Automatic Compilation of MATLAB Programs for Synergistic Execution on Heterogeneous Processors

MATLAB is an array language, initially popular for rapid prototyping, but is now being increasingly used to develop production code for numerical and scientific applications. Typical MATLAB programs have abundant data parallelism. These programs also have control flow dominated scalar regions that have an impact on the program's execution time. Today's computer systems have tremendous computing power in the form of traditional CPU cores and throughput oriented accelerators such as graphics processing units(GPUs). Thus, an approach that maps the control flow dominated regions to the CPU and the data parallel regions to the GPU can significantly improve program performance. In this paper, we present the design and implementation of MEGHA, a compiler that automatically compiles MATLAB programs to enable synergistic execution on heterogeneous processors. Our solution is fully automated and does not require programmer input for identifying data parallel regions. We propose a set of compiler optimizations tailored for MATLAB. Our compiler identifies data parallel regions of the program and composes them into kernels. The problem of combining statements into kernels is formulated as a constrained graph clustering problem. Heuristics are presented to map identified kernels to either the CPU or GPU so that kernel execution on the CPU and the GPU happens synergistically and the amount of data transfer needed is minimized. In order to ensure required data movement for dependencies across basic blocks, we propose a data flow analysis and edge splitting strategy. Thus our compiler automatically handles composition of kernels, mapping of kernels to CPU and GPU, scheduling and insertion of required data transfer. The proposed compiler was implemented and experimental evaluation using a set of MATLAB benchmarks shows that our approach achieves a geometric mean speedup of 19.8X for data parallel benchmarks over native execution of MATLAB.

Seismic Microzonation: Principles,Practices and Experiments

This paper presents an overview of the seismic microzonation and the grade/level based study along with methods used for estimating hazard. The principles of seismic microzonation along with some current practices are discussed. Summary of seismic microzonation experiments carried out in India is presented. A detailed work of seismic microzonation of Bangalore has been presented as a case study. In this case study, a seismotectonic map for microzonation area has been developed covering 350 km radius around Bangalore, India using seismicity and seismotectonic parameters of the region. For seismic microzonation Bangalore Mahanagar Palike (BMP) area of 220 km2 has been selected as the study area. Seismic hazard analysis has been carried out using deterministic as well as probabilistic approaches. Synthetic ground motion at 653 locations, recurrence relation and peak ground acceleration maps at rock level have been generated. A detailed site characterization has been carried out using borehole with standard penetration test (SPT) ―N‖ values and geophysical data. The base map and 3-dimensional sub surface borehole model has been generated for study area using geographical information system (GIS). Multichannel analysis of surface wave (MASW)method has been used to generate one-dimensional shear wave velocity profile at 58 locations and two- dimensional profile at 20 locations. These shear wave velocities are used to estimate equivalent shear wave velocity in the study area at every 5m intervals up to a depth of 30m. Because of wider variation in the rock depth, equivalent shear for the soil overburden thickness alone has been estimated and mapped using ArcGIS 9.2. Based on equivalent shear wave velocity of soil overburden thickness, the study area is classified as ―site class D‖. Site response study has been carried out using geotechnical properties and synthetic ground motions with program SHAKE2000.The soil in the study area is classified as soil with moderate amplification potential. Site response results obtained using standard penetration test (SPT) ―N‖ values and shear wave velocity are compared, it is found that the results based on shear wave velocity is lower than the results based on SPT ―N‖ values. Further, predominant frequency of soil column has been estimated based on ambient noise survey measurements using instruments of L4-3D short period sensors equipped with Reftek 24 bit digital acquisition systems. Predominant frequency obtained from site response study is compared with ambient noise survey. In general, predominant frequencies in the study area vary from 3Hz to 12Hz. Due to flat terrain in the study area, the induced effect of land slide possibility is considered to be remote. However, induced effect of liquefaction hazard has been estimated and mapped. Finally, by integrating the above hazard parameters two hazard index maps have been developed using Analytic Hierarchy Process (AHP) on GIS platform. One map is based on deterministic hazard analysis and other map is based on probabilistic hazard analysis. Finally, a general guideline is proposed by bringing out the advantages and disadvantages of different approaches.

Seismic Microzonation of Bangalore - Preliminary Maps

This paper presents an overview of the seismic microzonation and the grade/level based study along with methods used for estimating hazard. The principles of seismic microzonation along with some current practices are discussed. Summary of seismic microzonation experiments carried out in India is presented. A detailed work of seismic microzonation of Bangalore has been presented as a case study. In this case study, a seismotectonic map for microzonation area has been developed covering 350 km radius around Bangalore, India using seismicity and seismotectonic parameters of the region. For seismic microzonation Bangalore Mahanagar Palike (BMP) area of 220 km2 has been selected as the study area. Seismic hazard analysis has been carried out using deterministic as well as probabilistic approaches. Synthetic ground motion at 653 locations, recurrence relation and peak ground acceleration maps at rock level have been generated. A detailed site characterization has been carried out using borehole with standard penetration test (SPT) ―N‖ values and geophysical data. The base map and 3-dimensional sub surface borehole model has been generated for study area using geographical information system (GIS). Multichannel analysis of surface wave (MASW)method has been used to generate one-dimensional shear wave velocity profile at 58 locations and two- dimensional profile at 20 locations. These shear wave velocities are used to estimate equivalent shear wave velocity in the study area at every 5m intervals up to a depth of 30m. Because of wider variation in the rock depth, equivalent shear for the soil overburden thickness alone has been estimated and mapped using ArcGIS 9.2. Based on equivalent shear wave velocity of soil overburden thickness, the study area is classified as ―site class D‖. Site response study has been carried out using geotechnical properties and synthetic ground motions with program SHAKE2000.The soil in the study area is classified as soil with moderate amplification potential. Site response results obtained using standard penetration test (SPT) ―N‖ values and shear wave velocity are compared, it is found that the results based on shear wave velocity is lower than the results based on SPT ―N‖ values. Further, predominant frequency of soil column has been estimated based on ambient noise survey measurements using instruments of L4-3D short period sensors equipped with Reftek 24 bit digital acquisition systems. Predominant frequency obtained from site response study is compared with ambient noise survey. In general, predominant frequencies in the study area vary from 3Hz to 12Hz. Due to flat terrain in the study area, the induced effect of land slide possibility is considered to be remote. However, induced effect of liquefaction hazard has been estimated and mapped. Finally, by integrating the above hazard parameters two hazard index maps have been developed using Analytic Hierarchy Process (AHP) on GIS platform. One map is based on deterministic hazard analysis and other map is based on probabilistic hazard analysis. Finally, a general guideline is proposed by bringing out the advantages and disadvantages of different approaches.

Seismic Hazard and Microzonation Studies using Geographic Information System - A Case Study of Bangalore city

Seismic hazard and microzonation of cities enable to characterize the potential seismic areas that need to be taken into account when designing new structures or retrofitting the existing ones. Study of seismic hazard and preparation of geotechnical microzonation maps has been attempted using Geographical Information System (GIS). GIS will provide an effective solution for integrating different layers of information thus providing a useful input for city planning and in particular input to earthquake resistant design of structures in an area. Seismic hazard is the study of expected earthquake ground motions at any point on the earth. Microzonation is the process of sub division of region in to number of zones based on the earthquake effects in the local scale. Seismic microzonation is the process of estimating response of soil layers under earthquake excitation and thus the variation of ground motion characteristic on the ground surface. For the seismic microzonation, geotechnical site characterization need to be assessed at local scale (micro level), which is further used to assess of the site response and liquefaction susceptibility of the sites. Seismotectonic atlas of the area having a radius of 350km around Bangalore has been prepared with all the seismogenic sources and historic earthquake events (a catalogue of about 1400 events since 1906). We have attempted to carryout the site characterization of Bangalore by collating conventional geotechnical boreholes data (about 900 borehole data with depth) and integrated in GIS. 3-D subsurface model of Bangalore prepared using GIS is shown in Figure 1.Further, Shear wave velocity survey based on geophysical method at about 60 locations in the city has been carried out in 220 square Kms area. Site response and local site effects have been evaluated using 1-dimensional ground response analysis. Spatial variability of soil overburden depths, ground surface Peak Ground Acceleration’s(PGA), spectral acceleration for different frequencies, liquefaction susceptibility have been mapped in the 220 sq km area using GIS.ArcInfo software has been used for this purpose. These maps can be used for the city planning and risk & vulnerability studies. Figure 2 shows a map of peak ground acceleration at rock level for Bangalore city. Microtremor experiments were jointly carried out with NGRI scientists at about 55 locations in the city and the predominant frequency of the overburden soil columns were evaluated.

Modeling wrap faced reinforced soil retaining walls subjected to dynamic excitation

This paper describes some of the physical and numerical model tests of reinforced soil retaining walls subjected to dynamic excitation through uni-axial shaking tests. Models of retaining walls are constructed in a perspex box with geotextile reinforcement using the wrap around technique with dry sand backfill and instrumented with displacement sensors, accelerometers and soil pressure sensors. Numerical modelling of these shaking table tests is carried using FLAC. Numerical model is validated by comparing physical model results. Responses of wrap faced walls with different number of reinforcement layers are discussed from both the physical and numerical model tests. Results showed that the displacements are decreasing with the increase in number of reinforcement layers while acceleration amplifications are not affected significantly.

Analysis of a Novel MOEM Race Track Resontor based Vibration Sensor

In this paper we analyze a novel Micro Opto Electro Mechanical Systems (MOEMS) race track resonator based vibration sensor. In this vibration sensor the straight portion of a race track resonator is located at the foot of the cantilever beam with proof mass. As the beam deflects due to vibration, stress induced refractive change in the waveguide located over the beam lead to the wavelength shift providing the measure of vibration. A wavelength shift of 3.19 pm/g in the range of 280 g for a cantilever beam of 1750μm×450m×20μmhas been obtained. The maximum acceleration (breakdown) for these dimensions is 2900g when a safety factor of 2 is taken into account. Since the wavelength of operation is around 1.55μm hybrid integration of source and detector is possible on the same substrate. Also it is less amenable to noise as wavelength shift provides the sensor signal. This type of sensors can be used for aerospace application and other harsh environments with suitable design.

Limits of Data Level Parallelism

Abstract—A new breed of processors like the Cell Broadband Engine, the Imagine stream processor and the various GPU processors emphasize data-level parallelism (DLP) and threadlevel parallelism (TLP) as opposed to traditional instructionlevel parallelism (ILP). This allows them to achieve order-ofmagnitude improvements over conventional superscalar processors for many workloads. However, it is unclear as to how much parallelism of these types exists in current programs. Most earlier studies have largely concentrated on the amount of ILP in a program, without differentiating DLP or TLP. In this study, we investigate the extent of data-level parallelism available in programs in the MediaBench suite. By packing instructions in a SIMD fashion, we observe reductions of up to 91 % (84 % on average) in the number of dynamic instructions, indicating a very high degree of DLP in several applications. I.

Command Tracking in High Performance Aircrafts: A New Dynamic Inversion Design

This paper proposes a new straight forward technique based on dynamic inversion, which is applied for tracking the pilot commands in high performance aircrafts.Pilot commands assumed in longitudinal mode are normal acceleration and total velocity(while roll angle and lateral acceleration are maintained at zero). In lateral mode, roll rate and total velocity are used as pilot commands (while climb rate and lateral acceleration are maintained at zero). Ensuring zero lateral acceleration leads to a better turn co-ordination. A six degree-of-freedom model of F-16 aircraft is used for both control design as well as simulation studies. Promising results are obtained which are found to be superior as compared to an existing approach (which is also based on dynamic inversion). The new approach has two potential benefits, namely reduced oscillatory response and reduced control magnitude. Another advantage of this approach is that it leads to a significant reduction of tuning parameters in the control design process.

Model-following Neuro-adaptive approach for robust control of high performance aircrafts

Based on dynamic inversion, a relatively straightforward approach is presented in this paper for nonlinear flight control design of high performance aircrafts, which does not require the normal and lateral acceleration commands to be first transferred to body rates before computing the required control inputs. This leads to substantial improvement of the tracking response. Promising results are obtained from six degree-offreedom simulation studies of F-16 aircraft, which are found to be superior as compared to an existing approach (which is also based on dynamic inversion). The new approach has two potential benefits, namely reduced oscillatory response (including elimination of non-minimum phase behavior) and reduced control magnitude. Next, a model-following neuron-adaptive design is augmented the nominal design in order to assure robust performance in the presence of parameter inaccuracies in the model. Note that in the approach the model update takes place adaptively online and hence it is philosophically similar to indirect adaptive control. However, unlike a typical indirect adaptive control approach, there is no need to update the individual parameters explicitly. Instead the inaccuracy in the system output dynamics is captured directly and then used in modifying the control. This leads to faster adaptation, which helps in stabilizing the unstable plant quicker. The robustness study from a large number of simulations shows that the adaptive design has good amount of robustness with respect to the expected parameter inaccuracies in the model.

Real-time Computer Simulation of Three Dimensional Elastostatics using the Finite Point Method

Real-time simulation of deformable solids is essential for some applications such as biological organ simulations for surgical simulators. In this work, deformable solids are approximated to be linear elastic, and an easy and straight forward numerical technique, the Finite Point Method (FPM), is used to model three dimensional linear elastostatics. Graphics Processing Unit (GPU) is used to accelerate computations. Results show that the Finite Point Method, together with GPU, can compute three dimensional linear elastostatic responses of solids at rates suitable for real-time graphics, for solids represented by reasonable number of points.

Benchmark Control Problem for Highway Bridge based on FLC

Fuzzy logic control (FLC) systems have been applied as an effective control system in various fields, including vibration control of structures. The advantage of this approach is its inherent robustness and ability to handle non‐linearities and uncertainties in structural behavior and loading. The study evaluates the three‐dimensional benchmark control problem for a seismically excited highway bridge using an ANFIS driven hydraulic actuators. An ANN based training strategy that considers both velocity and acceleration feedback together with a fuzzy logic rule base is developed. Present study needs only 4 accelerometers and 4 fuzzy rule bases to determine the control force, instead of 8 accelerometers and 4 displacement transducers used in the benchmark study problem. The results obtained are better than that obtained from the benchmark control algorithm.

Experimental and Numerical Studies in a Compact Trapped Vortex Combustor: Stability Assessment and Augmentation

Fundamental studies on a compact trapped vortex combustor indicate that cavity injection strategies play a major role on flame stability. Detailed experiments indicate that blow-out occurs for a certain range of cavity air flow velocities. An unsteady RANS-based reacting flow simulation tool has been utilized to study the basic dynamics of cavity vortex for various flow conditions. The phenomenon of flame blow-out at certain intermediate cavity air velocities is explained on the basis of transition from a cavity-stabilized mode to an opposed flow stagnation mode. A novel strategy is proposed for achieving flame stability at all conditions. This involves using a flow guide vane in the path of the main flow to direct a portion of the main flow into the cavity. This seems to result in a desirable dual vortex structure, i.e., a small clockwise vortex behind the vane and large counterclockwise vortex in the cavity. Experimental results show stable flame at all flow conditions with the flow guide vane, and pressure drop is estimated to be within acceptable limits. Cold flow simulations show self-similar velocity profiles for a range of main inlet velocities, and high reverse velocity ratios (-0.3) are observed. Such a high-velocity ratio in the reverse flow shear layer profile leads to enhanced production of turbulence imperative to compact combustors. Reacting flow simulations show even higher reverse velocity ratios (above -0.7) due to flow acceleration. The flame is observed to be stable, even though minor shear layer oscillations are present in the form of vortex shedding. Self-similarity is also observed in reacting flow temperature profiles at combustor exit over the entire range of the mainstream velocity. This indicates that the present configuration holds a promise of delivering robust performance invariant of the flow operating conditions.

Aerodynamic Load Measurements at Hypersonic Speeds Using Internally Mounted Fiber-Optic Balance System

This paper describes the measurement of aerodynamic loads using fiber-optic strain gauge sensors and associated signal processors at hypersonic speeds in the 300mm hypersonic wind tunnel. at the Department of Aerospace Engineering, Indian Institute of Science. Fiber-optic sensors have been developed in USA since 1990, for variety of applications in experimental stress analysis, skin friction measurement in fluid flows, smart structures, smart materials, sensing of acoustic emission and more recently in the development of compact devices for measurement of displacement, stress/strain, pressure, temperature, acceleration etc. Our group at llSc has been playing a lead role in the use of these fiber - optic sensors for successful measurement of aerodynamic loads in wind tunnels and the first ever six-component wind tunnel strain gauge balance in the world based on fiber optic sensors was built at the Indian Institute of Science in the year 1999. We report here the results of our efforts in the development of an internal strain gauge balance for high-speed wind tunnel applications.

Parallel Computation of 2D Morse-Smale Complexes

The Morse-Smale complex is a useful topological data structure for the analysis and visualization of scalar data. This paper describes an algorithm that processes all mesh elements of the domain in parallel to compute the Morse-Smale complex of large two-dimensional data sets at interactive speeds. We employ a reformulation of the Morse-Smale complex using Forman's Discrete Morse Theory and achieve scalability by computing the discrete gradient using local accesses only. We also introduce a novel approach to merge gradient paths that ensures accurate geometry of the computed complex. We demonstrate that our algorithm performs well on both multicore environments and on massively parallel architectures such as the GPU.

Accelerating Numerical Linear Algebra Kernels on a Scalable Run Time Reconfigurable Platform

Numerical Linear Algebra (NLA) kernels are at the heart of all computational problems. These kernels require hardware acceleration for increased throughput. NLA Solvers for dense and sparse matrices differ in the way the matrices are stored and operated upon although they exhibit similar computational properties. While ASIC solutions for NLA Solvers can deliver high performance, they are not scalable, and hence are not commercially viable. In this paper, we show how NLA kernels can be accelerated on REDEFINE, a scalable runtime reconfigurable hardware platform. Compared to a software implementation, Direct Solver (Modified Faddeev's algorithm) on REDEFINE shows a 29X improvement on an average and Iterative Solver (Conjugate Gradient algorithm) shows a 15-20% improvement. We further show that solution on REDEFINE is scalable over larger problem sizes without any notable degradation in performance.