An HPC Programming Model for the Exascale Age

By Christian Simmendinger, T-Systems Solutions for Research and Daniel Grünewald, Fraunhofer ITWM, CC-HPC

June 26, 2012

As the supercomputing faithful prepare for exascale computing, there is a great deal of talk about moving beyond the two-decades-old MPI programming model . The HPC programmers of tomorrow are going to have to write codes that are able to deal with systems hundreds of times larger than the top supercomputers of today, and the general feeling is that MPI, by itself, will not make that transition gracefully. One of the alternatives being offered is a PGAS model known as GASPI, which was the subject of an extended session at last week’s International Supercomputing Conference.

GASPI, which stands for Global Address Space Programming Interface, is, as the name suggests, a partitioned global address space (PGAS) API. The GASPI standard is focused on three key objectives: scalability, flexibility and fault tolerance. It follows a single program multiple data (SPMD) approach and offers a small, yet powerful API composed of synchronization primitives, synchronous and asynchronous collectives, fine grained control over one-sided read and write communication primitives, global atomics, passive receives, communication groups and communication queues.

Essentially it uses one-sided RDMA-driven communication in a PGAS environment. As such, GASPI aims to initiate a paradigm shift from bulk-synchronous two-sided communication patterns towards an asynchronous communication and execution model.

With today’s ever increasing number of processes, a transition from bulk-synchronous communication towards an asynchronous programming model seems to be inevitable. Elapsed time for bulk-synchronous communication potentially scales with the logarithm of the number of processes, whereas the work assigned to a single process potentially scales with a factor of 1/(number of processes).

Hence, the scalability of bulk-synchronous communication patterns appears to be limited at best. Despite recent efforts to support true asynchronous communication, the message passing standard of MPI to a large extent still focuses on two-sided semantics and bulk-synchronous communication.

At the same time, fault tolerance also becomes a larger issue as machines expand in size. On systems with large number of processes, all non-local communication should be prepared for a potential failure of one of the communication partners. In GASPI this is accomplished by providing a timeout value as an argument to all non-local communication calls and the possibility to check for the state of each of the communication partners. The model also allows for the dynamic substitution of a failed process.

GASPI does not enforce a specific memory model, like, for example, the symmetric distributed memory management of OpenSHMEM. Rather GASPI offers PGAS in the form of configurable RDMA pinned memory segments. Since an application can request several segments in GASPI symmetric, asymmetric or stack based memory management models can readily coexist.

With PGAS, every thread can asynchronously read and write the entire global memory of an application. On modern machines with RDMA engines, an asynchronous PGAS programming model appears as a natural extension and abstraction of available hardware functionality. For systems with DMA engines (such as tile architectures), this also holds true for a node-local level.

While the GASPI API readily can support the various communication patterns of MPI by means of an add-on library, the reverse is not true. GASPI for example supports RDMA access to arbitrarily distributed data, which allows the programmer a direct RDMA write access from a local send halo of an unstructured mesh into the corresponding remote receive halo.

The GASPI API has been designed to coexist with MPI and hence in principle provides the possibility to complement MPI with a partitioned global address space. We note however, that while such an approach provides an opportunity for increased scalability, fault–tolerant execution will not be possible due to the corresponding limitations of MPI.

GASPI inherits much of its design from the Global address space Programming Interface (GPI), which was developed in 2005 at the Competence Center for High Performance Computing (CC-HPC) at Fraunhofer ITWM. GPI is implemented as a low-latency communication library and is designed for scalable, real-time parallel applications running on cluster systems. It provides a PGAS API and includes communication primitives, environment run-time checks and synchronization primitives such as fast barriers or global atomic counters.

GPI communication is asynchronous, one-sided and, most importantly, does not interfere with the computation on the CPU. Minimal communication overhead can be realized by overlapping communication and computation. GPI also provides a simple, run-time system to handle large data sets, as well as dynamic and irregular applications that are I/O- and compute-intensive. As of today, there are production-quality implementations for x86 and IBM Cell/B.E architectures.

GPI has been used to implement and optimize CC-HPC industry applications like the Generalized Radon Transform (GRT) method in seismic imaging or the seismic work flow and visualization suite PSPRO. Today, GPI is installed on Tier 0 supercomputer sites in Europe, including the HLRS in Stuttgart and the Jülich Supercomputing Centre.

The GPI library has yielded some promising results in a number of situations. In particular, GPI outperforms MPI in significant low-level benchmarks. For process to process communication, GPI asynchronous one-sided communication, as opposed to both MPI one-sided communication and MPI bulk-synchronous two sided-communication, delivers full hardware bandwidth. As a function of message size, GPI reaches its peak performance much earlier than MPI.

A slightly more complex type of low-level benchmark is the two dimensional fast Fourier transformation on a distributed data set. We have compared two almost identical MPI and GPI implementations which feature the same communication pattern. Contrary to MPI, GPI was able to deliver near perfect scalability in a strong scaling setup.

GPI has also shown excellent scalability in a broad spectrum of typical real world HPC applications like the Computational Fluid Dynamics (TAU code from the DLR), or BQCD, a four dimensional nearest neighbor stencil algorithm. GPI has also been used in the implementation of fastest Unbalanced Tree Search (UTS) benchmark on the market.

Since many of the GASPI key objectives are shared by GPI, these results show the inherent potential of GASPI.

In 2010 the request for a standardization of the GPI interface emerged, which ultimately lead to the inception of the GASPI project in 2011. The work was funded by the German Ministry of Education and Science and included project partners Fraunhofer ITWM and SCAI, T-Systems SfR, TU Dresden, DLR, KIT, FZJ, DWD and Scapos.

The standard is currently being implemented in two flavors: a highly portable open source implementation based on GASNet and a commercial implementation aimed at ultimate performance. This latter implementation will be based on GPI. The TU Dresden, ZIH will provide profiling support for GASPI by means of extending the VAMPIR tool suite.

The GASPI project intends to drive the dissemination and visibility of the API by means of highly visible lighthouse projects in specific application domains, including CFD, turbo-machinery, weather and climate, oil and gas, molecular dynamics, as well as in the area of sparse and dense matrices. Amongst other implementations, the GASPI project will provide an asynchronous GASPI version of the Linpack benchmark.

There are a number of other projects that pursue similar goals to GASPI, the closest in spirit being OpenSHMEM. Ultimately the GASPI project aims at establishing a de-facto standard for an API for scalable, fault-tolerant and flexible communication in a partitioned global address Space. Whether that newly emerging standard will be called GASPI, however, remains to be seen.

Subscribe to HPCwire's Weekly Update!

Be the most informed person in the room! Stay ahead of the tech trends with industry updates delivered to you every week!

OSI Open AI Definition Stops Short of Requiring Open Data for LLMs

November 6, 2024

The movement toward open source AI made progress today when the Open Source Initiative released the first Open Source AI Definition (OSAID). While the OSAID provides one step forward, the lack of requirements around open Read more…

D-Wave Readies 4,400-plus-qubit Advantage2 System for Use

November 6, 2024

Quantum computing pioneer D-Wave today announced it had completed calibration and benchmarking the latest latest version of its Advantage2 quantum processor, a 4,400-plus-qubit device. D-Wave said that compared with the Read more…

Bill Gropp on ‘Different Approaches to AI’

November 6, 2024

Around this same time last year, I expounded on what the “Future of AI” may entail. A lot has happened in the 12 months since then, including new approaches, new trends and, yes, new complications. A lot of the ne Read more…

Google Cloud Sporting a New Look in HPC and AI Hardware

November 5, 2024

It's raining hardware at Google Cloud, with the company making major upgrades in advance of bringing Nvidia's Blackwell GPUs into its fold next year. The upgrades announced in late October include a preview of its new Read more…

Go (Mountain) West, Quantum Workers! CU, CUbit, and Elevate Quantum Issue Workforce Roadmap

November 5, 2024

Last week the University of Colorado (Boulder), the CUbit Quantum Initiative, and the Elevate Quantum consortium released workforce roadmap for educating and building a quantum workforce. “This roadmap provides a foun Read more…

Microsoft Azure & AMD Solution Channel

Join Microsoft Azure and AMD at SC24

Atlanta, Georgia is the place to be this fall as the high-performance computing (HPC) community convenes for Supercomputing 2024. SC24 will bring together an unparalleled mix of scientists, engineers, researchers, educators, programmers, and developers for a week of learning and sharing. Read more…

Collaboration Speeds Complex Chemical Modeling

November 4, 2024

A recent collaboration among researchers from HUN-REN Wigner Research Centre for Physics in Hungary and the Department of Energy's Pacific Northwest National Laboratory (PNNL), along with industry collaborators SandboxAQ Read more…

OSI Open AI Definition Stops Short of Requiring Open Data for LLMs

November 6, 2024

The movement toward open source AI made progress today when the Open Source Initiative released the first Open Source AI Definition (OSAID). While the OSAID pro Read more…

Bill Gropp on ‘Different Approaches to AI’

November 6, 2024

Around this same time last year, I expounded on what the “Future of AI” may entail. A lot has happened in the 12 months since then, including new approaches Read more…

Shutterstock 1179408610

Google Cloud Sporting a New Look in HPC and AI Hardware

November 5, 2024

It's raining hardware at Google Cloud, with the company making major upgrades in advance of bringing Nvidia's Blackwell GPUs into its fold next year. The upg Read more…

Go (Mountain) West, Quantum Workers! CU, CUbit, and Elevate Quantum Issue Workforce Roadmap

November 5, 2024

Last week the University of Colorado (Boulder), the CUbit Quantum Initiative, and the Elevate Quantum consortium released workforce roadmap for educating and bu Read more…

Collaboration Speeds Complex Chemical Modeling

November 4, 2024

A recent collaboration among researchers from HUN-REN Wigner Research Centre for Physics in Hungary and the Department of Energy's Pacific Northwest National La Read more…

High-Performance Storage for AI and Analytics Panel

October 31, 2024

When storage is mentioned in an AI or Big Data analytics context, it is assumed to be a high-performance system. In practice, it may not be, and the user eventu Read more…

Shutterstock_556401859

Role Reversal: Google Teases Nvidia’s Blackwell as It Softens TPU Rivalry

October 30, 2024

Customers now have access to Google's homegrown hardware -- its Axion CPU and latest Trillium TPU -- in its Cloud service.  At the same time, Google gave custo Read more…

AI Has a Data Problem, Appen Report Says

October 30, 2024

AI may be a priority at American companies, but the difficulty in managing data and obtaining high quality data to train AI models is becoming a bigger hurdle t Read more…

Shutterstock_2176157037

Intel’s Falcon Shores Future Looks Bleak as It Concedes AI Training to GPU Rivals

September 17, 2024

Intel's Falcon Shores future looks bleak as it concedes AI training to GPU rivals On Monday, Intel sent a letter to employees detailing its comeback plan after Read more…

Granite Rapids HPC Benchmarks: I’m Thinking Intel Is Back (Updated)

September 25, 2024

Waiting is the hardest part. In the fall of 2023, HPCwire wrote about the new diverging Xeon processor strategy from Intel. Instead of a on-size-fits all approa Read more…

Ansys Fluent® Adds AMD Instinct™ MI200 and MI300 Acceleration to Power CFD Simulations

September 23, 2024

Ansys Fluent® is well-known in the commercial computational fluid dynamics (CFD) space and is praised for its versatility as a general-purpose solver. Its impr Read more…

xAI Colossus: The Elon Project

September 5, 2024

Elon Musk's xAI cluster, named Colossus (possibly after the 1970 movie about a massive computer that does not end well), has been brought online. Musk recently Read more…

Shutterstock 1024337068

Researchers Benchmark Nvidia’s GH200 Supercomputing Chips

September 4, 2024

Nvidia is putting its GH200 chips in European supercomputers, and researchers are getting their hands on those systems and releasing research papers with perfor Read more…

AMD Clears Up Messy GPU Roadmap, Upgrades Chips Annually

June 3, 2024

In the world of AI, there's a desperate search for an alternative to Nvidia's GPUs, and AMD is stepping up to the plate. AMD detailed its updated GPU roadmap, w Read more…

Comparing NVIDIA A100 and NVIDIA L40S: Which GPU is Ideal for AI and Graphics-Intensive Workloads?

October 30, 2023

With long lead times for the NVIDIA H100 and A100 GPUs, many organizations are looking at the new NVIDIA L40S GPU, which it’s a new GPU optimized for AI and g Read more…

Nvidia Shipped 3.76 Million Data-center GPUs in 2023, According to Study

June 10, 2024

Nvidia had an explosive 2023 in data-center GPU shipments, which totaled roughly 3.76 million units, according to a study conducted by semiconductor analyst fir Read more…

Leading Solution Providers

Contributors

IBM Develops New Quantum Benchmarking Tool — Benchpress

September 26, 2024

Benchmarking is an important topic in quantum computing. There’s consensus it’s needed but opinions vary widely on how to go about it. Last week, IBM introd Read more…

Intel Customizing Granite Rapids Server Chips for Nvidia GPUs

September 25, 2024

Intel is now customizing its latest Xeon 6 server chips for use with Nvidia's GPUs that dominate the AI landscape. The chipmaker's new Xeon 6 chips, also called Read more…

Zapata Computing, Early Quantum-AI Software Specialist, Ceases Operations

October 14, 2024

Zapata Computing, which was founded in 2017 as a Harvard spinout specializing in quantum software and later pivoted to an AI focus, is ceasing operations, accor Read more…

Quantum and AI: Navigating the Resource Challenge

September 18, 2024

Rapid advancements in quantum computing are bringing a new era of technological possibilities. However, as quantum technology progresses, there are growing conc Read more…

US Implements Controls on Quantum Computing and other Technologies

September 27, 2024

Yesterday the Commerce Department announced export controls on quantum computing technologies as well as new controls for advanced semiconductors and additive Read more…

Google’s DataGemma Tackles AI Hallucination

September 18, 2024

The rapid evolution of large language models (LLMs) has fueled significant advancement in AI, enabling these systems to analyze text, generate summaries, sugges Read more…

Microsoft, Quantinuum Use Hybrid Workflow to Simulate Catalyst

September 13, 2024

Microsoft and Quantinuum reported the ability to create 12 logical qubits on Quantinuum's H2 trapped ion system this week and also reported using two logical qu Read more…

On Paper, AMD’s New MI355X Makes MI325X Look Pedestrian

October 15, 2024

Advanced Micro Devices has detailed two new GPUs that unambiguously reinforce it as the only legitimate GPU alternative to Nvidia. AMD shared new facts on its n Read more…

  • arrow
  • Click Here for More Headlines
  • arrow
HPCwire