Difference between revisions of "Teleconference 2024-08-28"

From VTKM
Jump to navigation Jump to search
(Created page with "Attendees: == OASIS Deliverables == * Address requests for support, improvement, and new features by current and potential applications such as XGC and WarpX as well as oth...")
 
 
Line 1: Line 1:
Attendees:  
+
Attendees: Ken Moreland (ORNL), Hank Childs (UO), Cyrus Harrison (LLNL), Nicole Marsaglia (LLNL), Jeff Amstutz (Nvidia), Dave Pugmire (ORNL)
  
 
== OASIS Deliverables ==
 
== OASIS Deliverables ==
  
 
* Address requests for support, improvement, and new features by current and potential applications such as XGC and WarpX as well as other users of ParaView, VisIt, and Ascent.
 
* Address requests for support, improvement, and new features by current and potential applications such as XGC and WarpX as well as other users of ParaView, VisIt, and Ascent.
 +
** Ascent team solved ray tracing bug in VTK-m 2.2
 
* Strengthen operation of VTK-m on Aurora.
 
* Strengthen operation of VTK-m on Aurora.
 
** There are still some failing tests.
 
** There are still some failing tests.
Line 30: Line 31:
 
== Roundtable ==
 
== Roundtable ==
  
Dave has a branch that makes VTK-m a backend renderer for ANARI. Some slow progress with proof of concept.
+
Cyrus had a question about how VTK-m manages device streams and how that interfaces with Kokkos. Ken thinks VTK-m establishes a device stream per CPU core so that multiple CPU threads could schedule things simultaneously. Cyrus might need a way to set the device stream to use.
 
 
Potential progress Colleen Bertoni from ANL to put in Ascent configuration editor.
 
 
 
An issue was filed with the Intel compile team on the bug that was identified with VTK-m.
 

Latest revision as of 11:33, 28 August 2024

Attendees: Ken Moreland (ORNL), Hank Childs (UO), Cyrus Harrison (LLNL), Nicole Marsaglia (LLNL), Jeff Amstutz (Nvidia), Dave Pugmire (ORNL)

OASIS Deliverables

  • Address requests for support, improvement, and new features by current and potential applications such as XGC and WarpX as well as other users of ParaView, VisIt, and Ascent.
    • Ascent team solved ray tracing bug in VTK-m 2.2
  • Strengthen operation of VTK-m on Aurora.
    • There are still some failing tests.
    • They may be a problem with cell location, which might be an issue with Variant.
    • Having Intel look into the previously identified union/variant bug found.
  • Improve documentation with automatic generation and web deployment.
    • User's guide continues to improve.
  • Ensure proper building and running at OLCF, ALCF, and NERSC through Spack.
    • Ken verified that Slack builds work on Perlmutter and Frontier.
    • Silvio will establish contact for Ken of someone at ALCF tasked with compiling stuff with Spack. Request that they also compile VTK-m/Kokkos.

VTK-m Releases

Release for VTK-m 2.1.1 and 2.2.0 are out.

Viskores / HPSF

Have formalized the initial TSC. Ken is considering adding the following voting members to the TSC to represent their respective institutions and projects.

  • Silvio Rizzi, representing needs of ANL/ALCF
  • Nicole Marsaglia, representing Ascent and LLNL
  • Gunther Weber, representing topology, LBL, NERSC?

We have a plan to transition from VTK-m to Viskores. We are hoping for some funding from the HPSF to do this.

Roundtable

Cyrus had a question about how VTK-m manages device streams and how that interfaces with Kokkos. Ken thinks VTK-m establishes a device stream per CPU core so that multiple CPU threads could schedule things simultaneously. Cyrus might need a way to set the device stream to use.