source code on gitlab based on the newest vtk branch? If so, I can test it on my local paraview build :)

(0036676) Cory Quammen (developer) 2016-08-04 10:57	Yes, I believe VTK has been updated to include your commit: commit 56d7c50aaa9cb15d18cb7c2808503015bc6e1e7d Author: Haocheng Liu <haocheng.liu@kitware.com> Date: Thu Jun 23 16:51:28 2016 -0400 Bug 0016089: Fix vtkmaskPoints multi-process missing condition In old verion, if proportionalMaximumNumberOfPoints is enabled with one process, localMaxPts would be set to MaximumNumberOfPoints which obsolete the upper line. After adding the process number checking condition, the if statement in GetLocalSampleSize function is no longer needed. And we also need to compare maximumNumberOfPoints with the total for garbage memory safety reason.

(0036677) Haocheng Liu (developer) 2016-08-04 14:28	Hi Cory, I'm trying to find a pvtu test file for this patch, but somehow i can neither find one from paraview data nor internet... May I ask do you have any idea about where could i find such a test file?

(0036678) Cory Quammen (developer) 2016-08-04 14:51	Haocheng, I attached an example pvtp file with associated vtp files that should work for testing. To test, run ParaView, remote server with mpirun -np 4 In client, connect to server Load attached data file Add Glyph filter, set Glyph Mode to Uniform Spatial Distribution, Maximum Number of Sample Points to 5. Color by vtkProcessId You should see an even distribution of points across the different process ids. It seems that the Maximum Number of Sample Points is applied to each processor, so you wind up with 4 * 5 = 20 glyphs. That's a separate bug, though.

(0036683) Haocheng Liu (developer) 2016-08-04 15:20	Thanks for the guidance Cory! Currently I'm working on rewriting all the public and private module dependency in vtk and it's almost there:) If it's not urgent, hopefully I would test the file tomorrow.

(0036685) Cory Quammen (developer) 2016-08-04 15:26	Nope, not urgent. I just noticed the possible relationship between this bug and your work while going through the bug tracker today.

(0036689) Haocheng Liu (developer) 2016-08-08 17:30	Hi Cory, from my test result it seems that my patch didn't addressed this error. my patch fixed the problem that vtk would assign extreme large number of points when running in parallel, however it seems that the points passed to vtkMaskPoints are not uniform distributed. I suspect that there is a bug in random model of vtkMaskPoints rather than the parallel part. A screenshot is also attached. There are more blue and green arrows than red ones.

(0036690) Cory Quammen (developer) 2016-08-09 09:56	Thanks for checking, Haocheng!

(0038193) Kitware Robot (administrator) 2016-08-12 09:59	Resolving issue as `moved`. This issue tracker is no longer used. Further discussion of this issue may take place in the current ParaView Issues page linked in the banner at the top of this page.