17200

Lattice Quantum Chromodynamics on Intel Xeon Phi based supercomputers

Peter Labus
ICTP & SISSA Trieste, Italy
SISSA Trieste, 2016

@article{urbach2016lattice,

   title={Lattice Quantum Chromodynamics on Intel RO Xeon Phi TM based supercomputers},

   author={Urbach, Carsten and Labus, Peter},

   year={2016}

}

Download Download (PDF)   View View   Source Source   

1780

views

The aim of this master’s thesis project was to expand the QPhiX library for twisted-mass fermions with and without clover term. To this end, I continued work initiated by Mario Schrock et al. [63]. In writing this thesis, I was following two main goals. Firstly, I wanted to stress the intricate interplay of the four pillars of High Performance Computing: Algorithms, Hardware, Software and Performance Evaluation. Surely, algorithmic development is utterly important in Scientific Computing, in particular in LQCD, where it even outweighed the improvements made in Hardware architecture in the last decade-cf. the section about computational costs of LQCD. It is strongly influenced by the available hardware-think of the advent of parallel algorithms – but in turn also influenced the design of hardware itself. The IBM BlueGene series is only one of many examples in LQCD. Furthermore, there will be no benefit from the best algorithms, when one cannot implement the ideas into correct, performant, user-friendly, read- and maintainable (sometimes over several decades) software code. But again, truly outstanding HPC software cannot be written without a profound knowledge of its target hardware. Lastly, an HPC software architect and computational scientist has to be able to evaluate and benchmark the performance of a software program, in the often very heterogeneous environment of supercomputers with multiple software and hardware layers. My second goal in writing this thesis was to produce a self-contained introduction into the computational aspects of LQCD and in particular, to the features of QPhiX, so the reader would be able to compile, read and understand the code of one truly amazing pearl of HPC [40].
Rating: 2.5/5. From 1 vote.
Please wait...

* * *

* * *

HGPU group © 2010-2024 hgpu.org

All rights belong to the respective authors

Contact us: