In this tutorial we will review the KNL architecture, and discuss the differences between KNC and KNL. We will also discuss the impact of the different MCDRAM memory configurations and the different modes of cluster configuration. Recommendations regarding hybrid MPI+OMP execution when using KNL with the Intel OmniPath fabric will be provided.
We will also analyze the performance of some of the most popular applications in Stampede when running on KNL, and compare it to alternative platforms.
Algorithmic adaptations to use next-generation computers closer to their potential are underway in Oil & Gas and many other fields. Instead of squeezing out flops – the traditional goal of algorithmic optimality, which once served as a reasonable proxy for all associated costs – algorithms must now squeeze synchronizations, memory, and data transfers, while extra flops on locally cached data represent only small costs in time and energy. After decades of programming model stability with bulk synchronous processing, new programming models and new algorithmic capabilities (to make forays into, e.g., inverse problems, data assimilation, and uncertainty quantification) must be co-designed with the hardware. We briefly recap the architectural constraints, then concentrate on two kernels that each occupy a large portion of all scientific computing cycles: large dense symmetric/Hermitian systems (covariances, Hamiltonians, Hessians, Schur complements) and large sparse Poisson/Helmholtz systems (solids, fluids, electromagnetism, radiation diffusion, gravitation). We examine progress in porting solvers for these kernels (e.g., fast multipole, hierarchically low rank matrices, multigrid) to the hybrid distributed-shared programming environment, including the GPU and the MIC architectures.
We describe the coming convergence of traditional high performance computing and emerging cognitive computing workloads and discuss impacts to systems architecture and to the data center. To illustrate the opportunities and challenges, we provide some examples of applications in collaborative development in science domains and describe a proof-of-concept effort which deploys a container based high performance computing and cognitive computing software stack in IBM Research.
Historically, innovations in supercomputing have been eagerly adopted and leveraged by the O&G industry and the industry is poised to take advantage of successful exascale-class systems coming in the future. This talk will recap the last 10 years of supercomputing, highlighting major accomplishments, milestones, and technological advances. The talk will then speculate on what might be coming over the next 10 years as the supercomputing landscape continues to grow and change. We will look at the future in the context of supercomputing’s growing importance as a key component of competitive business practices, covering new and emerging technologies and also how supercomputing is enabling next generation applications and workflows.
