Piero Sguazzero

Migration of seismic data by phase shift plus interpolation

Under the horizontally layered velocity assumption, migration is defined by a set of independent ordinary differential equations in the wavenumber‐frequency domain. The wave components are extrapolated downward by rotating their phases. This paper shows that one can generalize the concepts of the phase‐shift method to media having lateral velocity variations. The wave extrapolation procedure consists of two steps. In the first step, the wave field is extrapolated by the phase‐shift method using l laterally uniform velocity fields. The intermediate result is l reference wave fields. In the second step, the actual wave field is computed by interpolation from the reference wave fields. The phase shift plus interpolation (PSPI) method is unconditionally stable and lends itself conveniently to migration of three‐dimensional data. The performance of the methods is demonstrated on synthetic examples. The PSPI migration results are then compared with those obtained from a finite‐difference method.

On the construction and efficiency of staggered numerical differentiators for the wave equation

Finite‐difference (FD) techniques have established themselves as viable tools for the numerical modeling of wave propagation. The accuracy and the computational efficiency of numerical modeling can be enhanced by using high‐order spatial differential operators (Dablain,1986).

Microtasking on IBM multiprocessors

The demand for very high computing performance has become increasingly common in many scientific and engineering environments. In addition to vector processing, parallel computing is now considered a useful way to enhance performance. However, parallel computing tends to be unpopular among users because, with presently available technology and software, it requires explicit programmer intervention to exploit architectural parallelism. This intervention can be minor in some cases, but it often requires a non-negligible amount of program restructuring, or even a reformulation of some of the algorithms used. In addition, it makes program debugging considerably more difficult. Tools for interprocedural program analysis, able to analyze at high levels large FORTRAN programs composed of many subroutines and to perform an automatic high-level parallelism could also be exploited automatically.

Parallel PIC plasma simulation through particle decomposition techniques

Abstract Parallelization of a particle-in-cell (PIC) code has been accomplished through a “particle decomposition” technique instead of the more usual “domain decomposition” one. The adopted technique requires a moderate effort in porting the code in parallel form and results in intrinsic load balancing and modest inter-processor communication. The resulting data parallel implementation has been carried out within the High Performance Fortran (HPF) framework, and tested on the IBM SP parallel system. The performance tests obtained confirm the hypothesis of high effectiveness of the strategy, if targeted towards moderately parallel architectures. Optimal use of resources is also discussed with reference to a specific physics problem.

Seismic computations on the IBM 3090 vector multiprocessor

Computerized seismic prospecting is an echo-ranging technique usually targeted at accurate mapping of oil and gas reservoirs. In seismic surveys an impulsive source, often an explosive charge, located at the earths surface generates elastic waves which propagate in the subsurface; these waves are scattered by the earths geological discontinuities back to the surface, where an array of receivers registers the reflected signals. The data recorded are then processed in a complex sequence of steps. Among them, seismic migration and stacking velocity estimation represent two characteristic components of the process solving the inverse problem of recovering the structure and the physical parameters of the earths geologic layers from echo measurements. A complementary tool in relating seismic data to the earths inhomogeneities is provided by seismic numerical models, which assume a subsurface structure and compute the seismic data which would be collected in a field survey, by solving the direct problem of exploration geophysics. This paper describes a vectorized and parallelized implementation of a two-dimensional seismic elastic model on the IBM 3090 VF Vector Multiprocessor. An implementation of a parallel seismic migration algorithm is then described. The paper also reports performance data for a vector/parallel implementation on the IBM 3090 of some typical seismic velocity estimation algorithms. The three problems chosen are representative of a wide class of geophysical computations, and the results summarized in this paper show their suitability for efficient implementation on the IBM 3090 Vector Multiprocessor; combined vector/parallel speedups in the range 15–25 are in fact observed.

Dispersion-bounded numerical integration of the elastodynamic equations with cost-effective staggered schemes

Abstract Known procedures for designing numerical schemes for the integration of elastodynamic equations with explicit control over numerical dispersion are reviewed. In the literature, the analysis of such schemes has concentrated on the discrete space differentiators, and has neglected the role played by time discretization in the overall accuracy. In this paper we define a computational cost for a given dispersion error bound which fully includes the effect of temporal differencing. For some representative schemes based on leap-frog time marching, we provide an optimal operating point (time sampling rate and number of grid points per shortest wavelength) which minimizes the computational cost for a given dispersion error threshold. Based on this notion of cost, we introduce new optimal operators for staggered grids. Additionally, we introduce the notion of composite differentiators to design still more cost-effective schemes. The cost of the proposed schemes is shown to be less than that of known finite difference (FD) operators and compares favorably with pseudo-spectral (PS) algorithms. Numerical simulations are presented to illustrate the effectiveness of the new operators.

Dense linear systems FORTRAN solvers on the IBM 3090 vector multiprocessor

Abstract The IBM 3090 is a vector multiprocessor with a hierarchical memory system. We show with two examples (the LU and Householder factorizations) that the complex memory system and the vector hardware can be used efficiently by recasting the basic algorithms in terms of high-level matrix-matrix modules.

Large-scale computing on clustered vector multiprocessors

A two-level parallel implementation of a large-scale geophysical simulation code is presented. The software/hardware environment utilizes IBM Clustered Fortran, allowing a single application program to execute concurrently on two IBM 3090 computers (first level of parallelism) while exploiting the multiple vector processors of each 3090 system (second level of parallelism). The experiments reported show that the problem is characterized by large task granularity and small communication/computation ratio, thus leading to sustained parallel speed-ups of nearly ten on a cluster of two 3090 computers, totalling twelve vector processors.<<ETX>>

Cost-Effective Staggered Numerical Integration of the Wave Equation

Finite-difference (FD) schemes for the numerical integration of the wave equation are generally designed with one of two criteria: (a) maximize the numerical accuracy order (as in the case of conventional FD operators); (b) minimize the simulation error of some physical feature of wave propagation (e.g. the dispersion relation) within a spectral frequency band (as introduced recently by Hotberg). The analysis of cost-effectiveness for a required error bound of the second schemes has neglected the role played by time discretization in the overall accuracy.

Parallelism in Seismic Computing

This paper describes a vectorized and parallelized implementation of a two-dimensional pseudo-spectralseismic elastic model and of a (frequency-domain)seismic migration algorithm on the (tightly-coupled) vector multiprocessor IBM 3090 VF. Performance data of alternative parallel implementations on an LCAP (loosely coupled system of array processors FPS 164) are also given.The paper reports additionally performance data of a vector/parallel implementation on the IBM 3090 of some characteristicseismic velocity estimation algorithms. The three problems chosen are representative of a wide class of geophysical computations and the results summarized in this paper show their suitability to efficient implementation on the IBM 3090 vector multiprocessor: combined vector/parallel speedups in the range 15–30 are in fact observed.