NERSCPowering Scientific Discovery for 50 Years
Login
My NERSC | A-Z Index
For Users

Need Help?

Help Portal

help.nersc.gov 
(requires login)

Accounts Portal

IRIS Accounts Interface 
(requires login)

Allocations Portal

ercap.nersc.gov 
(requires login)

Code of Conduct

NERSC Initiative for Scientific Exploration (NISE) 2011 Awards

Performance Enhancements for Three-Dimensional Fast Fourier Transforms Library P3DFFT

Dmitry Pekurovsky, University of California San Diego

NISE project m1243.

NISE Award: 350,000 Hours
Award Date: March 2011

Three-dimensional Fast Fourier Transforms are used in a wide range of scientific computing applications, such as DNS turbulence, astrophysics, material science, oceanography, 3D tomography. They are often the bottleneck for performance of such codes on petascale systems. Any performance improvements in P3DFFT will translate directly into benefits for many users in these fields, since this open-source library is well-known and used as a building block in cutting-edge research codes.

P3DFFT is an open-source library for ultra-scalable FFTs in three dimensions (http://code.google.com/p/p3dfft). It is used in many applications designed for large-scale simulations on Peta-scale systems. The library has demonstrated excellent scalability on up to O(10^5) cores on systems such as Jaguar and IBM BG.

This project will be focused on further improving performance and scalability of P3DFFT, putting it in a position to take advantage of the latest technological advances in supercomputer design. This work will follow two main directions.

  1. Currently P3DFFT relies on MPI_Alltoall(v) for a crucial global transpose operation. This is a blocking operation and does not allow any overlap of communication with computation.Such overlap will be explored, through use of one-sided communication (via MPI-2 or SHMEM).
  2. The multicore design of Hopper, as well as most other petascale platforms, necessitates adaptation of the default standard MPI programming model. In particular, a hybrid MPI/OpenMP implementation will be developed and tested,in order to make sure the nodes and interconnect are used in an optimal way.

Table of Contents

  1. Overview
  2. New potential energy surface for ozone molecule
  3. Hybrid OpenMP/MPI approach to R-matrix scattering
  4. Global Effects on the Dynamics of Plasmoids and Flux Ropes During Magnetic Reconnection
  5. MD Simulations of Liquid Water Films at the Boundary Between Gas Hydrate and Mineral Surfaces
  6. Systems Biology Knowledge Base
  7. Sparse Lanczos and GMRES Solvers in Co-Array Fortran (CAF)
  8. Numerical Study of Mode Conversion and Ion Heating in Laboratory-Relevant Plasmas
  9. Large Scale Numerical Study of Bootstrap Current in Edge Pedestal Plasma
  10. Ab initio simulation of mechanical properties of bulk metallic glass and Mo-bases alloys
  11. Improving Algorithms and Resolution for the Ocean-Atmosphere Reanalysis for Climate Applications OARCA (1850-2012)
  12. High resolution modeling to identify the location of the meltwater flood responsible for the Younger Dryas cold episode
  13. Numerical Investigations of Symmetry Breaking, Magnetism and the Pseudogap in the Three-Orbital Hubbard Model
  14. Breaking Magnetic Field Lines with Turbulence
  15. Development of CALPHAD type approach to thermal conductivity aided by first-principles phonon approach
  16. An Efficient Real-Space Approach to Calculating Auger Spectra
  17. Surface Instabilities and Subgrid Scale Models in Computational Fluid Dynamics
  18. Self-Healing of Polymer Films
  19. Computational Studies of Methanol Steam Reforming
  20. Laser Acceleration of Ions Using Few Times Critical Density Gas Targets
  21. An Integrative Strategy to Model Complex Biological Assemblies
  22. Unveiling Microbial Carbon Cycling Processes in Key U.S. Soils using "omics"
  23. Semiclassical Approaches for Clean Energy Resources
  24. Tall Tower Wind Energy Monitoring and Numerical Model Validation in Southern Nevada
  25. Threading EnergyPlus Using OpenMP
  26. First-Principles Calculation of Phonon-Assisted Optical Absorption in Indirect-Band-Gap Semiconductors.
  27. Deep Defect States and Valence Skipping in Narrow Band Gap Semiconductors
  28. Material Simulations in Joint Center for Artificial Photosynthesis (JCAP)
  29. High Fidelity Study of Multiscale Multiphase Lubricant Flow in High-Speed Gear Systems
  30. Semiclassical Approaches for Clean Energy Resources
  31. First Principles Characterization of van der Waals Dispersion Interactions
  32. Simulating ETG Plasma Turbulence with Impurities
  33. Coupled Electronic and Nuclear Dynamics in Enzymes and Photocatalysts
  34. Proton-coupled electron transfer in photocatalysis and protein translocation in biosynthesis: Bridging lengthscales and timescales in molecular simulation
  35. Three Dimensional Models for Martensitic Phase Transformations
  36. Molecular Simulations of the Release of Phosphate and ADP from F1-ATPase to Aid in the Understanding of the Function of F1-ATPase
  37. Turbulence over Complex Terrain: A Wind Energy Perspective
  38. Performance Enhancements for Three-Dimensional Fast Fourier Transforms Library P3DFFT
  39. Accelerated Materials Design Towards the Materials Genome
  40. Triple Photoionization of the Li atom
  41. Implementation of C-MAD Parallel Code to Include New Physics and Capabilities for Future Particle Colliders
  42. Morphology of Young Core-Collapse Supernova Remnants from Multi-Physics Three-Dimensional Simulations
  43. Large High-Resolution Galaxy Simulation Program
  44. Towards High Performance Urban Canopy Flows
  45. Evaluating PGAS scientific graph analysis codes on the Gemini interconnect
  46. Representing Convective Clouds in Global Circulation Models
  47. Simulation of Inelastic Decoherence in Electronic Transport Through Nanoscale Structures
  48. Improving Scalability of an Eigensolver for the Kohn-Sham Electronic Structure Problem
  49. Phonon, thermodynamic and diffusion properties of lithium-ion batteries from first-principles calculations
  50. CACHE novel architecture research
  51. High performance simulation of reactive transport processes at the pore scale
  52. First-Principles Calculations of Multiferroic Oxides
  53. High resolution integration of CAM5 for extreme weather research
  54. Modeling and Simulation of High Dimensional Stochastic Multiscale PDE Systems at the Exascale
  55. Theoretical Investigation on the Properties of Energy Materials
Last edited: 2013-11-18 13:33:50
Berkeley Lab logo U.S. Department of Energy Office of Science logo