24 May 2016 – Non-volatile memory (NVM) is playing a more important role in the memory architectures of HPC systems as illustrated by recent deployments and procurements. Yet there exist neither standard language constructs nor portable programming systems that provide support for these types of emerging memory architectures.
For older items, see our News Archive.
13 May 2016 – A new paper entitled, "OpenACC to FPGA: A Framework for Directive-based High-Performance Reconfigurable Computing" describes how the Future Technologies Group at Oak Ridge National Laboratory (ORNL) is attacking the well-known programmability and performance portability challenges of Field Programmable Gate Arrays (FPGAs). The researchers have extended the standard, portable OpenACC specification for FPGAs and evaluated it with a prototype implementation, demonstrating that OpenACC programs can be directly compiled for FPGAs.
A new FTG paper presents a survey of approximate computing techniques and has been highlighted on InsideHPC. Published in ACM Computing Surveys 2016, A Survey Of Techniques for Approximate Computing reviews nearly 85 papers on approximate computing in CPU, GPU and FPGA and various processor components (e.g. cache, main memory), along with approximate storage in SRAM, DRAM/eDRAM and non-volatile memories, e.g.
A new paper from FTG surveys asymmetric multicore processors, and has been highlighted on insideHPC. The paper reviews nearly 125 papers and has been accepted in ACM Computing Surveys 2015. Asymmetric multicore processors feature cores of different types (e.g. big and LITTLE) in the same processor and some commercial examples of them are Qualcomm Snapdragon 810, Samsung Exynos 5 Octa and Nvidia Tegra X1.
Three Future Technologies Group members recently presented 'Automated Characterization of Parallel Application Communication Patterns' at the 24th International ACM Symposium on High-Performance and Distributed Computing (HPDC'15) in Portland, Oregon in June, 2015. The paper describes an approach that uses automated search and a library of communication patterns to identify and parameterize a collection of patterns that best explains an MPI application's observed communication behavior. Philip Roth presented the work on behalf of his co-authors, Jeremy Meredith and Jeffrey Vetter.
24 Apr 2015 – This week, at the Exascale Application and Software Conference in Edinburgh, Scotland, the audience voted the FTG presentation on performance modeling as the conference’s Best Presentation. Jeffrey Vetter presented ‘Exploring Emerging Technologies in the Extreme Scale HPC Co-Design Space with Holistic Performance Modeling’ that discusses the importance of using performance modeling in the design of upcoming Extreme Scale platforms. The award was presented at the conference banquet at the The Royal College of Surgeons.
Oak Ridge National Lab, Penn State, and UCSB are pleased to announce the release of DESTINY (a 3D dEsign-Space exploraTIon tool for SRAM, eDRAM and Non-volatile memorY). DESTINY is a tool for modeling both 2D and 3D caches designed with five prominent conventional and emerging memory technologies: SRAM, eDRAM (embedded DRAM), PCM (or PCRAM), STT-RAM (or STT-MRAM) and ReRAM (or RRAM). DESTINY is intended to be a comprehensive tool, extending the capabilities of CACTI, CACTI-3DD, and NVSim, on which DESTINY is based. It can be used to model technology devices ranging from 22nm to 180nm.
In January, FTG hosted the Pegasus team from ISI/USC for a tutorial on their workflow system. This tutorial helped to kick off a new joint project, named Panorama, which examines predictive modeling and diagnostic monitoring of extreme science workflows.
A FTG paper on analytical modeling of application resiliency was selected as a Best Student Paper Finalist at SC14 http://sc14.supercomputing.org/schedule/event_detail?evid=pap192. The paper, entitled "Quantitately Modeling Application Resilency with the Data Vulnerability Factor," introduces a pragmatic, data-driven methodology to analyze application vulnerability based on a novel resilience metric: the data vulnerability factor (DVF).
Philip C. Roth, member of the ORNL Future Technologies Group since 2004, is part of the team organizing the 2014 Data Intensive Scalable Computing Systems (DISCS-2014) workshop at SC14. The workshop focuses on the intersection between traditional high performance computing and data intensive computing, and SC continues to be the perfect conference to host the workshop because it draws substantial numbers of attendees from across the globe interested in HPC and/or data intenstive computing.
Future Technologies Group members Philip Roth and Jeremy Meredith contributed to the 2012 CScADS workshop on Scientific Data and Analytics for Extreme-scale Computing, held July 30-August 2 at the Snowbird Ski & Summer Resort in Utah. Roth served on a panel titled "The effect of emerging architectures on data science," though he also included content on analyzing and visualizing performance data in his panel presentation. Meredith's work on a new scientific visualization and analysis library called EAVL was described in a presentation given by David Pugmire (also of ORNL).
A paper by Chao Chen, Yong Chen (former Future Tech postdoc and now on the faculty at Texas Tech University), and current Future Tech group member Philip Roth has been accepted to be presented at IEEE Cluster '12 in late September in Beijing, China. The paper, titled "DOSAS: Mitigating the Resource Contention in Active Storage Systems," quantifies the negative impact of contention in Active Storage systems and proposes a new approach for alleviating that negative impact.
Jeremy Meredith of ORNL’s Future Technologies group will present a new material interface reconstruction algorithm at EuroVis 2010, the Eurographics/IEEE Symposium on Visualization. This type of algorithm is used for generating geometric interfaces from material volume fraction data used in scientific applications, and is critical for a variety of visualization and analysis tasks. The paper, “Visualization and Analysis-Oriented Reconstruction of Material Interfaces”, by Meredith and collaborator Hank Childs from UC Davis and LBNL, evaluates this method against competing techniques.
Vinod Tipparaju, Weikuan Yu, and Jeffrey Vetter - members of ORNL’s Future Technologies group – have recently designed and developed a highly optimized Aggregate Remote Memory Copy Interface (ARMCI) runtime library on the Cray XT5 2.3 PetaFLOPs computer at Oak Ridge National Laboratory.
Collin McCurdy and Jeffrey Vetter - members of ORNL’s Future Technologies group - have recently developed Memphis: a tool that analyzes memory access patterns in scientific applications on Non-Uniform Memory Access (NUMA) architectures. The authors have been using Memphis to find and fix performance problems in several major DOE applications. These improvements have, so far, led to performance increases on the Cray XT5 at Oak Ridge of 23% for runs at scale of XGC1, and of 24% and 13% for single node runs of CAM and HYCOM, respectively.
Recent work by ORNL researcher Olaf Storaasli in collaboration with four Oslo colleagues at SINTEF (Norwegian NSF) has resulted in a comprehensive (31 page, 185 references) publication entitled \State-of-the-Art in Heterogeneous Computing\ recently published by the Journal of Scientific Programming, IOS Press. The publication identifies the key advances and trends over the years in computer hardware and software leading up to accelerators (Cell, GPU and FPGA) used in current and future High-Performance Supercomputers.
Vinod Tipparaju of the Future Technologies group contributed to a team that was selected as a finalist for 2009 ACM Gordon Bell Prize. ACM Gordon Bell Prize honors the world's highest-performing scientific computing applications. The team, comprising members from ORNL, Australian National University, Pacific Northwest National Laboratory, and Cray Inc., used a computational chemistry application known as NWChem to achieve 1.39 petaflops on Jaguar in a first principles, quantum mechanical exploration of the energy contained in clusters of water molecules.