EDGE wins ISC Research Poster Award

Illustration of the Knights Landing and Knights Mill architectures Illustration of Intel Xeon Phi Vector Processing Units (VPUs). Shown is a comparison of the Knights Landing and Knights Mill processors: a symmetric, single-pumped combo VPU is replaced by an asymmetric (single precision biased) VPU which is double-pumped for high efficiencies on the two-issue wide Xeon Phi frontend.

The poster “Deep Learning Hardware Accelerates Fused Discontinuous Galerkin Simulations” won the ISC Research Poster Award in the category “Artificial Intelligence & Machine Learning”. With seismic wave propagation in EDGE as the use-case, the work shows that high performance computing should exploit two important aspects of hardware developments in the years to come: a) new hardware features are biased towards improving single and lower precision performance, and b) computational capabilities increase faster than the processors’ memory performance. The award comes with a 300 Euro cash prize and a certificate. Details on the conference’s awards are available from http://isc-hpc.com. An electronic version of the award-winning poster is available from EDGE’s assets repository.

ISC High Performance 2018 and 11NCEE

Posters at ISC High Performance 2018 and 11NCEE

EDGE’s modeling and simulation pipeline is part of the two posters “Deep Learning Hardware Accelerates Fused Discontinuous Galerkin Simulations” and “EDGE: Benchmarking the Seismic Wave Propagation Solver”. The first poster will be presented at the ISC High Performance conference (06/24-06/28) in Frankfurt, Germany. The 11th National Conference on Earthquake Engineering (06/25-06/29) in Los Angeles, CA, USA is the venue of the second poster presentation.

The respective programs contain details on the presentations (ISC, 11NCEE). Additionally, both posters are also archived in EDGE’s assets repository.

LIBXSMM Brings Deep-learning “Lessons Learned” to Many HPC Applications

Floating point performance on 16 nodes for a large set of EDGE’s possible configurations Floating point performance on 16 nodes of the Knights Landing, Knights Mill and Skylake processor generation. Shown is a large set of EDGE’s possible configurations. All results are reported in terms of non-zero operations, contributing to EDGE’s solution. Dark gray bars represent the performance of single seismic forward simulations, while light gray bars show the performance of fused simulations.

The story “LIBXSMM Brings Deep-learning “Lessons Learned” to Many HPC Applications” by Rob Farber features EDGE as one of the first HPC applications, exploiting the Intel Xeon Phi processor for machine learning (Knights Mill). An important takeaway of the article is that only a comprehensive approach leads to optimal performance. Key factors for seismic wave propagation simulations in EDGE are: a) Extensive verification, including floating point precision as a modeling parameter, b) fused simulation technology for the exploitation of inter-simulation parallelism, and c) Just In Time (GIT) generated small sparse-matrix tensor kernels through the library LIBXSMM. The full story is available from the deep learning section of Medium. Further details are given in the slides of the IXPUG Middle East Conference 2018.


All Posts