Search NASA⌕ Search

SEARCH · Search NASA

Results for “performance autotuning”

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.

47 records · Page 3

Modeling Cooperative Threads to Project GPU Performance for Adaptive Parallelism

Most accelerators, such as graphics processing units (GPUs) and vector processors, are particularly suitable for accelerating massively parallel workloads. On the other hand, conventional workloads are developed for multi-core parallelism, which often scale to only a few dozen OpenMP threads. When hardware threads significantly outnumber the degree of parallelism in the outer loop, programmers are challenged with efficient hardware utilization. A common solution is to further exploit the parallelism hidden deep in the code structure. Such parallelism is less structured: parallel and sequential loops may be imperfectly nested within each other, neigh boring inner loops may exhibit different concurrency patterns (e.g. Reduction vs. Forall), yet have to be parallelized in the same parallel section. Many input-dependent transformations have to be explored. A programmer often employs a larger group of hardware threads to cooperatively walk through a smaller outer loop partition and adaptively exploit any encountered parallelism. This process is time-consuming and error-prone, yet the risk of gaining little or no performance remains high for such workloads. To reduce risk and guide implementation, we propose a technique to model workloads with limited parallelism that can automatically explore and evaluate transformations involving cooperative threads. Eventually, our framework projects the best achievable performance and the most promising transformations without implementing GPU code or using physical hardware. We envision our technique to be integrated into future compilers or optimization frameworks for autotuning.

Meng, Jiayuan↗

Performance Engineering Research Institute SciDAC-2 Enabling Technologies Institute Final Report

Enhancing the performance of SciDAC applications on petascale systems has high priority within DOE SC. As we look to the future, achieving expected levels of performance on high-end com-puting (HEC) systems is growing ever more challenging due to enormous scale, increasing archi-tectural complexity, and increasing application complexity. To address these challenges, PERI has implemented a unified, tripartite research plan encompassing: (1) performance modeling and prediction; (2) automatic performance tuning; and (3) performance engineering of high profile applications. The PERI performance modeling and prediction activity is developing and refining performance models, significantly reducing the cost of collecting the data upon which the models are based, and increasing model fidelity, speed and generality. Our primary research activity is automatic tuning (autotuning) of scientific software. This activity is spurred by the strong user preference for automatic tools and is based on previous successful activities such as ATLAS, which has automatically tuned components of the LAPACK linear algebra library, and other re-cent work on autotuning domain-specific libraries. Our third major component is application en-gagement, to which we are devoting approximately 30% of our effort to work directly with Sci-DAC-2 applications. This last activity not only helps DOE scientists meet their near-term per-formance goals, but also helps keep PERI research focused on the real challenges facing DOE computational scientists as they enter the Petascale Era.

97 MATHEMATICS AND COMPUTING↗

LDRD final report : autotuning for scalable linear algebra.

This report summarizes the progress made as part of a one year lab-directed research and development (LDRD) project to fund the research efforts of Bryan Marker at the University of Texas at Austin. The goal of the project was to develop new techniques for automatically tuning the performance of dense linear algebra kernels. These kernels often represent the majority of computational time in an application. The primary outcome from this work is a demonstration of the value of model driven engineering as an approach to accurately predict and study performance trade-offs for dense linear algebra computations.

99 GENERAL AND MISCELLANEOUS↗

A MultiGPU Performance-Portable Solution for Array Programming Based on Kokkos

Today, multiGPU nodes are widely used in high-performance computing and data centers. However, current programming models do not provide simple, transparent, and portable support for automatically targeting multiple GPUs within a node on application areas of array programming. In this paper, we describe a new application programming interface based on the Kokkos programming model to enable array computation on multiple GPUs in a transparent and portable way across both NVIDIA and AMD GPUs. We implement different variations of this technique to accommodate the exchange of stencils (array boundaries) among different GPU memory spaces, and we provide autotuning to select the proper number of GPUs, depending on the computational cost of the operations to be computed on arrays, that is completely transparent to the programmer. We evaluate our multiGPU extension on Summit (#5 TOP500), with six NVIDIA V100 Volta GPUs per node, and Crusher that contains identical hardware/software as Frontier (#1 TOP500), with four AMD MI250X GPUs, each with 2 Graphics Compute Dies (GCDs)for a total of 8 GCDs per node. We also compare the performance of this solution against the use of MPI + Kokkos, which is the cur-rent de facto solution for multiple GPUs in Kokkos. Our evaluation shows that the new Kokkos solution provides good scalability for many GPUs and a faster and simpler solution (from a programming productivity perspective) than MPI + Kokkos.

Valero Lara, Pedro↗

Compiler generation and autotuning of communication-avoiding operators for geometric multigrid

This paper describes a compiler approach to introducing communication-avoiding optimizations in geometric multigrid (GMG), one of the most popular methods for solving partial differential equations. Communication-avoiding optimizations reduce vertical communication through the memory hierarchy and horizontal communication across processes or threads, usually at the expense of introducing redundant computation. We focus on applying these optimizations to the smooth operator, which successively reduces the error and accounts for the largest fraction of the GMG execution time. Our compiler technology applies both novel and known transformations to derive an implementation comparable to manually-tuned code. To make the approach portable, an underlying autotuning system explores the tradeoff between reduced communication and increased computation, as well as tradeoffs in threading schemes, to automatically identify the best implementation for a particular architecture and at each computation phase. Results show that we are able to quadruple the performance of the smooth operation on the finest grids while attaining performance within 94% of manually-tuned code. Overall we improve the overall multigrid solve time by 2.5× without sacrificing programer productivity.

97 MATHEMATICS AND COMPUTING↗

Calibrating Building Energy Models Using Supercomputer Trained Machine Learning Agents

Building Energy Modeling (BEM) is an approach to model the energy usage in buildings for design and retrofit purposes. EnergyPlus is the flagship Department of Energy software that performs BEM for different types of buildings. The input to EnergyPlus can often extend in the order of a few thousand parameters which have to be calibrated manually by an expert for realistic energy modeling. This makes it challenging and expensive thereby making building energy modeling unfeasible for smaller projects. In this paper, we describe the Autotune research which employs machine learning algorithms to generate agents for the different kinds of standard reference buildings in the U.S. building stock. The parametric space and the variety of building locations and types make this a challenging computational problem necessitating the use of supercomputers. Millions of EnergyPlus simulations are run on supercomputers which are subsequently used to train machine learning algorithms to generate agents. These agents, once created, can then run in a fraction of the time thereby allowing cost-effective calibration of building models.

Sanyal, Jibonananda↗

Supercomputer assisted generation of machine learning agents for the calibration of building energy models. In: XSEDE '13 Proceedings of the Conference on Extreme Science and Engineering Discovery Environment: Gateway to Discovery, Article No. 22

Building Energy Modeling (BEM) is an approach to model the energy usage in buildings for design and retrofit purposes. EnergyPlus is the flagship Department of Energy software that performs BEM for different types of buildings. The input to EnergyPlus can often extend in the order of a few thousand parameters which have to be calibrated manually by an expert for realistic energy modeling. This makes it challenging and expensive thereby making building energy modeling unfeasible for smaller projects. In this paper, we describe the "Autotune" research which employs machine learning algorithms to generate agents for the different kinds of standard reference buildings in the U.S. building stock. The parametric space and the variety of building locations and types make this a challenging computational problem necessitating the use of supercomputers. Millions of EnergyPlus simulations are run on supercomputers which are subsequently used to train machine learning algorithms to generate agents. These agents, once created, can then run in a fraction of the time thereby allowing cost-effective calibration of building models.

Sanyal, Jibonananda↗

Supercomputer Assisted Generation of Machine Learning Agents for the Calibration of Building Energy Models

Building Energy Modeling (BEM) is an approach to model the energy usage in buildings for design and retrot pur- poses. EnergyPlus is the agship Department of Energy software that performs BEM for dierent types of buildings. The input to EnergyPlus can often extend in the order of a few thousand parameters which have to be calibrated manu- ally by an expert for realistic energy modeling. This makes it challenging and expensive thereby making building en- ergy modeling unfeasible for smaller projects. In this paper, we describe the \Autotune" research which employs machine learning algorithms to generate agents for the dierent kinds of standard reference buildings in the U.S. building stock. The parametric space and the variety of building locations and types make this a challenging computational problem necessitating the use of supercomputers. Millions of En- ergyPlus simulations are run on supercomputers which are subsequently used to train machine learning algorithms to generate agents. These agents, once created, can then run in a fraction of the time thereby allowing cost-eective cali- bration of building models.

autotune↗

Toward Robust Autotuning of Noisy Quantum dot Devices

The current autotuning approaches for quantum dot (QD) devices, while showing some success, lack an assessment of data reliability. This leads to unexpected failures when noisy or otherwise low-quality data is processed by an autonomous system. In this work, we propose a framework for robust autotuning of QD devices that combines a machine learning (ML) state classifier with a data quality control module. The data quality control module acts as a “gatekeeper” system, ensuring that only reliable data are processed by the state classifier. Lower data quality results in either device recalibration or termination. To train both ML systems, we enhance the QD simulation by incorporating synthetic noise typical of QD experiments. We confirm that the inclusion of synthetic noise in the training of the state classifier significantly improves the performance, resulting in an accuracy of 95.0(9)% when tested on experimental data. We then validate the functionality of the data quality control module by showing that the state classifier performance deteriorates with decreasing data quality, as expected. Our results establish a robust and flexible ML framework for autonomous tuning of noisy QD devices.

75 CONDENSED MATTER PHYSICS, SUPERCONDUCTIVITY AND↗

Autotuning of Double-Dot Devices In Situ with Machine Learning

The current practice of manually tuning quantum dots (QDs) for qubit operation is a relatively time-consuming procedure that is inherently impractical for scaling up and applications. In this work, we report on the in situ implementation of a recently proposed autotuning protocol that combines machine learning (ML) with an optimization routine to navigate the parameter space. In particular, we show that a ML algorithm trained using exclusively simulated data to quantitatively classify the state of a double-QD device can be used to replace human heuristics in the tuning of gate voltages in real devices. Here, we demonstrate active feedback of a functional double-dot device operated at millikelvin temperatures and discuss success rates as a function of the initial conditions and the device performance. Modifications to the training network, fitness function, and optimizer are discussed as a path toward further improvement in the success rate when starting both near and far detuned from the target double-dot range.

71 CLASSICAL AND QUANTUM MECHANICS, GENERAL PHYSIC↗

TUNE: Compiler-Directed Automatic Performance Tuning

This project has developed compiler-directed performance tuning technology targeting the Cray XT4 Jaguar system at Oak Ridge, which has multi-core Opteron nodes with SSE-3 SIMD extensions, and the Cray XE6 Hopper system at NERSC. To achieve this goal, we combined compiler technology for model-guided empirical optimization for memory hierarchies with SIMD code generation, which have been developed by the PIs over the past several years. We examined DOE Office of Science applications to identify performance bottlenecks and apply our system to computational kernels that operate on dense arrays. Our goal for this performance-tuning technology has been to yield hand-tuned levels of performance on DOE Office of Science computational kernels, while allowing application programmers to specify their computations at a high level without requiring manual optimization. Overall, we aim to make our technology for SIMD code generation and memory hierarchy optimization a crucial component of high-productivity Petaflops computing through a close collaboration with the scientists in national laboratories.

97 MATHEMATICS AND COMPUTING↗