Difference between revisions of "Teleconference 2024-02-07"

From VTKM
Jump to navigation Jump to search
(Created page with "Attendees: Ken Moreland (ORNL), Dave Pugmire (ORNL), Hank Childs (UO), Jay Wang (ORNL), Sujin Philip (Kitware), Abhishek Yenpure (Kitware), Nicole Marsaglia (LLNL) == ECP Upd...")
 
 
(One intermediate revision by the same user not shown)
Line 1: Line 1:
Attendees: Ken Moreland (ORNL), Dave Pugmire (ORNL), Hank Childs (UO), Jay Wang (ORNL), Sujin Philip (Kitware), Abhishek Yenpure (Kitware), Nicole Marsaglia (LLNL)
+
Attendees: Ken Moreland (ORNL), Silvio Rizzi (ANL), Mark Bolstad (SNL), Abhishek Yenpure (Kitware), Dave Pugmire (ORNL), Tushar Athawale (ORNL)
  
 
== ECP Updates ==
 
== ECP Updates ==
Line 17: Line 17:
 
No meeting next two weeks.
 
No meeting next two weeks.
  
An Intel engineer found that the build issues we are having with building for ascent types are fixed in the latest compiler. That compiler is not available on Aurora yet, but will be soon.
+
Silvio reports that compiles on Aurora work better if you set the SIMD width to 16 (from 32). Ken says that will probably be OK for VTK-m in general.
 +
 
 +
The Kokkos device adapter test is failing. This might be similar to earlier issues that were fixed. Mark will take a look at the detailed log.

Latest revision as of 12:27, 7 February 2024

Attendees: Ken Moreland (ORNL), Silvio Rizzi (ANL), Mark Bolstad (SNL), Abhishek Yenpure (Kitware), Dave Pugmire (ORNL), Tushar Athawale (ORNL)

ECP Updates

  • Aurora available with no allocations or restrictions
    • Access through April 2024
    • Email support@alcf.anl.gov to get an account on Aurora (if you already have one on Sunspot)
  • Confluence and JIRA are gone
  • Zoom going away May 31
    • Planning to move to Google Meet
    • Check to make sure you will be able to connect with that
  • IJHPCA paper
  • High Performance Software Foundation

Roundtable

No meeting next two weeks.

Silvio reports that compiles on Aurora work better if you set the SIMD width to 16 (from 32). Ken says that will probably be OK for VTK-m in general.

The Kokkos device adapter test is failing. This might be similar to earlier issues that were fixed. Mark will take a look at the detailed log.