The semi-automatic parallelisation of scientific application codes using a computer aided parallelisation toolkit

The shared-memory programming model can be an effective way to achieve parallelism on shared memory parallel computers. Historically however, the lack of a programming standard using directives and the limited scalability have affected its take-up. Recent advances in hardware and software technologies have resulted in improvements to both the performance of parallel programs with compiler directives and the issue of portability with the introduction of OpenMP. In this study, the Computer Aided Parallelisation Toolkit has been extended to automatically generate OpenMP-based parallel programs with nominal user assistance. We categorize the different loop types and show how efficient directives can be placed using the toolkit's in-depth interprocedural analysis. Examples are taken from the NAS parallel benchmarks and a number of real-world application codes. This demonstrates the great potential of using the toolkit to quickly parallelise serial programs as well as the good performance achievable on up to 300 processors for hybrid message passing-directive parallelisations.

Portability, predictability and performance for parallel computing: BSP in practice

We report on practical experience using the Oxford BSP Library to parallelize a large electromagnetic code, the British Aerospace finite-difference time-domain code EMMA T:FD3D. The Oxford BS Library is one of the first realizations of the Bulk Synchronous Parallel computational model to be targeted at numerically intensive scientific (typically Fortran) computing. The BAe EMMA code is one of the first large-scale applications to be parallelized using this library, and it is an important demonstration of the cost effectiveness of the BSP approach. We illustrate how BSP cost-modelling techniques can be used to predict and optimize performance for single-source programs across different parallel platforms. We provide predicted and observed performance figures for an industrial-strength, single-source parallel code for a variety of real parallel architectures: shared memory multiprocessors, workstation clusters and massively parallel platforms.

CAPLib - A 'thin layer' message passing library to support computational mechanics codes on distributed memory parallel systems

The Computer Aided Parallelisation Tools (CAPTools) [Ierotheou, C, Johnson SP, Cross M, Leggett PF, Computer aided parallelisation tools (CAPTools)-conceptual overview and performance on the parallelisation of structured mesh codes, Parallel Computing, 1996;22:163±195] is a set of interactive tools aimed to provide automatic parallelisation of serial FORTRAN Computational Mechanics (CM) programs. CAPTools analyses the user's serial code and then through stages of array partitioning, mask and communication calculation, generates parallel SPMD (Single Program Multiple Data) messages passing FORTRAN. The parallel code generated by CAPTools contains calls to a collection of routines that form the CAPTools communications Library (CAPLib). The library provides a portable layer and user friendly abstraction over the underlying parallel environment. CAPLib contains optimised message passing routines for data exchange between parallel processes and other utility routines for parallel execution control, initialisation and debugging. By compiling and linking with different implementations of the library, the user is able to run on many different parallel environments. Even with today's parallel systems the concept of a single version of a parallel application code is more of an aspiration than a reality. However for CM codes the data partitioning SPMD paradigm requires a relatively small set of message-passing communication calls. This set can be implemented as an intermediate `thin layer' library of message-passing calls that enables the parallel code (especially that generated automatically by a parallelisation tool such as CAPTools) to be as generic as possible. CAPLib is just such a `thin layer' message passing library that supports parallel CM codes, by mapping generic calls onto machine specific libraries (such as CRAY SHMEM) and portable general purpose libraries (such as PVM an MPI). This paper describe CAPLib together with its three perceived advantages over other routes: - as a high level abstraction, it is both easy to understand (especially when generated automatically by tools) and to implement by hand, for the CM community (who are not generally parallel computing specialists); - the one parallel version of the application code is truly generic and portable; - the parallel application can readily utilise whatever message passing libraries on a given machine yield optimum performance.

Automatic relative debugging of OpenMP programs

In this work we show how automatic relative debugging can be used to find differences in computation between a correct serial program and an OpenMP parallel version of that program that does not yield correct results. Backtracking and re-execution are used to determine the first OpenMP parallel region that produces a difference in computation that may lead to an incorrect value the user has indicated. Our approach also lends itself to finding differences between parallel computations, where executing with M threads produces expected results but an N thread execution does not (M, N > 1, M ≠ N). OpenMP programs created using a parallelization tool are addressed by utilizing static analysis and directive information from the tool. Hand-parallelized programs, where OpenMP directives are inserted by the user, are addressed by performing data dependence and directive analysis.

Modelling continuum mechanics phenomena using three dimensional unstructured meshes on massively parallel processors

The difficulties encountered in implementing large scale CM codes on multiprocessor systems are now fairly well understood. Despite the claims of shared memory architecture manufacturers to provide effective parallelizing compilers, these have not proved to be adequate for large or complex programs. Significant programmer effort is usually required to achieve reasonable parallel efficiencies on significant numbers of processors. The paradigm of Single Program Multi Data (SPMD) domain decomposition with message passing, where each processor runs the same code on a subdomain of the problem, communicating through exchange of messages, has for some time been demonstrated to provide the required level of efficiency, scalability, and portability across both shared and distributed memory systems, without the need to re-author the code into a new language or even to support differing message passing implementations. Extension of the methods into three dimensions has been enabled through the engineering of PHYSICA, a framework for supporting 3D, unstructured mesh and continuum mechanics modeling. In PHYSICA, six inspectors are used. Part of the challenge for automation of parallelization is being able to prove the equivalence of inspectors so that they can be merged into as few as possible.

Automatic parallel code generation for message passing on distributed memory systems

The availability of a very accurate dependence graph for a scalar code is the basis for the automatic generation of an efficient parallel implementation. The strategy for this task which is encapsulated in a comprehensive data partitioning code generation algorithm is described. This algorithm involves the data partition, calculation of assignment ranges for partitioned arrays, addition of a comprehensive set of execution control masks, altering loop limits, addition and optimisation of communications for all data. In this context, the development and implementation of strategies to merge communications wherever possible has proved an important feature in producing efficient parallel implementations for numerical mesh based codes. The code generation strategies described here are embedded within the Computer Aided Parallelisation tools (CAPTools) software as a key part of a toolkit for automating as much as possible of the parallelisation process for mesh based numerical codes. The algorithms used enables parallelisation of real computational mechanics codes with only minor user interaction and without any prior manual customisation of the serial code to suit the parallelisation tool.

Integrating user knowledge with information from parallelisation tools to facilitate the automatic generation of efficient parallel FORTRAN code

User supplied knowledge and interaction is a vital component of a toolkit for producing high quality parallel implementations of scalar FORTRAN numerical code. In this paper we consider the necessary components that such a parallelisation toolkit should possess to provide an effective environment to identify, extract and embed user relevant user knowledge. We also examine to what extent these facilities are available in leading parallelisation tools; in particular we discuss how these issues have been addressed in the development of the user interface of the Computer Aided Parallelisation Tools (CAPTools). The CAPTools environment has been designed to enable user exploration, interaction and insertion of user knowledge to facilitate the automatic generation of very efficient parallel code. A key issue in the user's interaction is control of the volume of information so that the user is focused on only that which is needed. User control over the level and extent of information revealed at any phase is supplied using a wide variety of filters. Another issue is the way in which information is communicated. Dependence analysis and its resulting graphs involve a lot of sophisticated rather abstract concepts unlikely to be familiar to most users of parallelising tools. As such, considerable effort has been made to communicate with the user in terms that they will understand. These features, amongst others, and their use in the parallelisation process are described and their effectiveness discussed.

Parallel dynamic graph-partitioning for unstructured meshes

A parallel method for the dynamic partitioning of unstructured meshes is described. The method introduces a new iterative optimization technique known as relative gain optimization which both balances the workload and attempts to minimize the interprocessor communications overhead. Experiments on a series of adaptively refined meshes indicate that the algorithm provides partitions of an equivalent or higher quality to static partitioners (which do not reuse the existing partition) and much more rapidly. Perhaps more importantly, the algorithm results in only a small fraction of the amount of data migration compared to the static partitioners.

Converting best manual practice into generic automatable strategies for unstructured mesh parallelisation

The manual effort required to convert sequential computational mechanics programs into a useful, scalable parallel form is considerable. Tools that can assist in the conversion process are clearly required. Computer aided parallelisation tools (CAPTools) have been developed to generate efficient parallel code for real world structured grid application codes such as Computational Fluid Dynamics. Automatable single-program multi-data (SPMD) overlapping domain decomposition (DD) techniques established for structured grid codes have been adapted by the authors to manually parallelise unstructured mesh applications. Inspector loops have been used to provide generic techniques for the run-time support necessary to extend the capabilities of CAPTools to automatic implementation of SPMD DD techniques in the parallelisation of unstructured mesh codes. Copyright © 1999 John Wiley & Sons, Ltd.

An automatable generic strategy for dynamic load balancing in parallel structured mesh CFD codes

Parallel computing is now widely used in numerical simulation, particularly for application codes based on finite difference and finite element methods. A popular and successful technique employed to parallelize such codes onto large distributed memory systems is to partition the mesh into sub-domains that are then allocated to processors. The code then executes in parallel, using the SPMD methodology, with message passing for inter-processor interactions. In order to improve the parallel efficiency of an imbalanced structured mesh CFD code, a new dynamic load balancing (DLB) strategy has been developed in which the processor partition range limits of just one of the partitioned dimensions uses non-coincidental limits, as opposed to coincidental limits. The ‘local’ partition limit change allows greater flexibility in obtaining a balanced load distribution, as the workload increase, or decrease, on a processor is no longer restricted by the ‘global’ (coincidental) limit change. The automatic implementation of this generic DLB strategy within an existing parallel code is presented in this chapter, along with some preliminary results.

Computational modelling of multi-physics processes on high performance parallel computer systems

The demands of the process of engineering design, particularly for structural integrity, have exploited computational modelling techniques and software tools for decades. Frequently, the shape of structural components or assemblies is determined to optimise the flow distribution or heat transfer characteristics, and to ensure that the structural performance in service is adequate. From the perspective of computational modelling these activities are typically separated into: • fluid flow and the associated heat transfer analysis (possibly with chemical reactions), based upon Computational Fluid Dynamics (CFD) technology • structural analysis again possibly with heat transfer, based upon finite element analysis (FEA) techniques.

The implementation of an aeronautical CFD flow code onto distributed memory parallel systems

The parallelization of an industrially important in-house computational fluid dynamics (CFD) code for calculating the airflow over complex aircraft configurations using the Euler or Navier–Stokes equations is presented. The code discussed is the flow solver module of the SAUNA CFD suite. This suite uses a novel grid system that may include block-structured hexahedral or pyramidal grids, unstructured tetrahedral grids or a hybrid combination of both. To assist in the rapid convergence to a solution, a number of convergence acceleration techniques are employed including implicit residual smoothing and a multigrid full approximation storage scheme (FAS). Key features of the parallelization approach are the use of domain decomposition and encapsulated message passing to enable the execution in parallel using a single programme multiple data (SPMD) paradigm. In the case where a hybrid grid is used, a unified grid partitioning scheme is employed to define the decomposition of the mesh. The parallel code has been tested using both structured and hybrid grids on a number of different distributed memory parallel systems and is now routinely used to perform industrial scale aeronautical simulations. Copyright © 2000 John Wiley & Sons, Ltd.

Three parallel programming paradigms: Comparisons on an archetypal PDE computation

Three paradigms for distributed-memory parallel computation that free the application programmer from the details of message passing are compared for an archetypal structured scientific computation -- a nonlinear, structured-grid partial differential equation boundary value problem -- using the same algorithm on the same hardware. All of the paradigms -- parallel languages represented by the Portland Group's HPF, (semi-)automated serial-to-parallel source-to-source translation represented by CAP-Tools from the University of Greenwich, and parallel libraries represented by Argonne's PETSc -- are found to be easy to use for this problem class, and all are reasonably effective in exploiting concurrency after a short learning curve. The level of involvement required by the application programmer under any paradigm includes specification of the data partitioning, corresponding to a geometrically simple decomposition of the domain of the PDE. Programming in SPMD style for the PETSc library requires writing only the routines that discretize the PDE and its Jacobian, managing subdomain-to-processor mappings (affine global-to-local index mappings), and interfacing to library solver routines. Programming for HPF requires a complete sequential implementation of the same algorithm as a starting point, introduction of concurrency through subdomain blocking (a task similar to the index mapping), and modest experimentation with rewriting loops to elucidate to the compiler the latent concurrency. Programming with CAPTools involves feeding the same sequential implementation to the CAPTools interactive parallelization system, and guiding the source-to-source code transformation by responding to various queries about quantities knowable only at runtime. Results representative of "the state of the practice" for a scaled sequence of structured grid problems are given on three of the most important contemporary high-performance platforms: the IBM SP, the SGI Origin 2000, and the CRAYY T3E.