Search results

This paper aims to review in-memory computing (IMC) for machine learning (ML) applications from history, architectures and options aspects. In this review, the authors investigate different architectural aspects and collect and provide our comparative evaluations.

Collecting over 40 IMC papers related to hardware design and optimization techniques of recent years, then classify them into three optimization option categories: optimization through graphic processing unit (GPU), optimization through reduced precision and optimization through hardware accelerator. Then, the authors brief those techniques in aspects such as what kind of data set it applied, how it is designed and what is the contribution of this design.

ML algorithms are potent tools accommodated on IMC architecture. Although general-purpose hardware (central processing units and GPUs) can supply explicit solutions, their energy efficiencies have limitations because of their excessive flexibility support. On the other hand, hardware accelerators (field programmable gate arrays and application-specific integrated circuits) win on the energy efficiency aspect, but individual accelerator often adapts exclusively to ax single ML approach (family). From a long hardware evolution perspective, hardware/software collaboration heterogeneity design from hybrid platforms is an option for the researcher.

IMC’s optimization enables high-speed processing, increases performance and analyzes massive volumes of data in real-time. This work reviews IMC and its evolution. Then, the authors categorize three optimization paths for the IMC architecture to improve performance metrics.

This study aims to modify the standard probabilistic lattice Boltzmann methodology (LBM) cellular automata (CA) algorithm to enable a more realistic and accurate computation of the ensemble rather than individual particle trajectories that need to be updated from one time step to the next (allowing, as such, a fraction of the collection of particles in any lattice grid cell to be updated in a time step, rather than the entire collection of particles as in the standard LBM-CA algorithm leading to a better representation of the dynamic interaction between the particles and the background flow). Exploitation of the inherent parallelism of the modified LBM-CA algorithm to provide a computationally efficient scheme for computation of particle-laden flows on readily available commodity general-purpose graphics processing units (GPGPUs).

This paper presents a framework for the implementation of a LBM for the simulation of particle transport and deposition in complex flows on a GPGPU. Towards this objective, the authors have shown how to map the data structure of the LBM with a multiple-relaxation-time (MRT) collision operator and the Smagorinsky subgrid-scale turbulence model (for turbulent fluid flow simulations) coupled with a CA probabilistic method (for particle transport and deposition simulations) to a GPGPU to give a high-performance computing tool for the calculation of particle-laden flows.

A fluid-particle simulation using our LBM-MRT-CA algorithm run on a single GPGPU was 160 times as computationally efficient as the same algorithm run on a single CPU.

The method is limited by the available computational resources (e.g. GPU memory size).

A new 3D LBM-MRT-CA model was developed to simulate the particle transport and deposition in complex laminar and turbulent flows with different hydrodynamic characteristics (e.g. vortex shedding, impingement, free shear layer, turbulent boundary layer). The solid particle information is encapsulated locally at the lattice grid nodes, allowing for straightforward mapping of the datastructure onto a GPGPU enabling a massive parallel execution of the LBM-MRT-CA algorithm. The new particle transport algorithm was based on the local (bulk) particle density and velocity and provides more realistic results for the particle transport and deposition than the standard LBM-CA algorithm.

Once in a while, you should take stock of your personal computing environment. What is on your system? How did it get there? What do you actually use? How did you arrive at your hardware configuration, and does it still meet your needs? You may find that you can free up some disk space in the process; at the very least, you'll understand your situation better. The author goes through this exercise both as an example of what it can show and because full disclosure is important for this series of articles. You need to know the background for the advice that appears here. The author discloses his current computing environments, how they got that way, and what that may mean. He also points out the real limits within which he operates as a PC commentator. When you go through the software on your system, you should check to see whether it represents ethical computing. The author offers a few notes on ethical issues related to software. The author also provides notes from PC literature for January‐June 1992.

You can buy and use a personal computer without understanding the language—but it's a little hard to read these articles, and a lot harder to understand the Held thoroughly. The author starts a new sequence of Trailing Edge articles by denning some of the terms used in the articles and in the field and mentioning some of the other terms you'll see used but rarely defined. The author also provides notes from PC literature for July‐September 1992. For a year in which prices were supposed to stabilize, it's been a remarkable summer: the new general‐purpose machine is supposed to be a 50MHz 486DX2, a well‐equipped 486SX now goes for less than $2,000 complete, and even the hottest new machines are coming out at reasonable prices.

The purpose of this paper is to investigate possibilities to adopt state-of-the-art computer graphics technologies for big data visualization in engineering applications. Toward this purpose, a conceptual heterogeneous system is proposed for graphical rendering, which is established with multiple central processing unit cores and multiple graphics processing unit GPUs.

The design of the system supports both general-purpose computation and graphics-related computation. Three processing components are discussed to fulfill the execution requirements in load balancing, data streaming and display. This design fully uses computational and memory resources and enhances the performance with the support of GPU-based parallelization.

The advantages and disadvantages of particular technical methods for each processing component are discussed. The possible ways to integrate them are analyzed.

This work has contributions of using computer graphics technologies in engineering applications.

The purpose of this paper is to demonstrate a real time 3D localization and mapping approach for the USAR (Urban Search and Rescue) robotic application, focusing on the performance and the accuracy of the General‐purpose computing on graphics processing units (GPGPU)‐based iterative closest point (ICP) 3D data registration implemented using modern GPGPU with FERMI architecture.

The authors put all the ICP computation into GPU, and performed the experiments with registration up to 106 data points. The main goal of the research was to provide a method for real‐time data registration performed by a mobile robot equipped with commercially available laser measurement system 3D. The main contribution of the paper is a new GPGPU based ICP implementation with regular grid decomposition. It guarantees high accuracy as equivalent CPU based ICP implementation with better performance.

The authors have shown an empirical analysis of the tuning of GPUICP parameters for obtaining much better performance (acceptable level of the variance of the computing time) with minimal lost of accuracy. Loop closing method is added and demonstrates satisfactory results of 3D localization and mapping in urban environments. This work can help in building the USAR mobile robotic applications that process 3D cloud of points in real time.

This work can help in developing real time mapping for USAR robotic applications.

The paper proposes a new method for nearest neighbor search that guarantees better performance with minimal loss of accuracy. The variance of computational time is much less than SoA.

If confession is good for the soul, this is Crawford's personal revival meeting. Yes, the hardcore text maven and trailing‐edge devotee has gone GUI: most of his home computing now uses a true graphical user interface. The author says that the taste of crow has passed and that the new environment works very well, albeit not without a few frustrations. This article discusses the author's move to Windows and some of the good and bad points of that interface. The author includes some tips on Windows, as seems inevitable for any article on that topic. The author also provides some additional notes related to previous columns, on clip‐art collections and the actual construction of TrueType typefaces. As usual, the article concludes with notes on the recent PC literature.

The purpose of this paper is the examination of fluid flow around NACA0012 airfoil, with the aim of the numerical validation between the experimental results in the wind tunnel and the Lattice Boltzmann method (LBM) analysis, for the medium Reynolds number (Re = 191,000). The LBM–large Eddy simulation (LES) method described in this paper opens up opportunities for faster computational fluid dynamics (CFD) analysis, because of the LBM scalability on high performance computing architectures, more specifically general purpose graphics processing units (GPGPUs), pertaining at the same time the high resolution LES approach.

Process starts with data collection in open-circuit wind tunnel experiment. Furthermore, the pressure coefficient, as a comparative variable, has been used with varying angle of attack (2°, 4°, 6° and 8°) for both experiment and LBM analysis. To numerically reproduce the experimental results, the LBM coupled with the LES turbulence model, the generalized wall function (GWF) and the cumulant collision operator with D3Q27 velocity set has been used. Also, a mesh independence study has been provided to ensure result congruence.

The proposed LBM methodology is capable of highly accurate predictions when compared with experimental data. Besides, the special significance of this work is the possibility of experimental and CFD comparison for the same domain dimensions.

Considering the quality of results, root-mean-square error (RMSE) shows good correlations both for airfoil’s upper and lower surface. More precisely, maximal RMSE for the upper surface is 0.105, whereas 0.089 for the lower surface, regarding all angles of attack.

Recent Trailing Edge articles have discussed typefaces and graphics. This column discusses putting it all together: economical desktop publishing. There has never been a better time for libraries to become desktop publishers, and some will find that doing so requires no new software or hardware. The author discusses changes that have made desktop publishing such an appealing and reasonably‐priced proposition in 1994 and some of your options for getting started and moving on. He brings the typeface discussion up to date with a startling recent development and defines the difference between true desktop publishing and the spare‐no‐expense field that the “desktop publishing” magazines cover. A sidebar notes a series of desktop publishing workshops that the author is offering as part of LITA's regional institutes program. Finally, the author adds notes on the personal computing literature for January to March 1994, now including some Macintosh magazines and, soon, CD‐ROM/multimedia publications.