Patrick Groeneveld
ABSTRACT
Solving 3-D partial differential equations in a Finite Element model is computationally intensive and requires extremely high memory and communication bandwidth. This paper describes a novel way where the Finite Element mesh points of varying resolution are mapped on a large 2-D homogenous array of processors. Cerebras developed a novel supercomputer that is powered by a 21.5cm by 21.5cm Wafer-Scale Engine (WSE) with 850,000 programmable compute cores. With 2.6 trillion transistors in a 7nm process this is by far the largest chip in the world. It is structured as a regular array of 800 by 1060 identical processing elements, each with its own local fast SRAM memory and direct high bandwidth connection to its neighboring cores. For the 2021 ISPD competition we propose a challenge to optimize placement of computational physics problems to achieve the highest possible performance on the Cerebras supercomputer. The objectives are to maximize performance and accuracy by optimizing the mapping of the problem to cores in the system. This involves partitioning and placement algorithms.
- Kamil Rocki, Dirk Van Essendelft, Ilya Sharapov, Robert Schreiber, Michael Morrison, Vladimir Kibardin, Andrey Portnoy, Jean Francois Dietiker, Madhava Syamlal, Michael James, "Fast Stencil-Code Computation on a Wafer-Scale Processor", Supercomputing 2020, https://arxiv.org/pdf/2010.03660.pdfGoogle Scholar
- Michael James, Marvin Tom, Patrick Groeneveld, Vladimir Kibardin, "Physical Mapping of Neural Networks on a Wafer-Scale Deep Learning Accelerator", Proc International Symposium on Physical Design 2020, pp 140--145,Google Scholar
- M. James, P. Groeneveld, M. Tom,"ISPD 2020 Contest: Wafer Scale Deep Learning Accelerator Placement", Jan 2, 2020, Accessed on Jan 31, 2020. [online] Available: https://www.cerebras.net/ispd-2020-contest/Google Scholar
Index Terms
- ISPD 2021 Wafer-Scale Physics Modeling Contest: A New Frontier for Partitioning, Placement and Routing
Recommendations
MP-Trees: A Packing-Based Macro Placement Algorithm for Modern Mixed-Size Designs
In this paper, we present a new multipacking-tree (MP-tree) representation for macro placements to handle modern mixed-size designs with large macros and high chip utilization rates. Based on binary trees, the MP-tree is very efficient, effective, and ...
Modern floorplanning based on B*-tree and fast simulated annealing
Unlike classical floorplanning that usually handles only block packing to minimize silicon area, modern very large scale integration (VLSI) floorplanning typically needs to pack blocks within a fixed die (outline), and additionally considers the packing ...
The ISPD-2011 routability-driven placement contest and benchmark suite
ISPD '11: Proceedings of the 2011 international symposium on Physical designThe last few years have seen significant advances in the quality of placement algorithms. This is in part due to the availability of large, challenging testcases by way of the ISPD-2005 [17] and ISPD-2006 [16] placement contests. These contests ...
Comments