Sharif Digital Repository

Newly developed 3D die stacking technologies affords us the possibility to revisit the idea of Processing-in-Memory (PIM) as implementation hurdles are overcome. We now have the opportunity to offload the data intensive parts of our program to the PIM in form of kernels to be able to take advantage of the high internal bandwidth of the memory modules. Memory access latency and bandwidth are two major bottlenecks in today’s high-performance computers and new use-cases are moving faster than ever before towards this mode of computing. With new graph processing and neural network applications being developed every day, having a performance model of such systems helps in predicting the behavior... 

In this study we will show that the variation rate of the memristance of the memristive device depends completely on its current memristance which means that it can change significantly with time during the learning phase. This phenomenon can degrade the performance of learning methods like Spike Timing-Dependent Plasticity (STDP) and cause the corresponding neuromorphic systems to become unstable  

In this paper, a simpler approach compared to the existing approaches is developed to analyze nuclear reactor dynamics based on the explicit Monte Carlo method. A new population control method is also introduced to prevent neutron population growth and consequent computer memory shortages, which also increases simulation speed. The scheme is applied for time-dependent particle tracking in three-dimensional arbitrary geometries in the presence of feedbacks through a code named MCSP-Explicit. Changes in material density, as well as geometry dimensions, are also considered during simulation. MCSP-Explicit can be run with either continuous or multigroup data libraries, and it is further boosted... 

This work presents a case for general support for processing-in-memory (PIM) in compilers and puts forth an approach to face it along with a simple model. The ultimate goal of the work is to implement the features in a general-purpose compiler that can compile for any homogeneous ISA system, so the benefits from PIM are not limited to niche use-cases. © 2013 IEEE  

In this paper, a simpler approach compared to the existing approaches is developed to analyze nuclear reactor dynamics based on the explicit Monte Carlo method. A new population control method is also introduced to prevent neutron population growth and consequent computer memory shortages, which also increases simulation speed. The scheme is applied for time-dependent particle tracking in three-dimensional arbitrary geometries in the presence of feedbacks through a code named MCSP-Explicit. Changes in material density, as well as geometry dimensions, are also considered during simulation. MCSP-Explicit can be run with either continuous or multigroup data libraries, and it is further boosted... 

Theoretical and numerical analyses of rotating disks with non-uniform thickness and material properties subjected to thermo-mechanical loadings have been carried out by variable material properties (VMP), Runge-Kutta's (RK) and finite element (FE) methods. The material is assumed to be elastic-linear hardening. A power form function is used to describe the temperature gradient with the higher temperature at outer surface. Von-Mises theory has been used as failure criterion. The effects of geometry, material and thermal loading parameters as well as boundary conditions on radial, hoop and equivalent stress distributions which have not been studied in much detail in previous works have been... 

The propagation of intense optical beams in a gas undergoing ionization is analyzed through a threedimensional finite-difference time-domain (3D-FDTD) scheme. The propagation dynamics include the effects of diffraction, nonlinear self-focusing, and ionization. For sufficiently intense optical beams the neutral gas undergoes ionization, generating a plasma which tends to defocus the beam. Balancing of diffraction, plasma defocusing, and nonlinear self-focusing may lead to self-guided results. In this paper, necessary relations have been introduced into the conventional FDTD formulation to account for the nonlinear behaviors. Furthermore, a concurrent utilization of computer memory and disk... 

A genetic-based pore network extraction method from micro-computed tomography (micro-CT) images is proposed in this paper. Several variables such as the number, radius and location of pores, the coordination number, as well as the radius and length of the throats are used herein as the optimization parameters. Two approaches to generate the pore network structure are presented. Unlike previous algorithms, the presented approaches are directly based on minimizing the error between the extracted network and the real porous medium. This leads to the generation of more accurate results while reducing required computational memories. Two different objective functions are used in building the... 

Overall turbine analysis requires large CPU time and computer memory, even in the present days. As a result, choosing an appropriate computational domain accompanied by a suitable boundary condition can dramatically reduce the time cost of computations. This work compares different geometries for numerical investigation of the 3D flow in the runner of a Francis turbine, and presents an optimum geometry with least computational effort and desirable numerical accuracy. The numerical results are validated with a GAMM Francis Turbine runner, which was used as a test case (GAMM workshop on 3D computation of incompressible internal flows, 1989) in which the geometry and detailed best efficiency...