MantisBT - ParaView
View Issue Details
IDProjectCategoryView StatusDate SubmittedLast Update
0013051ParaView(No Category)public2012-04-02 11:532016-08-12 09:59
ReporterChristopher Mitchell 
Assigned ToAashish Chaudhary 
PrioritynormalSeverityminorReproducibilityhave not tried
StatusclosedResolutionmoved 
PlatformOSOS Version
Product Version3.14 
Target VersionFixed in Version 
ProjectTBD
Topic Name
Typeperformance
Summary0013051: Slow Rendering Performance when using OpenGL (Hardware Rendering) and MPI
DescriptionSummery:
When using ParaView in parallel with either MVAPICH2 or OpenMPI and hardware rendering (nVidia Quadro 5000/6000, Tesla C2050 tested), rendering performance drops significantly. Problem reproduced using a VTK test program and when running on a single node with MPI simply linked to the executable (no composting occurring).

Notes:
1) Have ruled out composting as the issue by running on one node with MPI simply linked but not invoked (no mpirun or mpiexec called).
2) No X forwarding or Mesa in play on the systems exhibiting this issue.
3) Immediate mode rendering toggled on and off with no effect when test data set loaded into PV GUI.

To Reproduce:
1) Build ParaView w\ MPI support and OpenGL support for the installed graphics card.
2) Issue a pvbatch job with timer logs to capture run time. Recompile PV w/o MPI and re-time.
3) Convert the pvbatch job to a VTK C++ program. Link against vtkRendering and time execution. Relink program against vtkParallel and time execution.

Results Seen (VTK Program w\ MPI and 4 million triangles):
HP DL585 compute node w\ Quadro 5000, MVAPICH2: ~10 fps
HP Z800 workstation w\ Quadro 6000, OpenMPI: ~4 fps
HP Z800 workstation w\ Tesla C2050, OpenMPI: ~2.5 fps

Results w/o MPI Linked for DL585 Compute Node, 4 million triangles: 0000162:0000114 fps
Steps To Reproduce
Additional Information
TagsNo tags attached.
Relationships
Attached Files
Issue History
Date ModifiedUsernameFieldChange
2012-04-02 11:53Christopher MitchellNew Issue
2012-04-02 11:56Christopher MitchellNote Added: 0028420
2012-04-06 18:36Christopher MitchellNote Added: 0028433
2012-05-10 12:10Christopher MitchellNote Added: 0028586
2012-07-19 11:44Aashish ChaudharyAssigned To => Aashish Chaudhary
2012-07-19 11:46Aashish ChaudharyNote Added: 0028849
2012-07-19 11:46Aashish ChaudharyStatusbacklog => todo
2012-10-25 14:41Aashish ChaudharyNote Added: 0029516
2012-10-25 14:41Aashish ChaudharyStatustodo => active development
2016-08-12 09:59Kitware RobotNote Added: 0038177
2016-08-12 09:59Kitware RobotStatusactive development => closed
2016-08-12 09:59Kitware RobotResolutionopen => moved

Notes
(0028420)
Christopher Mitchell   
2012-04-02 11:56   
Tracking system incorrectly transformed the last line into a link. Original Text:

Results w/o MPI Linked for DL585 Compute Node, 4 million triangles: ~ 114 fps
(0028433)
Christopher Mitchell   
2012-04-06 18:36   
Re-ran tests with the following nVidia driver versions:

- 260.1936 (release date: 1/21/2011)
- 275.43 (release date: 12/14/2011)
- 295.20 (release date: 2/13/2012)
- 295.33 (release date: 3/22/2012)

Saw no change in rendering performance of MPI linked VTK test program between driver versions.
(0028586)
Christopher Mitchell   
2012-05-10 12:10   
Tested with MPICH2 version 1.4.1p1 (no IB support) and found that no performance hit was taken. Appears to narrow the problem to a conflict with OFED and/or the IB-enabled portions of MPI.
(0028849)
Aashish Chaudhary   
2012-07-19 11:46   
I was able to reproduce it and run various performance tests. Also I ran my tests on ATI cards and found no noticeable difference in performance. We are working towards a solution.
(0029516)
Aashish Chaudhary   
2012-10-25 14:41   
I submitted a report after extensive testing.
(0038177)
Kitware Robot   
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.