Search NASA⌕ Search

SEARCH · Search NASA

Results for “Cache memory”

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.

At least 109 records · Page 6
All sourcesNASA NTRSDOE OSTI
DOE OSTI2026

FitCache: A Transparent Drop-In Framework for Multi-Tier Caching to Accelerate Distributed Deep Learning Workloads

Training in Deep learning (DL) remains highly compute- and data-intensive, with I/O becoming a critical bottleneck as models and datasets scale. Recent studies report that data loading can dominate training time, especially on large-scale HPC systems with shared parallel file systems (PFS). Existing caching approaches either rely on single-tier designs or require intrusive modifications to training pipelines, limiting their portability and effectiveness. In this work, we present FitCache, a transparent drop-in framework for multi-tier caching to accelerate distributed DL training by coordinating fast local memory (e.g., DRAM, Persistent Memory (PMem)) and NVMe as hierarchical caches atop PFS. Our design adapts to hardware diversity, i.e., if NVMe is missing, memory transparently acts as a caching tier, ensuring stable performance. FitCache transparently intercepts I/O requests and issues concurrent fetches across all tiers, returning data from the fastest responder without centralized metadata or static redirection paths. FitCache adapts to dynamic workloads and heterogeneous clusters while maintaining POSIX compatibility. Experiments on Frontier (2048 GPUs) and smaller research clusters show that FitCache reduces training time by up to 40% and per-batch I/O latency by up to 71.6% compared to Lustre Orion PFS, offering a drop-in solution for scalable DL training.

Hu, Guangxing [ORNL] (ORCID:0009000283203614)↗
DOE OSTI2022

Performance assessment of ensembles of in situ workflows under resource constraints

Summary Scientific breakthroughs in biomolecular methods and improvements in hardware technology have shifted from a long‐running simulation to a large set of shorter simulations running simultaneously, called an ensemble. In an ensemble, simulations are usually coupled with analyses of data produced by the simulations. In situ methods can be used to analyze large volumes of data generated by scientific simulations at runtime (i.e., simulations and analyses are performed concurrently). In this work, we study the execution of ensemble‐based simulations paired with in situ analyses using in‐memory staging methods. Using an ensemble of molecular dynamics in situ workflows with multiple simulations and analyses, we first show that collecting traditional metrics such as makespan, instructions per cycle, memory usage, or cache miss ratio is not sufficient to characterize complex behaviors of ensembles. We propose a method to evaluate the performance of ensembles of workflows that captures multiple resource usage aspects: resource efficiency, resource allocation, and resource provisioning. Experimental results demonstrate that the proposed method can effectively distinguish the performance of different component placements in an ensemble with up to 32 ensemble members. By evaluating different co‐location scenarios, our proposed performance indicators demonstrate benefits of co‐locating simulation and coupled analyses within a compute node.

Do, Tu Mai Anh↗
DOE OSTI2017

Managing the Performance/Error Tradeoff of Floating-point Intensive Applications

Modern embedded systems are becoming more reliant on real-valued arithmetic as they employ mathematically complex vision algorithms and sensor signal processing. Double-precision floating point is the most commonly used precision in computer vision algorithm implementations. A single-precision floating point can provide a performance boost due to less memory transfers, less cache occupancy, and relatively faster mathematical operations on some architectures. However, adopting it can result in loss of accuracy. Identifying which parts of the program can run in single-precision floating point with low impact on error is a manual and tedious process. In this paper, we propose an automatic approach to identify parts of the program that have a low impact on error using shadow-value analysis. Our approach provides the user with a performance/error tradeoff, using which the user can decide how much accuracy can be sacrificed in return for performance improvement. We illustrate the impact of the approach using a well known implementation of Apriltag detection used in robotics vision. In conclusion, we demonstrate that an average 1.3x speedup can be achieved with no impact on tag detection, and a 1.7x speedup with only 4% false negatives.

97 MATHEMATICS AND COMPUTING↗
DOE OSTI2020

Parallelization and Performance Portability in Hydrodynamics Codes

With the eve of Exascale computing, performance and portability are at the forefront of all scientific codes. Adding more cores and more energy to a system is no longer a sustainable way to achieve performance, and extra effort must now be made to improve performance in all areas of code and code development. Using hydrodynamic codes as a basis, this work explores numerous techniques to achieve performance in different ways. Adaptive mesh refinement (AMR) is a necessary technique to improve memory optimization in mesh-based simulations. However it is invasive and conventionally difficult to integrate into existing applications, so we present a new branch of AMR to create a smooth transition to these optimizations, which not only improves performance, but also greatly reduces developer effort. We introduce the concept of this improvement as Phantom-Cell AMR, and assess theoretically the improvements, as well as present an application of its use. Other work included involves and investigation into an efficient data structure that ensures optimal memory layout for cache performance, with a target of making codes performant and portable across all architectures. All of the work targets both performance and portability, not just on CPU hardware, but specifically across GPU architectures. Parallel performance is key to all of the methods presented, but the research makes a great effort to improve the portability of all applications to prepare for current high performance computing systems and those on the horizon.

97 MATHEMATICS AND COMPUTING↗
NASA NTRS2003

Scalability of a Low-Cost Multi-Teraflop Linux Cluster for High-End Classical Atomistic and Quantum Mechanical Simulations

Scalability of a low-cost, Intel Xeon-based, multi-Teraflop Linux cluster is tested for two high-end scientific applications: Classical atomistic simulation based on the molecular dynamics method and quantum mechanical calculation based on the density functional theory. These scalable parallel applications use space-time multiresolution algorithms and feature computational-space decomposition, wavelet-based adaptive load balancing, and spacefilling-curve-based data compression for scalable I/O. Comparative performance tests are performed on a 1,024-processor Linux cluster and a conventional higher-end parallel supercomputer, 1,184-processor IBM SP4. The results show that the performance of the Linux cluster is comparable to that of the SP4. We also study various effects, such as the sharing of memory and L2 cache among processors, on the performance.

Kikuchi, Hideaki↗
NASA NTRS2008

The Interplanetary Overlay Networking Protocol Accelerator

A document describes the Interplanetary Overlay Networking Protocol Accelerator (IONAC) an electronic apparatus, now under development, for relaying data at high rates in spacecraft and interplanetary radio-communication systems utilizing a delay-tolerant networking protocol. The protocol includes provisions for transmission and reception of data in bundles (essentially, messages), transfer of custody of a bundle to a recipient relay station at each step of a relay, and return receipts. Because of limitations on energy resources available for such relays, data rates attainable in a conventional software implementation of the protocol are lower than those needed, at any given reasonable energy-consumption rate. Therefore, a main goal in developing the IONAC is to reduce the energy consumption by an order of magnitude and the data-throughput capability by two orders of magnitude. The IONAC prototype is a field-programmable gate array that serves as a reconfigurable hybrid (hardware/ firmware) system for implementation of the protocol. The prototype can decode 108,000 bundles per second and encode 100,000 bundles per second. It includes a bundle-cache static randomaccess memory that enables maintenance of a throughput of 2.7Gb/s, and an Ethernet convergence layer that supports a duplex throughput of 1Gb/s.

Pang, Jackson↗
DOE OSTI2017

Fusion PIC code performance analysis on the Cori KNL system

We study the attainable performance of Particle-In-Cell codes on the Cori KNL system by analyzing a miniature particle push application based on the fusion PIC code XGC1. We start from the most basic building blocks of a PIC code and build up the complexity to identify the kernels that cost the most in performance and focus optimization efforts there. Particle push kernels operate at high AI and are not likely to be memory bandwidth or even cache bandwidth bound on KNL. Therefore, we see only minor benefits from the high bandwidth memory available on KNL, and achieving good vectorization is shown to be the most beneficial optimization path with theoretical yield of up to 8x speedup on KNL. In practice we are able to obtain up to a 4x gain from vectorization due to limitations set by the data layout and memory latency.

97 MATHEMATICS AND COMPUTING↗
DOE OSTI2016

Calculating Reuse Distance from Source Code

The efficient use of a system is of paramount importance in high-performance computing. Applications need to be engineered for future systems even before the architecture of such a system is clearly known. Static performance analysis that generates performance bounds is one way to approach the task of understanding application behavior. Performance bounds provide an upper limit on the performance of an application on a given architecture. Predicting cache hierarchy behavior and accesses to main memory is a requirement for accurate performance bounds. This work presents our static reuse distance algorithm to generate reuse distance histograms. We then use these histograms to predict cache miss rates. Experimental results for kernels studied show that the approach is accurate.

Cache memory↗
DOE OSTI2022

Design and analysis of CXL performance models for tightly-coupled heterogeneous computing

Truly heterogeneous systems enable partitioned workloads to be mapped to the hardware that nets the best performance. However, current practice requires that inter-device communication between different vendors' hardware use host memory as an intermediary step. To date, there are no widely adopted solutions that allow accelerators to directly transfer data. A new cache-coherent protocol, CXL, aims to facilitate easier, fine-grained sharing between accelerators. In this work we analyze existing methods for designing heterogeneous applications that target GPUs and FPGAs working collaboratively, followed by an exploration to show the benefits of a CXL-enabled system. Specifically, we develop a test application that utilizes both an NVIDIA P100 GPU and a Xilinx U250 FPGA to show current communication limitations. From this application, we capture overall execution time and throughput measurements on the FPGA and GPU. We use these measurements as inputs to novel CXL performance models to show that using CXL caching instead of host memory results in a 1.31X speedup, while a more tightly-coupled pipelined implementation using CXL-enabled hardware would result in a speedup of 1.45X.

Cabrera, Anthony↗
DOE OSTI2021

Method and apparatus for using compression to improve performance of low voltage caches

A method of operating a cache in a computing device includes, in response to receiving a memory access request at the cache, determining compressibility of data specified by the request, selecting in the cache a destination portion for storing the data based on the compressibility of the data and a persistent fault history of the destination portion, and storing a compressed copy of the data in a non-faulted subportion of the destination portion, wherein the persistent fault history indicates that the non-faulted subportion excludes any persistent faults.

Kalamatianos, John↗
NASA NTRS1998

Performance Modeling and Measurement of Parallelized Code for Distributed Shared Memory Multiprocessors

This paper presents a model to evaluate the performance and overhead of parallelizing sequential code using compiler directives for multiprocessing on distributed shared memory (DSM) systems. With increasing popularity of shared address space architectures, it is essential to understand their performance impact on programs that benefit from shared memory multiprocessing. We present a simple model to characterize the performance of programs that are parallelized using compiler directives for shared memory multiprocessing. We parallelized the sequential implementation of NAS benchmarks using native Fortran77 compiler directives for an Origin2000, which is a DSM system based on a cache-coherent Non Uniform Memory Access (ccNUMA) architecture. We report measurement based performance of these parallelized benchmarks from four perspectives: efficacy of parallelization process; scalability; parallelization overhead; and comparison with hand-parallelized and -optimized version of the same benchmarks. Our results indicate that sequential programs can conveniently be parallelized for DSM systems using compiler directives but realizing performance gains as predicted by the performance model depends primarily on minimizing architecture-specific data locality overhead.

Waheed, Abdul↗
NASA NTRS2005

Parallelization of Lower-Upper Symmetric Gauss-Seidel Method for Chemically Reacting Flow

Development of technologies for exploration of the solar system has revived an interest in computational simulation of chemically reacting flows since planetary probe vehicles exhibit non-equilibrium phenomena during the atmospheric entry of a planet or a moon as well as the reentry to the Earth. Stability in combustion is essential for new propulsion systems. Numerical solution of real-gas flows often increases computational work by an order-of-magnitude compared to perfect gas flow partly because of the increased complexity of equations to solve. Recently, as part of Project Columbia, NASA has integrated a cluster of interconnected SGI Altix systems to provide a ten-fold increase in current supercomputing capacity that includes an SGI Origin system. Both the new and existing machines are based on cache coherent non-uniform memory access architecture. Lower-Upper Symmetric Gauss-Seidel (LU-SGS) relaxation method has been implemented into both perfect and real gas flow codes including Real-Gas Aerodynamic Simulator (RGAS). However, the vectorized RGAS code runs inefficiently on cache-based shared-memory machines such as SGI system. Parallelization of a Gauss-Seidel method is nontrivial due to its sequential nature. The LU-SGS method has been vectorized on an oblique plane in INS3D-LU code that has been one of the base codes for NAS Parallel benchmarks. The oblique plane has been called a hyperplane by computer scientists. It is straightforward to parallelize a Gauss-Seidel method by partitioning the hyperplanes once they are formed. Another way of parallelization is to schedule processors like a pipeline using software. Both hyperplane and pipeline methods have been implemented using openMP directives. The present paper reports the performance of the parallelized RGAS code on SGI Origin and Altix systems.

Yoon, Seokkwan↗
DOE OSTI2019

Selective data retrieval based on access latency

A processor includes multiple processing units (e.g., processor cores), with each processing unit associated with at least one private, dedicated cache. The processor is also associated with a system memory that stores all data that can be accessed by the multiple processing units. A coherency manager (e.g., a coherence directory) of the processor enforces a specified coherency scheme to ensure data coherency between the different caches and between the caches and the system memory. In response to a memory access request to a given cache resulting in a cache miss, the coherency manager identifies the current access latency to the system memory as well as the current access latencies to other caches of the processor. The coherency manager transfers the targeted data to the given cache from the cache or system memory having the lower access latency.

97 MATHEMATICS AND COMPUTING↗
DOE OSTI2019

Analytical Modeling of Cache Behavior for Affine Programs

Modern optimizing compiler design integrates program transformation strategies aimed at reducing data movement to/from main memory by exploiting the data cache hierarchy. But in reality, the effect of these transformations on the actual cache miss count is ignored in compilers due to the lack of precise compile-time models of misses for hierarchical caches. The current state of practice for cache miss analysis is based on accurate simulation. However, simulation requires time proportional to the dataset/problem size as well as the number of distinct cache configurations of interest to be evaluated. This paper takes a fundamentally different approach, made possible by focusing on polyhedral programs with static control flow: instead of relying on costly simulation, this paper develops the first closed-form solution for exact modeling of misses in a set-associative cache hierarchy. It enabling simulation-less selection of program transformations at compile-time to optimize cache misses. A push-button tool implementing the approach is developed and used for extensive validation of the complete framework.

affine programs, cache modeling, set associativity↗
NASA NTRS2013

A Simple GPU-Accelerated Two-Dimensional MUSCL-Hancock Solver for Ideal Magnetohydrodynamics

We describe our experience using NVIDIA's CUDA (Compute Unified Device Architecture) C programming environment to implement a two-dimensional second-order MUSCL-Hancock ideal magnetohydrodynamics (MHD) solver on a GTX 480 Graphics Processing Unit (GPU). Taking a simple approach in which the MHD variables are stored exclusively in the global memory of the GTX 480 and accessed in a cache-friendly manner (without further optimizing memory access by, for example, staging data in the GPU's faster shared memory), we achieved a maximum speed-up of approx. = 126 for a sq 1024 grid relative to the sequential C code running on a single Intel Nehalem (2.8 GHz) core. This speedup is consistent with simple estimates based on the known floating point performance, memory throughput and parallel processing capacity of the GTX 480.

graphics processing units↗
DOE OSTI2017

Two-level main memory co-design: Multi-threaded algorithmic primitives, analysis, and simulation

A challenge in computer architecture is that processors often cannot be fed data from DRAM as fast as CPUs can consume it. Therefore, many applications are memory-bandwidth bound. With this motivation and the realization that traditional architectures (with all DRAM reachable only via bus) are insufficient to feed groups of modern processing units, vendors have introduced a variety of non-DDR 3D memory technologies (Hybrid Memory Cube (HMC),Wide I/O 2, High Bandwidth Memory (HBM)). These offer higher bandwidth and lower power by stacking DRAM chips on the processor or nearby on a silicon interposer. We will call these solutions “near-memory,” and if user-addressable, “scratchpad.” High-performance systems on the market now offer two levels of main memory: near-memory on package and traditional DRAM further away. In the near term we expect the latencies near-memory and DRAM to be similar. Here, it is natural to think of near-memory as another module on the DRAM level of the memory hierarchy. Vendors are expected to offer modes in which the near memory is used as cache, but we believe that this will be inefficient.

97 MATHEMATICS AND COMPUTING↗
DOE OSTI2022

GPU cache management based on locality type detection

Wavefront loading in a processor is managed and includes monitoring a selected wavefront of a set of wavefronts. Reuse of memory access requests for the selected wavefront is counted. A cache hit rate in one or more caches of the processor is determined based on the counted reuse. Based on the cache hit rate, subsequent memory requests of other wavefronts of the set of wavefronts are modified by including a type of reuse of cache lines in requests to the caches. In the caches, storage of data in the caches is based on the type of reuse indicated by the subsequent memory access requests. Reused cache lines are protected by preventing cache line contents from being replaced by another cache line for a duration of processing the set of wavefronts. Caches are bypassed when streaming access requests are made.

Zhang, Xianwei↗