Teleconference 2019-04-10
Attendees: Ken Moreland (SNL), Nick Davis (SNL), Ollie Lo, (LANL), Rob Maynard and others (Kitware), Dave Pugmire (ORNL), Abhishek Yenpure (UO), Hank Childs (UO), Matt Larsen (LLNL)
ECP Updates
- Thanks to Hank for getting KPP-3 integration metric done
- Final Design Report section submitted
- Starting to plan work detail for FY20 and beyond
- Basis of Estimate due May 1
- Basically need to provide budgets for FY20-23.
- Need to work with your financial people.
- Last chance to decide on subcontracts.
- We have been given lump sums for FY20 and FY21-FY23
- Ken will provide divisions among institutes shortly
- Lump sum includes a 9% decrease per year decrease
- Projects that deal with porting (our project was called out) should actually use more funds in the later years.
- Next steps
- Ken will assign money to each institution
- All PIs will report their breakdown of budget per year
- Each PI will work with their budget persons to fill out spreadsheet forms
- Questions to ask Mike:
- Is there a way to transfer FY20 funds to later (carryover)?
- Do non-labs submit their own budget or should it be included in the lab holding the subcontract?
- Review confluence page: https://docs.google.com/document/d/1L44kTtwKi0SgeSJxt62OGGI7hpfpte-OHm0V0B6vsg8/edit#heading=h.hk0so9awvgd1
- Budget pages: https://confluence.exascaleproject.org/download/attachments/61904390/BOE%20CD-2.xlsx?version=1&modificationDate=1554499415120&api=v2
- Basically need to provide budgets for FY20-23.
ECP deliverable updates
FY19Q3
- [MS-19/08] Point Movement 2
- Done except user's guide.
FY19Q4
- [MS-19/09] Path Geometry
- Not started
- Some of the functionality is already there
- [MS-20/01] Lightweight Cell Library
- Been doing some changes on the interface to the VTK-m cell library.
- Most of the underlying library is ready.
- Almost ready.
- Still need to implement VTK-m data around it.
- [MS-20/02] Specialized Data Models
- Some work on how XGC used a rotated triangular grid for a bunch of wedges.
- Need to generalize some code written by Matt.
- Pulling in code from other contributors.
- Connect up with Berk who is picking up XGC work from before.
- Need to brainstorm on what other features would be useful.
- [MS-20/03] Feature-Sensitive Surface Norm
- First pass in a MR
- Need to wait until CUDA 8 off all dashboards
- Should be walking the graph traversal
- First pass in a MR
- [MS-20/04] VTK-m Release 3
- On track
- Probably June would be a good time for a release.
- [MS-20/05] Cell Metrics
- Brent is getting back on this.
- [MS-20/06] Contouring
- Will model the problem after triangulation. Have different contouring for each cell type.
- Ollie looking into VTK implementation and flying edge paper.
- [MS-20/07] Advanced Flow Algorithms
- Merged in fix to do double precision correctly.
- Still deciding what we want to do.
- Planning on talking during DOECGF
- Hybrid parallel seems like a good way to go
- Dave is working on generalizing cell lookups and evaluators to work on all grid types
- Dave has been working on distributed memory advection
Rountable notes
We should try to collect some ECP data so that as we do reporting we can quickly generate relevant images. It would be great if we could also share across ALPINE and other vis projects.
- Terry Turton has collected some stuff from LANL. Not sure if can be sent out.
- She might have a list of distributed data in different places
Hackathon?
Rob has been working on getting rid of warnings from CUDA code. Some warnings have popped up because we are no longer suppressing lots of CUDA warnings about device calling host. These are legitimate issues.
Matt has been working with early CI testing on machines.
There is a customer building VTK-m with different compilers (with Sierra) with XLC and PGI. XLC and PGI take a long time to compile (with CUDA).
