SEARCH · Engineering Papers
Results for “Performance optimization”
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.
Enhanced thermoelectric performance of optimized Ba, Yb filled and Fe substituted skutterudite compounds
Explore the source record for details and available documents.
Evaluating adaptive and predictive power management strategies for optimizing visualization performance on supercomputers
Power is becoming an increasingly scarce resource on the next generation of supercomputers, and should be used wisely to improve overall performance. One strategy for improving power usage is hardware overprovisioning, i.e., systems with more nodes than can be run at full power simultaneously without exceeding the system-wide power limit. With this study, we compare two strategies for allocating power throughout an overprovisioned system – adaptation and prediction – in the context of visualization workloads. While adaptation has been suitable for workloads with more regular execution behaviors, it may not be as suitable on visualization workloads, since they can have variable execution behaviors. This study considers a total of 104 experiments, which vary the rendering workload, power budget, allocation strategy, and node concurrency, including tests processing data sets up to 1 billion cells and using up to 18,432 cores across 512 nodes. Overall, we find that prediction is a superior strategy for this use case, improving performance up to 27% compared to an adaptive strategy.
Etching-Chemistry-Driven Ruthenium Doping on Ti 3 C 2 T x MXene for Optimizing Electrochemical Performance
We demonstrate that the etching chemistry used during MXene synthesis from Ti 3 AlC 2 MAX phase significantly influences surface functionalization and structural vacancies, which in turn affect ruthenium (Ru) ion interactions. Using hydrofluoric acid (HF) and ammonium bifluoride (NH 4 HF 2 ) as etchants, we obtained MXene surfaces with distinct functional groups and Ti vacancies that impact Ru ion interactions and electrochemical performance. Both MXene variants (labeled MX(H) and MX(N), respectively) exhibited negative zeta potentials in their pristine state, but upon the addition of Ru the zeta potential for MX(H) reached 12.9 mV while that for MX(N) remained negative at −6.4 mV. This adsorption resulted in a 14.4-fold increase in the specific capacitance of MX(H)/Ru compared to pristine MX(H), whereas MX(N)/Ru exhibited only a 4.4-fold increase over its pristine counterpart. X-ray diffraction analysis identified the formation of ammonium titanium oxide fluoride, (NH 4 ) 3 TiOF 5 , on MX(N), which likely contributed to its reduced Ru adsorption. X-ray photoelectron spectroscopy suggested the presence of Ti vacancies in both MXene variants; however, their behavior toward Ru accommodation differed markedly, with MX(H) showing the most obvious shift in the Ti 2p peak in the XPS survey spectrum, while MX(N) showed the most obvious shift in the C 1s peak. Electron paramagnetic resonance spectroscopy further demonstrated a distinct alteration in the spectral signatures of MX(H) upon Ru addition, in contrast to the negligible changes in MX(N), indicating effective passivation of the Ti defect sites in MX(H) via vacancy-assisted Ru doping. Cyclic voltammetry showed that Ru-incorporated MX(H) nanocomposites exhibit more efficient redox-active sites, as reflected in their higher capacitance values. These findings highlight the pivotal role of MXene surface chemistry in controlling cation adsorption, providing valuable insights for the rational design of high-performance electrodes.
Optimizing photovoltaic performance in CuInS 2 and CdS quantum dot-sensitized solar cells by using an agar-based gel polymer electrolyte
Quantum dot-sensitized solar cells (QDSSCs) offer new opportunities to address the clean energy challenge, being one of the top candidates for third generation photovoltaics. Like dye-sensitized solar cells (DSSCs), QDSSCs normally use liquid electrolytes that suffer from issues such as evaporation or leakage. In this study a gel polysulfide electrolyte was prepared containing a natural polymer, agar, and was used as a quasi-solid-state electrolyte in solar cells to replace the conventional liquid electrolytes. This gel electrolyte shows almost the same conductivity as the liquid one. The solar cells were fabricated using CuInS 2 quantum dots (QDs), previously synthesized, deposited on TiO 2 photoanodes by electrophoretic deposition (EPD). CdS was deposited on TiO 2 by successive ionic layer adsorption and reaction (SILAR). Reduced graphene oxide (RGO)–Cu 2 S, brass, and thin film CuxS were used as counter electrodes. Compared to a liquid polysulfide water based electrolyte, solar cells based on CuInS 2 and CdS using gel polymer electrolyte (GPE) exhibit greater incident photon to current conversion efficiency (IPCE = 51.7% at 520 nm and 72.7% at 440 nm), photocurrent density (J sc = 10.75 and 13.51 mA cm -2 ), and power conversion efficiency (η = 2.97 and 2.98%) while exhibiting significantly enhanced stability. The solar cells employing the agar-based gel polymeric electrolyte are about a factor of 0.20 more stable than using a liquid electrolyte. The higher photovoltaic performance is due to the good conductivity and high wettability as well as the superior permeation capability of the gel electrolyte into the mesoporous matrix of a TiO 2 film
Criticality analysis of nuclear binding energy neural networks
Machine learning methods, in particular deep learning methods such as artificial neural networks (ANNs) with many layers, have become widespread and useful tools in nuclear physics. However, these ANNs are typically treated as ‘black boxes’, with their architecture (width, depth, and weight/bias initialization) and the training algorithm and parameters chosen empirically by optimizing learning based on limited exploration. We test a non-empirical approach to understanding and optimizing nuclear physics ANNs by adapting a criticality analysis based on renormalization group flows in terms of the hyperparameters for weight/bias initialization, training rates, and the ratio of depth to width. This treatment utilizes the statistical properties of neural network initialization to find a generating functional for network outputs at any layer, allowing for a path integral formulation of the ANN outputs as a Euclidean statistical field theory. We use a prototypical example to test the applicability of this approach: a simple ANN for nuclear binding energies. We find that with training using a stochastic gradient descent optimizer, the predicted criticality behavior is realized, and optimal performance is found with critical tuning. However, the use of an adaptive learning algorithm leads to somewhat superior results without concern for tuning and thus obscures the analysis. Nevertheless, the criticality analysis offers a way to look within the black box of ANNs, which is a first step towards potential improvements in network performance beyond using adaptive optimizers.
CERN’s Super Proton Synchrotron 200 MHz cavity regulation upgrade: Modeling, design optimization, and performance estimation
Not Available
Quantum Approximate Optimization Algorithm: Performance, Mechanism, and Implementation on Near-Term Devices
Not Available
Optimizing the performance of parallel applications on a 5D torus via task mapping
Abstract not provided
Impacts of Mathematical Optimizations on Reinforcement Learning Policy Performance.
Abstract not provided.
Piper: Pipelining OpenMP Offloading Execution Through Compiler Optimization For Performance
Explore the source record for details and available documents.
ACStor: Optimizing Access Performance of Virtual Disk Images in Clouds
In virtualized data centers, virtual disk images (VDIs) serve as the containers in virtual environment, so their access performance is critical for the overall system performance. Some distributed VDI chunk storage systems have been proposed in order to alleviate the I/O bottleneck for VM management. As the system scales up to a large number of running VMs, however, the overall network traffic would become unbalanced with hot spots on some VMs inevitably, leading to I/O performance degradation when accessing the VMs. Here, we propose an adaptive and collaborative VDI storage system (ACStor) to resolve the above performance issue. In comparison with the existing research, our solution is able to dynamically balance the traffic workloads in accessing VDI chunks, based on the run-time network state. Specifically, compute nodes with lightly loaded traffic will be adaptively assigned more chunk access requests from remote VMs and vice versa, which can effectively eliminate the above problem and thus improves the I/O performance of VMs. We also implement a prototype based on our ACStor design, and evaluate it by various benchmarks on a real cluster with 32 nodes and a simulated platform with 256 nodes. Experiments show that under different network traffic patterns of data centers, our solution achieves up to 2-8 performance gain on VM booting time and VM’s I/O throughput, in comparison with the other state-of-the-art approaches.
Enhanced photocathode performance through optimization of film thickness and substrate
Here, it is shown that the efficiency of photoelectron emission may be enhanced, several-fold, through optimization of photocathode film thickness and appropriate substrate configuration. Such an enhancement is based on a careful consideration of wave interference effects in the film and the consequent modulation of the absorption profiles and electron emission probabilities. Finally, the inadequacy of the well-known Lambert-Beer law for modeling photon absorption in thin films is also discussed.
Adaptive Time Step Control for Multirate Infinitesimal Methods
Multirate methods have been used for decades to temporally evolve initial-value problems in which different components evolve on distinct time scales, and thus use of different step sizes for these components can result in increased computational efficiency. Generally, such methods select these different step sizes based on experimentation or stability considerations. For problems that evolve on a single time scale, adaptivity approaches that strive to control local temporal error are widely used to achieve numerical results of a desired accuracy with minimal computational effort, while alleviating the need for manual experimentation with different time step sizes. However, there is a notable gap in the publication record on the development of adaptive time step controllers for multirate methods. In this paper, we extend the single-rate controller work of Gustafsson [ACM Trans. Math. Software, 20 (1994), pp. 496-517] to the multirate method setting. Specifically, we develop controllers based on polynomial approximations to the principal error functions for both the "fast" and "slow" time scales within multirate infinitesimal (MRI) methods. We additionally investigate a variety of approaches for estimating the errors arising from each time scale within MRI methods. We then numerically evaluate the proposed multirate controllers and error estimation strategies on a range of multirate test problems, comparing their performance against an estimated optimal performance. Through this work, we combine the most performant of these approaches to arrive at a set of multirate adaptive time step controllers that robustly achieve desired solution accuracy with minimal computational effort.
Insights from Optimizing HPL Performance on Exascale Systems: A Comparative Analysis of Panel Factorization
High performance LINPACK (HPL) remains the primary benchmark for evaluating supercomputing performance. It includes many parts with substantial internal complexity, and its performance is affected by a large number of parameters that interact in ways that are difficult to predict on large-scale heterogeneous supercomputer systems. We present a comprehensive performance analysis of HPL on Frontier, the world’s first exascale supercomputer, which achieved HPL performance of 1.35 exaflops. Through empirical parameter tuning, detailed modeling, and comparative evaluation, we uncover critical performance insights, share lessons learned, and outline best practices for effective parameter tuning on exascale systems. We introduce and evaluate two novel PDFACT strategies: a dedicated-thread (DT) variant and a GPU-based variant (GPUPDFACT) implementation using HIP cooperative groups, demonstrating that GPU-based factorization outperforms conventional CPU-based PDFACT on Frontier’s architecture. Our findings establish key performance factors for HPL on exascale systems and offer valuable guidance for future high-performance computing and benchmarking efforts.
Optimizing the Performance of Sparse-Matrix Vector Products on Next-Generation Processors
Matrix-vector products are ubiquitous in high-performance scientific applications and have a growing set of occurrences in advanced data analysis activities. Achieving high performance for these kernels is therefore paramount, in part, because these operations can consume vast amounts of application execution time. In this report we document the development of several sparse-matrix vector product kernel implementations using a variety of programming models and approaches. Each kernel is run on a broad set of matrices selected to demonstrate the wide variety of matrix structure and sparsity that is possible with a single, generic kernel. For benchmarking and performance analysis, we utilize leading computing architectures for the NNSA/ASC program including Intel's Knights Landing processor and IBM's POWER8.