Design of a Portable Implementation of Partitioned Point-to-Point Communication Primitives.
Abstract not provided.
SEARCH · Search NASA
Search indexed NASA NTRS and DOE OSTI research on propulsion, heat transfer, battery materials and energy systems. Follow report and document links to the original sources.
Quote a phrase for an exact phrase match. Source license links do not imply unrestricted reuse.
Abstract not provided.
Krylov methods are a key way of solving large sparse linear systems of equations but suffer from poor strong scalability on distributed memory machines. Furthermore, this is due to high synchronization costs from large numbers of collective communication calls alongside a low computational workload. Enlarged Krylov methods address this issue by decreasing the total iterations to convergence, an artifact of splitting the initial residual and resulting in operations on block vectors. In this article, we present a performance study of an enlarged Krylov method, Enlarged Conjugate Gradients (ECG), noting the impact of block vectors on parallel performance at scale. Most notably, we observe the increased overhead of point-to-point communication as a result of denser messages in the sparse matrix-block vector multiplication kernel. Additionally, we present models to analyze expected performance of ECG, as well as motivate design decisions. Most importantly, we introduce a new point-to-point communication approach based on node-aware communication techniques that increases efficiency of the method at scale.
Irregular communication often limits both the performance and scalability of parallel applications. Typically, applications individually implement irregular communication as point-to-point, and any optimizations are integrated directly into the application. As a result, these optimizations lack portability. It is difficult to optimize point-to-point messages within MPI, as the interface for single messages provides no information on the collection of all communication to be performed. However, the persistent neighbor collective API, released in the MPI 4 standard, provides an interface for portable optimizations of irregular communication within MPI libraries. This paper presents methods for implementing existing optimizations for irregular communication within neighborhood collectives, analyzes the impact of replacing point-to-point communication in existing codebases such as Hypre BoomerAMG with neighborhood collectives, and finally shows up to a 1.38x speedup on sparse matrix-vector multiplication communication within a BoomerAMG solve through the use of our optimized neighbor collectives. Here, the authors analyze three implementations of persistent neighborhood collectives for Alltoallv: an unoptimized wrapper of standard point-to-point communication, and two locality-aware aggregating methods. The second locality-aware implementation exposes an non-standard interface to perform additional optimization, and the authors present the additional 0.07x speedup from the extended interface. All optimizations are available in an open-source codebase, MPI Advance, which sits on top of MPI, allowing for optimizations to be added into existing codebases regardless of the system MPI install.
Practical applications of quantum computing are currently limited by the number of qubits that can be set with reasonable fidelities for each system. Therefore, a distributed quantum computing system with multiple quantum computers coherently connected is highly demanding. To realize the internode communication of quantum information, the software interface, Quantum Message Passing Interface (QMPI), leveraging the framework built for classical MPI but taking advantage of quantum teleportation to communicate between different quantum nodes was proposed. In this project, we develop the QMPI with point-to-point and collective operations in Qiskit and characterize its performance by demonstrating the application implementations. Moreover, we developed a new technique for optimizing collective communication of the distributed quantum programs with Multi-Controlled Toffoli gates. This technique beats the state-of-the-art in terms of fidelity and the number of remote EPR pairs consumed in both simulations and experiments.
The attention mechanism has become foundational for remarkable AI breakthroughs since the introduction of the Transformer, driving the demand for increasingly longer context to power frontier models such as large-scale reasoning language models and high-resolution image/video generators. However, its quadratic computational and memory complexities present substantial challenges. Current state-of-the-art parallel attention methods, such as ring attention, are widely adopted for long-context training but utilize a point-to-point communication strategy that fails to fully exploit the capabilities of modern HPC network architectures. In this work, we propose ringX, a scalable family of parallel attention methods optimized explicitly for HPC systems. By enhancing workload partitioning, refining communication patterns, and improving load balancing, ringX achieves up to 3.4 × speedup compared to conventional ring attention on the Frontier supercomputer. Optimized for both bi-directional and causal attention mechanisms, ringX demonstrates its effectiveness through training benchmarks of a Vision Transformer (ViT) on a climate dataset and a Generative Pre-Trained Transformer (GPT) model, Llama3 8B. Our method attains an end-to-end training speedup of approximately 1.5 × in both scenarios. To our knowledge, the achieved 38% model FLOPs utilization (MFU) for training Llama3 8B with a 1M-token sequence length on 4,096 GPUs represents one of the highest training efficiencies reported for long-context learning on HPC systems. Our code implementation is available at https://github.com/jqyin/ringX-attention.
Phloem is a Message Passing Interface (MPI) micro-benchmarking suite featuring sub-communicator collectives, methods for finding slow links on MPI interconnects, and point-to-point MPI benchmarks, including a messaging rate benchmark. All of the benchmarks except for ones related exclusively to finding slow links are GPU-aware via the Umpire resource management library.
The major achievements enabled by QMC Endstation grant include * Performance improvement on clusters of x86 multi-core systems, especially on Cray XT systems * New and improved methods for the wavefunction optimizations * New forms of trial wavefunctions * Implementation of the full application on NVIDIA GPUs using CUDA The scaling studies of QMCPACK on large-scale systems show excellent parallel efficiency up to 216K cores on Jaguarpf (Cray XT5). The GPU implementation shows speedups of 10-15x over the CPU implementation on older generation of x86. We have implemented hybrid OpenMP/MPI scheme in QMC to take advantage of multi-core shared memory processors of petascale systems. Our hybrid scheme has several advantages over the standard MPI-only scheme. * Memory optimized: large read-only data to store one-body orbitals and other shared properties to represent the trial wave function and many-body Hamiltonian can be shared among threads, which reduces the memory footprint of a large-scale problem. * Cache optimized: the data associated with an active Walker are in cache during the compute-intensive drift-diffusion process and the operations on an Walker are optimized for cache reuse. Thread-local objects are used to ensure the data affinity to a thread. * Load balanced: Walkers in an ensemble are evenly distributed among threads and MPI tasks. The two-level parallelism reduces the population imbalance among MPI tasks and reduces the number of point-to-point communications of large messages (serialized objects) for the Walker exchange. * Communication optimized: the communication overhead, especially for the collective operations necessary to determine ET and measure the properties of an ensemble, is significantly lowered by using less MPI tasks. The multiple forms of parallelism afforded by QMC algorithms make them ideal candidates for acceleration in the many-core paradigm. We presented the results of our effort to port the QMCPACK simulation code to the NVIDIA CUDA GPU platform. We restructured the CPU algorithms to express additional parallelism, minimize GPU-CPU communication, and efficiently utilize the GPU memory hierarchy. Using mixed precision on GT200 GPUs and MPI for intercommunication and load balancing, we observe typical full-application speedups of approximately 10x to 15x relative to quad-core Xeon CPUs alone, while reproducing the double-precision CPU results within statistical error. We developed an all-electron quantum Monte Carlo (QMC) method for solids that does not rely on pseudopotentials, and used it to construct a primary ultra-high-pressure calibration based on the equation of state of cubic boron nitride. We computed the static contribution to the free energy with the QMC method and obtained the phonon contribution from density functional theory, yielding a high-accuracy calibration up to 900 GPa usable directly in experiment. We computed the anharmonic Raman frequency shift with QMC simulations as a function of pressure and temperature, allowing optical pressure calibration. In contrast to present experimental approaches, small systematic errors in the theoretical EOS do not increase with pressure, and no extrapolation is needed. This all-electron method is applicable to first-row solids, providing a new reference for ab initio calculations of solids and benchmarks for pseudopotential accuracy. We compared experimental and theoretical results on the momentum distribution and the quasiparticle renormalization factor in sodium. From an x-ray Compton-profile measurement of the valence-electron momentum density, we derived its discontinuity at the Fermi wavevector finding an accurate measure of the renormalization factor that we compared with quantum-Monte-Carlo and G0W0 calculations performed both on crystalline sodium and on the homogeneous electron gas. Our calculated results are in good agreement with the experiment. We have been studying the heat of formation for various Kubas complexes of molecular hydrogen on Ti(1,2)ethylene-nH2 using Diffusion Monte Carlo. This work has been started and is ongoing. We are studying systems involving 1 and 2 Ti bonding sites with up to 10 hydrogen molecules in numerous configurations. This work will establish a benchmark that will test the accuracy of density functional calculations and establish the feasibility of our methods for similar systems.
LAMMPS "Large-scale Atomic/Molecular Massively Parallel Simulator" is an open-source molecular dynamics package from Sandia National Laboratories. Significant performance improvements in strong-scaling and time-to-solution for this application on IBM's Blue Gene/Q have been achieved through computational optimizations of the OpenMP versions of the short-range Lennard-Jones term of the CHARMM force field and the long-range Coulombic interaction implemented with the PPPM (particle-particle-particle mesh) algorithm, enhanced by runtime parameter settings controlling thread utilization. Additionally, MPI communication performance improvements were made to the PPPM calculation by re-engineering the parallel 3D FFT to use MPICH collectives instead of point-to-point. Performance testing was done using an 8.4-million atom simulation scaling up to 16 racks on the Mira system at Argonne Leadership Computing Facility (ALCF). Speedups resulting from this effort were in some cases over 2x.
In this work, we focus on compact schemes resulting in tridiagonal systems of equations, specifically the fifth-order CRWENO scheme. We propose a scalable implementation of the nonlinear compact schemes by implementing a parallel tridiagonal solver based on the partitioning/substructuring approach. We use an iterative solver for the reduced system of equations; however, we solve this system to machine zero accuracy to ensure that no parallelization errors are introduced. It is possible to achieve machine-zero convergence with few iterations because of the diagonal dominance of the system. The number of iterations is specified a priori instead of a norm-based exit criterion, and collective communications are avoided. The overall algorithm thus involves only point-to-point communication between neighboring processors. Our implementation of the tridiagonal solver differs from and avoids the drawbacks of past efforts in the following ways: it introduces no parallelization-related approximations (multiprocessor solutions are exactly identical to uniprocessor ones), it involves minimal communication, the mathematical complexity is similar to that of the Thomas algorithm on a single processor, and it does not require any communication and computation scheduling.
This project aimed to improve long term global climate simulations by resolving and enhancing the representation of the processes involved in the cycling of freshwater through estuaries and coastal regions. This was a collaborative multi-institution project consisting of physical oceanographers, climate model developers, and computational scientists. It specifically targeted the DOE objectives of advancing simulation and predictive capability of climate models through improvements in resolution and physical process representation. The main computational objectives were: 1. To develop computationally efficient, but physically based, parameterizations of estuary and continental shelf mixing processes for use in an Earth System Model (CESM). 2. To develop a two-way nested regional modeling framework in order to dynamically downscale the climate response of particular coastal ocean regions and to upscale the impact of the regional coastal processes to the global climate in an Earth System Model (CESM). 3. To develop computational infrastructure to enhance the efficiency of data transfer between specific sources and destinations, i.e., a point-to-point communication capability, (used in objective 1) within POP, the ocean component of CESM.
Although high-performance computing traditionally focuses on the efficient execution of large-scale applications, both energy and power have become critical concerns when approaching exascale. Drastic increases in the power consumption of supercomputers affect significantly their operating costs and failure rates. In modern microprocessor architectures, equipped with dynamic voltage and frequency scaling (DVFS) and CPU clock modulation (throttling), the power consumption may be controlled in software. Additionally, network interconnect, such as Infiniband, may be exploited to maximize energy savings while the application performance loss and frequency switching overheads must be carefully balanced. This work first studies two important collective communication operations, all-to-all and allgather and proposes energy saving strategies on the per-call basis. Next, it targets point-to-point communications to group them into phases and apply frequency scaling to them to save energy by exploiting the architectural and communication stalls. Finally, it proposes an automatic runtime system which combines both collective and point-to-point communications into phases, and applies throttling to them apart from DVFS to maximize energy savings. The experimental results are presented for NAS parallel benchmark problems as well as for the realistic parallel electronic structure calculations performed by the widely used quantum chemistry package GAMESS. Close to the maximum energy savings were obtained with a substantially low performance loss on the given platform.
This document is a lecture for the Parallel Computing Summer Research Internship at LANL. It describes the Message Passing Interface and its application.
This project investigated the cyber-security impacts of moving from an all analog, point-to-point, instrumentation and control (I&C) system to a digital I&C system based on Modbus and a shared communication medium. A formalism called a hybrid attack graph was expanded to support the nuclear research reactor system. The hybrid attack graph allows one to check a system for vulnerabilities, in this case cyber-security vulnerabilities, and to document the attack vectors (scenarios) causing those vulnerabilities. In parallel, a simulation of the system was developed to model both the physical reactor parameters and operations, as well as the network interconnects and communications. This simulation platform was modeled on the nuclear research reactor located at Washington State University. The simulation platform provided a sandbox to evaluate and quantify the impact of identified and proposed vulnerabilities in the system and to determine the effectiveness of countermeasures at stopping these attacks. The simulation and hybrid attack graph tools were integrated to provide a streamlined process of generating attack scenarios, playing those scenarios out in the simulation, and then analyzing the results to correlate system state to states in the hybrid attack graph. This process was used to (1) quantify the impact of attack scenarios and (2) to determine if the system moved through the hybrid attack graph as anticipated. The hybrid attack graph tool was extended and customized to produce a tool to automatically identify critical assets (CAs) and critical digital assets (CDAs) as defined by NRC Regulatory Guide 5.71. This tool was verified using the nuclear research reactor at Washington State University. Finally, a series of educational modules covering the findings of the different aspects of this research have been created.
Nearest-neighbor communication is at the heart of many high-performance parallel computations. We report on the performance of such communication on the Oak Ridge Leadership Computing Facility system Summit in the context of the PETSc communication module. The analysis in this report includes basic Ping-Pong point-to-point communication and regular and irregular nearest-neighbor communication.
High speed wireless communication has proven elusive in subsea environments due to the inherent bandwidth limitations of acoustics and range limitations of other transmission modalities. A truly connected subsea system necessitates a high-speed, resilient architecture that can enable the integration of new sensor technologies and edge analytics and allow closed-loop monitoring and control of subsea operations for integrity monitoring and optimization. Like terrestrial Internet of Things applications, the realization of this “digital subsea” vision requires the application of high speed, point-to-point wireless technologies to complement rather than replace “hard-wired” communications such as optical fiber or acoustic systems. This work addresses the development of ULTRA (Underwater LASER Telemetry and Remote Access), an ultra-long range underwater laser communications system for use in critical points of the subsea communications architecture to increase reliability, operational flexibility, and reduce communication system maintenance associated with physical subsea connections.
Wireless spectrum is a limiting resource for continued economic growth in the United States. As discussed in the recent National Spectrum Strategy (NSS), wireless spectrum underpins several aspects of the U.S. economy and the demand for additional spectrum is driving the need for realizing spectrum sharing to enable continued development. At the same time, wireless spectrum is an essential foundation of critical energy infrastructure, including electric, oil, and natural gas resources. In particular, wireless point-to-point (P2P) links are the backbone of vast infrastructure networks that enable the flow of sensor and control information needed to manage critical energy sector infrastructure in the United States. These links will only become more important as the energy sector incorporates more diverse sensing and more efficient control mechanisms, which increases system complexity and data volume requiring more resilient communications. Therefore, the continued economic development of the U.S. depends on determining novel approaches to spectrum management that balance both broad access for advanced wireless technologies and resilience for critical infrastructure.
recision timing distribution systems require determin istic and repeatable high speed serial data link latency. The EIC Timing Data Link will employ a specialized link layer protocol for deterministic multigigabit communication us ing AMD Ultrascale+ GTH and GTY transceivers. Link la tency must also be measurable to accurately compensate in real time for latency variations of the physical medium. The point-to-point link is full duplex and fully synchronous with a latency control algorithm to align the internal clocks of each system with picosecond resolution. Deterministic clock domain crossing between systems is ensured using static timing analysis of the elastic buffer control signals.
This study introduces a novel “Digital Grid Twin–Direct Communication Scheme” test bed. This advanced platform evaluates point-to-point communication between transmitter and receiver relays with optical fiber and radio omnidirectional antenna systems, implemented at the Advanced Protection lab in the Grid Research Innovation and Development Center at Oak Ridge National Laboratory. The increased diversity of energy sources has led to more protective relay misoperations. In North America, microgrid protection schemes now use point-to-point communication along distribution lines between relays to implement advanced logic in nonradial grids that include both high- and low-inertia generators. This trend challenges utilities to minimize misoperations while ensuring rapid fault clearance and accurate selectivity coordination between primary and backup relays. This study assesses relay-to-relay communication schemes by introducing an advanced testing platform based on a digital grid twin protection test bed using a synchronized time source system. The platform evaluates the communication system using radio antennas or optical fiber links by integrating protective relays that operate breakers within the digital twin and record relay events and communication signals. In the experiments, transmitter and receiver relays were configured with inverse time overcurrent and breaker trip detection logic to assess the total time of the communication protection schemes based on the sum of the relay protection element operating time, radio latency, propagation delay, baud rate delay, and relay processing time. These delays were derived from recorded relay events and communication signals from the interface of a real-time simulator set as a digital grid twin. The test bed successfully simulated various electrical faults along a distribution line while ensuring effective and reliable point-to-point communication between transmitter and receiver relays. The radio antenna communication system exhibited latency because of the radio. This latency depends on the baud rate setting and type of radio application; in general, the higher the baud rate, the lower the radio latency. The measured radio latency (for Mirrored Bits with an encryption card at 9,600 bps) was about 9–10 ms. Additionally, calculated propagation delay per mile for radio antennas and optical fiber was 5.36 µs/mi and 8.04 µs/mi, respectively. Optical fiber communication did not demonstrate radio latency. Instead, the protection element operating time depends mainly on the protection logic function set in the relay, and the relay processing time depends on the processing rate of the relay in samples per power system cycle.