1 of 1

VTK-m / Ascent in situ with Raleigh-Taylor Instability on 16k Sierra GPUs

VTK-m/

Catalyst/ with PyFR on Titan with 5000 GPUS

Dynamic types

Kenneth Moreland (PI) Oak Ridge National Laboratory, David Pugmire (Deputy PI) Oak Ridge National Laboratory,

David Rogers and Ollie Lo, Los Alamos National Laboratory, Berk Geveci Kitware, Inc., Hank Childs University of Oregon,

Mark Bolstad, Sandia National Laboratory, Silvio Rizzi, Argonne National Laboratory

The VTK-m software is DOE’s solution to develop and deploy scientific visualization software that take advantage of the shared-memory parallelism available on many-core CPUs and GPUs.

ECP scope:

Redeveloping, implementing, and supporting necessary visualization algorithms on many-core under VTK-m. There is a large base of complex, computationally intensive algorithms in regular use that need to be redesigned for advanced architectures. Updating the many critical scientific visualization algorithms in use today requires significant effort.

VTK-m partners with other ECP visualization efforts, including providing the many-core support that powers ECP ALPINE’s in situ efforts like Ascent, ParaView Catalyst, and VisIt LibSim.

Project Description

VTK-m: Updating HPC Visualization Software

for Exascale-Era Processors

2016

2017

2018

2019

2020

2021

2022

2023

VTK-m / Ascent integration into WarpX

VTK-m particle tracing accelerates fusion Poincaré for WDMApp by 60x

VTK-m

live in situ with Catalyst/ PyFR

VTK-m

in VisIt

VTK-m in

VTK and ParaView

Timeline

1.0

1.1

1.3

1.4

1.5

1.6

1.7

1.8

1.9

2.0

VTK-m Releases

Kokkos on Spock

Refactored filters

Articus

Faster HIP compiles

Raycasting renderer

Flow filters

New features:

- Pathlines

- Ghost zones

- Connected components

- Point merge

Thread safety

In situ visualization with WDMApp

Virtual methods removed

ECP Tool Integration

VTKm Functionality

ECP Hardware

ECP App Integration

Legend

Contour

Streams

Clip

Render

x86

CUDA

Xeon Phi

Radeon

Xe

Surface

Normals

Ghost Cells

Warp

Demonstrated

This image is of an idealized Inertial Confinement Fusion (ICF) simulation of a Rayleigh-Taylor instability with two fluids mixing in a spherical geometry.