Using NVIDIA Index in ALYA MultiPhysics Christopher Lux (NV), Vishal - - PowerPoint PPT Presentation

▶
using nvidia index in alya multiphysics
SMART_READER_LITE
LIVE PREVIEW

Using NVIDIA Index in ALYA MultiPhysics Christopher Lux (NV), Vishal - - PowerPoint PPT Presentation

Jan 19, 2024 102 likes •593 views

www.bsc.es Interactive HPC: Large Scale In-Situ Visualization Using NVIDIA Index in ALYA MultiPhysics Christopher Lux (NV), Vishal Mehta (BSC) and Marc Nienhaus (NV) May 8 th 2017 Barcelona Supercomputing Center Marenostrum 4 13.7

slide-1
SLIDE 1

www.bsc.es

Interactive HPC: Large Scale In-Situ Visualization Using NVIDIA Index in ALYA MultiPhysics

Christopher Lux (NV), Vishal Mehta (BSC) and Marc Nienhaus (NV) May 8th 2017

slide-2
SLIDE 2

Barcelona Supercomputing Center

Marenostrum 4

  • 13.7 PetaFlop/s
  • General Purpose Computing

▪ 3400 nodes of Xeon, 11 PF/s

  • Emerging Technologies

▪ Power 9 + Pascal  1.5 PF/s ▪ Knights Landing and Knights Hill  0.5 PF/s ▪ 64bit ARMv8  0.5 PF/s

2

slide-3
SLIDE 3

Research at BSC

3

EARTH SCIENCES

To develop and implement global and regional state-

  • f-the-art models for short-

term air quality forecast and long-term climate applications

LIFE SCIENCES

To understand living

  • rganisms by means of

theoretical and computational methods (molecular modeling, genomics, proteomics) CASE

To develop scientific and engineering software to efficiently exploit super- computing capabilities (biomedical, geophysics, atmospheric, energy, social and economic simulations)

COMPUTER SCIENCES

To influence the way machines are built, programmed and used: programming models, performance tools, Big Data, computer architecture, energy efficiency

slide-4
SLIDE 4

4

ALYA System: Large Scale Computational Mechanics

slide-5
SLIDE 5

5

slide-6
SLIDE 6

6

ALYA HPC Context

slide-7
SLIDE 7

7

ALYA HPC Context

slide-8
SLIDE 8

ALYA RED

8

slide-9
SLIDE 9

9

Computational Cardiac Model

Applications ▪ Pacemaker applications ▪ Computational analysis of malfunctioning tissue patches ▪ Computational Drug testing on cardiac tissue.

slide-10
SLIDE 10

10

Computational Cardiac Model

slide-11
SLIDE 11

11

Pacemaker Application

slide-12
SLIDE 12

12

Pacemaker Application

slide-13
SLIDE 13

13

Computational Drug Testing

slide-14
SLIDE 14

Christopher Lux (NV), Vishal Mehta (BSC) and Marc Nienhaus (NV) May 8th 2017

INTERACTIVE HPC: LARGE SCALE

IN-SITU VISUALIZATION USING NVIDIA INDEX IN ALYA MULTIPHYSICS

slide-15
SLIDE 15

15

NVIDIA INDEX

GPU-cluster aware solution High-quality and scalable visualization of large-scale datasets In-situ visualization Commercial software Available and deployed in production

Scalable, Interactive Visual Computing

slide-16
SLIDE 16

16

NVIDIA Quadro Workstation NVIDIA Quadro VCA or DGX-1

NVIDIA HPC Clusters

Performance, dataset size, number of pixels, visual quality…

SCALABILITY AS ENABLER

slide-17
SLIDE 17

17

Scientific Data Visualization

slide-18
SLIDE 18

18

Time-Varying Data Visualization

slide-19
SLIDE 19

19

Time-Varying Data Visualization

Simulation data source: A Numerical Study of High-Pressure Oxygen/Methane Mixing and Combustion of a Shear Coaxial Injector, Nan Zong & Vigor Yang, AIAA 2005

slide-20
SLIDE 20

20

In-Trans and In-Situ Visualization

slide-21
SLIDE 21

21

Computational Heart

slide-22
SLIDE 22

22

BACKGROUND

slide-23
SLIDE 23

23

DISTRIBUTED PARALLEL RENDERING

[..]

* image compositing

[..]

Sort-Last Rendering (multi-GPU)

slide-24
SLIDE 24

24

DISTRIBUTED PARALLEL RENDERING

[..]

* image compositing

Sort-Last Rendering (Cluster of multi-GPU Nodes)

[..]

Cluster of VCAs

slide-25
SLIDE 25

25

DISTRIBUTED DATATYPES

Volume datatypes

▪ Regular ▪ Sparse ▪ Unstructured/Irregular

Surface-geometry datatypes

▪ Height field ▪ Triangle mesh

Various Application Domains

slide-26
SLIDE 26

26

IN-TRANS AND IN-SITU VISUALIZATION

slide-27
SLIDE 27

27

TRADITIONAL VISUALIZATION PIPELINE

5/15/2 017 Simulation Cluster Data Storage

e.g. Unstructured Data

NVIDIA IndeX Visualization Cluster

slide-28
SLIDE 28

28

Stream

Visualize and Animate Interact and Explore

TIME-SERIES DATA VISUALIZATION

Visualize Pre-calculated ALYA Simulation Results

Terabyte time-varying simulation data of nasal system

slide-29
SLIDE 29

29

IN-SITU (IN-TRANS) VISUALIZATION PIPELINE

Simulation Cluster NVIDIA IndeX Visualization Cluster Network

Unstructured Data Unstructured Data Unstructured Data

slide-30
SLIDE 30

30

IN-SITU VISUALIZATION PIPELINE

5/15/2 017 Combined Simulation and Visualization Cluster

slide-31
SLIDE 31

31

IN-SITU/IN-TRANS SUPPORT

Parallel jobs executed locally or remotely Direct access to local host and device data Fast RDMA memory transfers User-defined affinity and spatial subdivision Application-driven updates

▪ Push updated data when ready

Rendering-driven updates

▪ Request computation updates for active data

Compute Result Integration

Clustered neuron activity

slide-32
SLIDE 32

32

IN-TRANS RESULT TRANSFERS

GPUDirect RDMA NVLink high-speed interconnect between system memory and GPU (IBM and NVIDIA)

Fast Data Transfers to Rendering Nodes/GPUs

Page-Locked System Memory Page-Locked System Memory

RDMA

(over InfiniBand) CUDA GPU Memory CUDA GPU Memory

slide-33
SLIDE 33

33

Visualize Interact and Explore

COMPUTATIONAL HEART

In-Situ Simulation and Visualization

Simulate/Compute (ALYA) Parameterize and Steer

slide-34
SLIDE 34

34

NVIDIA INDEX AVAILABILITY

slide-35
SLIDE 35

35

NVIDIA INDEX 1.4

In-Situ/In-Trans Visualization Support

NVIDIA IndeX 1.4 (released 07/2016)

Support of 32bit, 16bit, 8bit fixed point, 32bit floating point and RGBA (8bit) regular volumes Dynamic streaming and GPU caching of time-varying volume data Irregular volumes and sparse volumes Built-in volume shading capabilities Multi-view capabilities Zero-copy RDMA/GPUDirect compute integration infrastructure User-defined affinity and spatial subdivision support Architecture and API for in-situ/in-trans visualization (compute integration) Dynamic workload balancing Advanced CUDA memory management, error handling and logging MPI/NVIDIA IndeX interprocess coupling (CUDA IPC and shared memory)

slide-36
SLIDE 36

36

NVIDIA INDEX 1.5 OUTLOOK

User-defined Rendering Kernel Components

slide-37
SLIDE 37

37

NVIDIA INDEX 1.5 OUTLOOK

User-defined Rendering Kernel Components

slide-38
SLIDE 38

IN-SITU VISUALIZATIONS

38

slide-39
SLIDE 39

39

Challenges

Simulation

MPI MPI MPI

. .

Unstructured Mesh Uneven Spatial partitions Balanced computations

Index Rendering

MPI MPI MPI

. .

Cubical Scene region Even spatial partitions Balanced rendering load

  • Parallel Operations
  • Maintain frame rates
  • Steering Simulation
  • Data Affinity
  • Render @compute
  • In-trans approach
slide-40
SLIDE 40

40

Multi-code Coupling in ALYA

  • All

spatial interpolation

  • n

spatial domain in structured and unstructured meshes

  • Allows

setting send and receive frequencies to synchronize simulation times.

  • Allows coupling with third party

codes

  • Parallel

and Asynchronous MPI coupling

slide-41
SLIDE 41

41

Ingredients of Coupling in ALYA

  • WHAT

The underlying variables

  • WHERE

Surface, Volume, etc.

  • WHEN

Time step, iteration step

  • HOW

Algorithmic interpolation mpirun -np 8 Alya.x fluid : -np 4 Test.x : -np 4 Alya.x solidz

slide-42
SLIDE 42

42

Coupling for IN-SITU Visualizations

Simulation

MPI MPI MPI

. . Index Rendering

MPI MPI MPI

. .

  • Allows optimizing resources for compute

and render

  • Application

Driven Updates, push simulation data.

  • Allows inter-operability between coarser

and finer meshes, adjusting data updates.

  • Maintains high frame rates and allows

interaction with the volume.

  • Can couple multiple physics apps to a

single rendering app.

slide-43
SLIDE 43

43

Steering Simulations

  • Steering is Application Specific
  • Steering

simulations requires handling interrupts.

  • Interrupt communicated through

backward coupling.

  • A

general approach by scalar/vector interrupts, and user defined function to handle the variables of simulation.

Simulation Time S1 Time S2 Time S3

. .

Rendering

Time S1 Time S2 Time S3

. . Coupling

Interrupt handler Function applied to fields

slide-44
SLIDE 44

44

Summary: In-situ Visualization

  • Index enables better insights into simulation data through professional visualization techniques
  • Scalability is the enabler for HPC in-situ visualization.
  • Multiphysics coupling is the key to scalability, and resource management for in-situ.
slide-45
SLIDE 45

45

SELF-PACED LABS

NVIDIA IndeX for ParaView plugin hands-on Location: ▪ Self-paced lab area on lower level Dates: ▪ Monday 1:00 – 5:00pm ▪ Tuesday 9:30 – 11.30am ▪ Wednesday 1:00 – 5:00pm

Interactive HPC Volume Visualization in ParaView

slide-46
SLIDE 46

46

INTERACTIVE DEMO

Live demonstration of in-situ visualization Interactive steering of simulation parameters Location: ▪ NVIDIA demo-booth in exhibit hall 1

Interactiver HPC: Large Scale In-Situ Visualization using NVIDIA IndeX in ALYA MultiPhysics

slide-47
SLIDE 47

Christopher Lux, NVIDIA Vishal Mehta, BSC Marc Nienhaus, NVIDA