Xeon Phi Reseach & Development

As well as enabling and optimising software on Intel Xeon Phi platforms, ICHEC also delivers Education, Training and Outreach activities around Xeon Phi technology to both academia and industry. Below, we list some of our recent activities in this area (including publications), which we aim to expand on. If you are interested to receive tailored training or would like to increase your team's understanding and awareness of Xeon Phi technology please contact: Dr Simon Wong

• IPCC EMEA User Group Meeting Dublin 2015
• IPCC Code Dungeon Dublin 2015
• High Performance Parallelism Pearls 1st Ed. Book Chapter
• High Performance Parallelism Pearls 2nd Ed. Book Chapter

Enabling DL_POLY_4 on Intel Xeon Phi-based clusters

DL_POLY_4 is a general-purpose classical molecular dynamics (MD) simulation application developed at the STFC’s Daresbury Laboratory. The package is used to model the atomistic evolution of the full spectrum of models commonly employed in the materials science, solid-state chemistry, biological simulation and soft condensed-matter communities. Using MPI and OpenMP, ICHEC aims to enable and optimise a wide range of functionalities within DL_POLY_4 with the particular goal of accelerating discovery in areas of materials science that have been identified as national research priority areas.

More information about DL_POLY and how to download the code can be found here.

Publications/Proceedings/Conferences/Workshops:

Work to date has been presented at:

• Intel Software Conference (ISTEP EMEA), Seville, Spain (2015)
• IPCC EMEA User Group Meeting, Royal Irish Academy, Dublin (2015)
• IPCC EMEA User Group Meeting, Zuse Institute Berlin, Berlin (2014)
• CECAM Many-Core Workshop 2014, Science Technology and Facilities Council (STFC), Daresbury UK (2014)
• Intel Theater, Supercomputing Conference (SC'14), New Orleans (2014)

Enabling the HARMONIE weather forecasting code on Intel Xeon Phi-based clusters

HARMONIE is a regional weather forecast model developed and used operationally for short-range forecasts by the HIRLAM consortium of mostly northern-European countries, including Ireland (see www.hirlam.org). There is an insatiable demand from this community for better HPC performance without sacrificing code portability and transparency. ICHEC aims to harness the parallelism of Intel’s new Intel Xeon Phi platform to accelerate HARMONIE in order to provide tangible operational benefits such as higher-resolution forecasts in shorter run-times. At the same time, any changes made to HARMONIE should be minimally disruptive, and be sustainable over future versions of HARMONIE as well as future versions of Xeon Phi hardware. Here ICHEC is building on over a decade of experience of working closely with Ireland’s national meteorology agency, Met Éireann, who use ICHEC’s latest HPC system Fionn.

More information about HARMONIE can be found here

Publications/Proceedings/Conferences/Workshops:

Work to date has been presented at:

• ALADIN/HIRLAM 25th Workshop/All-Staff Meeting, Helsingor, Denmark (2015)
• IPCC EMEA User Group Meeting, Royal Irish Academy, Dublin (2015)
• IPCC EMEA User Group Meeting, Zuse Institute Berlin, Berlin (2014)
• ECMWF Workshop on Scalability, Reading UK (2014)
• ALADIN/HIRLAM 24th Workshop/All-Staff Meeting, Bucharest (2014)

Preparing a Seismic Imaging Code for Intel's Knights Landing Xeon Phi processor

ICHEC, in collaboration with Tullow Oil plc, is preparing a seismic imaging code for the Knight’s Landing (KNL) Xeon Phi processor. The seismic imaging code in question is a Reverse Time Migration (RTM) code based on an unconventional rotated staggered grid (RSG) method and has undergone a significant amount of Xeon/Xeon Phi-focused re-engineering at ICHEC. The code uses an explicit finite difference (FD) scheme of variable spatial order, and first order in time, to model wave propagation in three isotropy cases. The code is parallelised with MPI and OpenMP, for wide hardware compatability. Vectorisation, and efficient cache utilisation were carefully considered in the kernel design, while attempting to maintain portability and maintainability. The stencil-based kernels of the code have low arithmetic intensity and are bound by the main memory bandwidth of the Xeon/Xeon Phi, which we have successfully alleviated by making highly efficient use of shared last level cache (LLC) on Xeon. The latter optimization is one that throws up some interesting challenges for achieving optimal performance on Xeon Phi, which we are investigating further in strong collaboration with Intel engineers.

Publications/Proceedings/Conferences/Workshops:

Work to date has been presented at:

• PARCO'15-Symposium on Experiences of porting and optimising code for Xeon Phi processors, Preparing a Seismic Imaging Code for the Intel Knights Landing
• ISC'15 Workshop: "The Road to Application Performance on Intel Xeon Phi Coprocessor”

Preparing a Quantum Collisions code for Intel's Knights Landing Xeon Phi processor

ICHEC, in collaboration with the UK's STFC, is preparing the quantum collisions package, PFARM, for the Knight’s Landing (KNL) Xeon Phi processor. PFARM is itself part of a suite of programs based on the ‘R- matrix’ ab-initio approach to the variational solution of the many-electron Schrödinger equation for electron-molecule, electron-atom, and electron-ion scattering. The package has been used to calculate electron collision data for astrophysical applications (such as: the interstellar medium, planetary atmospheres) , as well as other applications such as plasma modelling and fusion reactor impurities. The code has been adapted for electron (positron) molecule collisions in these and other application areas, and also to study RNA/DNA base molecule resonances relevant to biological radiation damage. ICHEC is focused on modernising the EXAS code within the PFARM suite so as to be able to exploit future extreme-scale Xeon Phi processor-based systems. With the recent significant re-engineering of the code to reduce communication over-head, early stage results strongly indicate the potential performance benefits of running the code on Xeon Phi in both native and "MPI symmetric mode".

Publications/Proceedings/Conferences/Workshops:

Work to date has been presented at:

• EMiT'15 Conference Paper: "Optimising Performance of the HPC electron collisions R-matrix code PFARM on the Intel Xeon Phi"

Enabling R packages on the Intel Xeon Phi platform

Data analytics is broadly recognised as a valuable tool for civil society and industry alike, for example it was clearly identified in the Irish Government's Action Plan for Jobs 2013. In this context the R language is becoming more widely used for increasingly challenging computational tasks. ICHEC's role involves identifying key use-cases and working with users and the R community to ensure that it can best exploit the computational power offered by the Xeon Phi platform.

Enablement of R packages on Intel Xeon Phi is driven by industry needs and is carried out in partnership with industry on a case by case basis. If you are interested in partnering with ICHEC in this area, please contact us

Publications/Proceedings/Conferences/Workshops:

Work to date has been presented at:

• Super-R Birds of a Feather (BOF), International Supercomputing Conference (ISC'14), Leipzig (2014)

Below, you can find a list of some of the past Xeon Phi R&D projects that ICHEC has recently been involved in:

• Enabling Smeagol on Xeon Phi: Lessons Learned (FP7 PRACE 1IP). 
• Enabling the UCD-SPH code on Xeon Phi (FP7 PRACE 1IP).
• Enabling PRMAT, on Xeon Phi-based clusters (FP7 PRACE 2IP). 
• Solving the Minimum Number of Sudoku Clues Problem on Xeon Phi (User Support).