Engineering topics
Publications and source records attributed to Saini, Subhash.
This viewgraph presentation covers the advantages of mesh adaptation, unstructured grids, and dynamic load balancing. It illustrates parallel adaptive communications, and explains PLUM (Parallel dynamic load balancing for adaptive unstructured meshes), and PSAW (Proper Self Avoiding Walks).
Porting applications to new high performance parallel and distributed platforms is a challenging task. Writing parallel code by hand is time consuming and costly, but this task can be simplified by high level languages and would even better be automated by parallelizing tools and compilers. The definition of HPF (High Performance Fortran, based on data parallel model) and OpenMP (based on shared memory parallel model) standards has offered great opportunity in this respect. Both provide simple and clear interfaces to language like FORTRAN and simplify many tedious tasks encountered in writing message passing programs. In our study, we implemented the parallel versions of the NAS Benchmarks with HPF and OpenMP directives. Comparison of their performance with the MPI implementation and pros and cons of different approaches will be discussed along with experience of using computer-aided tools to help parallelize these benchmarks. Based on the study, potentials of applying some of the techniques to realistic aerospace applications will be presented.
Development of HPF versions of NPB and ARC3D showed that HPF has potential to be a high level language for parallelization of CFD applications. The use of HPF requires an intimate knowledge of the applications and a detailed analysis of data affinity, data movement and data granularity. Since HPF hides data movement from the user even with this knowledge it is easy to overlook pieces of the code causing low performance of the application. In order to simplify and accelerate the task of developing HPF versions of existing CFD applications we have designed and partially implemented ADAPT (Automatic Data Distribution and Placement Tool). The ADAPT analyzes a CFD application working on a single structured grid and generates HPF TEMPLATE, (RE)DISTRIBUTION, ALIGNMENT and INDEPENDENT directives. The directives can be generated on the nest level, subroutine level, application level or inter application level. ADAPT is designed to annotate existing CFD FORTRAN application performing computations on single or multiple grids. On each grid the application can considered as a sequence of operators each applied to a set of variables defined in a particular grid domain. The operators can be classified as implicit, having data dependences, and explicit, without data dependences. In order to parallelize an explicit operator it is sufficient to create a template for the domain of the operator, align arrays used in the operator with the template, distribute the template, and declare the loops over the distributed dimensions as INDEPENDENT. In order to parallelize an implicit operator, the distribution of the operator's domain should be consistent with the operator's dependences. Any dependence between sections distributed on different processors would preclude parallelization if compiler does not have an ability to pipeline computations. If a data distribution is "orthogonal" to the dependences of an implicit operator then the loop which implements the operator can be declared as INDEPENDENT.
The Einstein relation between spontaneous emission and absorption was originally derived for a system consists of a two-state subsystem representing matter and harmonic fields representing radiation. The derivation is based on the detailed balance between these two subsystems under thermal equilibrium. The relationship was later investigated in connection with the interactions between radiation field and solids or semiconductors. The simple derivation dose not hold for semiconductors in general. In certain limiting cases, simple relation was obtained. The validity of this relation is important not only because of its fundamental role connecting two of the most fundamental optical processes in semiconductors, but mostly also because of its wide use as a practical method to measure the optical gain of a semiconductor. The validity of this relation for semiconductors has been an issue of controversial for some time. In this paper we numerically examine the validity of this relationship for several different lineshapes including Lorentzian, Gaussian, Sech, and a convoluted double Lorentzians (CDL). We find out that at relatively low density above transparency level, all first three lineshapes violate the Einstein relation. The relation is approximately valid at high density. At very high density, the validity of the Einstein relation holds well for all three lineshapes. The reason behind this observation is explained. The CDL lineshape has been shown analytically to obey the Einstein relationship previously. We show that for a 2D semiconductor with parabolic bands, the CDL lineshape can be integrated analytically. This analytic lineshape is compared with a simple Lorentzian lineshape.
We present a model for quantum computation using $n$ steady 3-level atoms kept inside a quantum cavity, or using $n$ quantum-dots (QDs) kept inside a quantum cavity. In this model one external laser is pointed towards all the atoms/QDs, and $n$ pairs of electrodes are addressing the atoms/QDs, so that each atom is addressed by one pair. The energy levels of each atom/QD are controlled by an external Stark field given to the atom/QD by its external pair of electrodes. Transition between two energy levels of an individual atom/ QD are controlled by the voltage on its electrodes, and by the external laser. Interactions between two atoms/ QDs are performed with the additional help of the cavity mode (using on-resonance condition). Laser frequency, cavity frequency, and energy levels are far off-resonance most of the time, and they are brought to the resonance (using the Stark effect) only at the time of operations. Steps for a controlled-NOT gate between any two atoms/QDs have been described for this model. Our model demands some challenging technological efforts, such as manufacturing single-electron QDs inside a cavity. However, it promises big advantages over other existing models which are currently implemented, and might enable a much easier scale-up, to compute with many more qubits.
Modeling and simulation are important to understand laser operation and to optimize and design device functions. Numerical simulation of VCSEL (Vertical Cavity Surface Emitting Lasers) has been largely based on solving time-independent Helmholtz equation or time dependent coupled mode equations. There are various advantages for choosing these approaches. However, the disadvantages are also apparent. The former cannot handle dynamical mode competition seen in VCSELs, while the latter assumes a given type and number of modes a priori. Furthermore, the microscopic physics of heterstructures and electron-hole plasma is very often represented by a few parameters such as linear gain coefficients and the linewidth enhancement factor. These are over simplification of space and frequency (wavelength) dependent gain and refractive index functions. When the space-time dynamical operation of VCSELs becomes important, these simple approximations become questionable. In this paper, we apply a recently developed model for edge-emitting lasers to a gain guided VCSEL for space-time domain simulation. This model takes into account the actual nonlinear dependence of gain and refractive index on frequency and carrier density within the frame work of the effective Bloch equations. The corresponding partial differential equations are solved directly by finite difference methods. Laser behavior with increasing pumping current is investigated in detail. Special attention is paid to the dynamical competition of the transverse modes.
This presentation discusses coupling between carbon nanotubes (CNT), simple metals (FEG) and a graphene sheet. The graphene sheet did not couple well with FEG, but the combination of a graphene strip and CNT did couple well with most simple metals.
We show that quantum effects are likely to significantly degrade the performance of MOSFETs (metal oxide semiconductor field effect transistor) as these devices are scaled below 100 nm channel length and 2 nm oxide thickness over the next decade. A general and computationally efficient electronic device model including quantum effects would allow us to monitor and mitigate these effects. Full quantum models are too expensive in multi-dimensions. Using a general but efficient PDE solver called PROPHET, we implemented the density-gradient (DG) quantum correction to the industry-dominant classical drift-diffusion (DD) model. The DG model efficiently includes quantum carrier profile smoothing and tunneling in multi-dimensions and for any electronic device structure. We show that the DG model reduces DD model error from as much as 50% down to a few percent in comparison to thin oxide MOS capacitance measurements. We also show the first DG simulations of gate oxide tunneling and transverse current flow in ultra-scaled MOSFETs. The advantages of rapid model implementation using the PDE solver approach will be demonstrated, as well as the applicability of the DG model to any electronic device structure.
Porting applications to new high performance parallel and distributed platforms is a challenging task. Writing parallel code by hand is time consuming and costly, but the task can be simplified by high level languages and would even better be automated by parallelizing tools and compilers. The definition of HPF (High Performance Fortran, based on data parallel model) and OpenMP (based on shared memory parallel model) standards has offered great opportunity in this respect. Both provide simple and clear interfaces to language like FORTRAN and simplify many tedious tasks encountered in writing message passing programs. In our study we implemented the parallel versions of the NAS Benchmarks with HPF and OpenMP directives. Comparison of their performance with the MPI implementation and pros and cons of different approaches will be discussed along with experience of using computer-aided tools to help parallelize these benchmarks. Based on the study,potentials of applying some of the techniques to realistic aerospace applications will be presented
Nanomechanics of single-wall C, BN and BC$_3$ and B doped C nanotubes under axial compression and tension are investigated through a generalized tight-binding molecular dynamics (GTBMD) and {\it ab-initio} electronic structure methods. The dynamic strength of BN, BC$_3$ and B doped C nanotubes for small axial strain are comparable to each other. The main difference is in the critical strain at which structural collapse occurs. For example, even a shallow doping with B lowers the value of critical strain for C nanotubes. The critical strain for BN nanotube is found to be more than that for the similar C nanotube. Once the structural collapse starts to occur we find that carbon nanotubes irreversibly go into plastic deformation regime via the formation of tetrahedral (four-fold coordinated) bonds at the location of sharp pinches or kinks. This finding is considerably different from the classical MD (molecular dynamics) simulation results known so far. The energetics and electronic densities of states of the collapsed structures, investigated with {\it ab-initio) methods, will also be discussed.
This presentation reviews publicly available information related to carbon based nanotechnology. Topics covered include nanomechanics, carbon based electronics, nanodevice/materials applications, nanotube motors, nano-lithography and H2O storage in nanotubes.
This talk considers simplified finite element discretization techniques for first-order systems of conservation laws equipped with a convex (entropy) extension. Using newly developed techniques in entropy symmetrization theory, simplified forms of the Galerkin least-squares (GLS) and the discontinuous Galerkin (DG) finite element method have been developed and analyzed. The use of symmetrization variables yields numerical schemes which inherit global entropy stability properties of the POE system. Central to the development of the simplified GLS and DG methods is the Degenerative Scaling Theorem which characterizes right symmetrizes of an arbitrary first-order hyperbolic system in terms of scaled eigenvectors of the corresponding flux Jacobean matrices. A constructive proof is provided for the Eigenvalue Scaling Theorem with detailed consideration given to the Euler, Navier-Stokes, and magnetohydrodynamic (MHD) equations. Linear and nonlinear energy stability is proven for the simplified GLS and DG methods. Spatial convergence properties of the simplified GLS and DO methods are numerical evaluated via the computation of Ringleb flow on a sequence of successively refined triangulations. Finally, we consider a posteriori error estimates for the GLS and DG demoralization assuming error functionals related to the integrated lift and drag of a body. Sample calculations in 20 are shown to validate the theory and implementation.
The possibility of nanoscale etching and indentation of Si(001)(2x1) surface by (8,0) and (10,10) carbon nanotube tips is demonstrated, for the first time, by classical molecular dynamics simulations employing Tersoff's many-body potential for a mixed C/Si/Ge system. In the nanotube tip barely touching the surface scenario atomistic etching is observed, where as in the nanoindentation scenario nanotube tip penetrates the surface without much hindrance. The results are explained in terms of the relative strength of C-C, C-Si, and Si-Si bonds.
A series of efforts have been devoted to investigating methods of porting and parallelizing applications quickly and efficiently for new architectures, such as the SCSI Origin 2000 and Cray T3E. This report presents the parallelization of a CFD application, ARC3D, using the computer-aided tools, Cesspools. Steps of parallelizing this code and requirements of achieving better performance are discussed. The generated parallel version has achieved reasonably well performance, for example, having a speedup of 30 for 36 Cray T3E processors. However, this performance could not be obtained without modification of the original serial code. It is suggested that in many cases improving serial code and performing necessary code transformations are important parts for the automated parallelization process although user intervention in many of these parts are still necessary. Nevertheless, development and improvement of useful software tools, such as Cesspools, can help trim down many tedious parallelization details and improve the processing efficiency.
We present the HPF implementation of BT, SP, LU, FT, and MG of NPB2.3-serial benchmark set, The implementation is based on HPF performance model of the benchmark specific operations with distributed arrays. We present profiling and performance data on SGI origin 2000 and compare the results with NPB2.3. We discuss advantages and limitations of HPF and pghpf compiler.
There have been many studies of the linear response ac conductance of a double barrier resonant tunneling structure (DBRTS), both at zero and finite dc biases. While these studies are important, they fail to self consistently include the effect of the time dependent charge density in the well. In this paper, we calculate the ac conductance at both zero and finite do biases by including the effect of the time dependent charge density in the well in a self consistent manner. The charge density in the well contributes to both the flow of displacement currents in the contacts and the time dependent potential in the well. We find that including these effects can make a significant difference to the ac conductance and the total ac current is not equal to the simple average of the non-selfconsistently calculated conduction currents in the two contacts. This is illustrated by comparing the results obtained with and without the effect of the time dependent charge density included correctly. Some possible experimental scenarios to observe these effects are suggested.
Mesh adaption is a powerful tool for efficient unstructured-grid computations but causes load imbalance among processors on a parallel machine. We present a novel method called PLUM to dynamically balance the processor workloads with a global view. This paper presents the implementation and integration of all major components within our dynamic load balancing strategy for adaptive grid calculations. Mesh adaption, repartitioning, processor assignment, and remapping are critical components of the framework that must be accomplished rapidly and efficiently so as not to cause a significant overhead to the numerical simulation. A data redistribution model is also presented that predicts the remapping cost on the SP2. This model is required to determine whether the gain from a balanced workload distribution offsets the cost of data movement. Results presented in this paper demonstrate that PLUM is an effective dynamic load balancing strategy which remains viable on a large number of processors.
Dynamic mesh adaptation on unstructured grids is a powerful tool for computing unsteady three-dimensional problems that require grid modifications to efficiently resolve solution features. By locally refining and coarsening the mesh to capture phenomena of interest, such procedures make standard computational methods more cost effective. Highly refined meshes are required to accurately capture shock waves, contact discontinuities, vortices, and shear layers in fluid flow problems. Adaptive meshes have also proved to be useful in several other areas of computational science and engineering like computer vision and graphics, semiconductor device modeling, and structural mechanics. Local mesh adaptation provides the opportunity to obtain solutions that are comparable to those obtained on globally-refined grids but at a much lower cost. Additional information is contained in the original extended abstract.