Cray Research T3D Research Articles

Performance Modeling and Evaluation of MPI

Users of parallel machines need to have a good grasp for how different communication patterns and styles affect the performance of message-passing applications. LogGP is a simple performance model that reflects the most important parameters required to estimate the communication performance of parallel computers. The message passing interface (MPI) standard provides new opportunities for developing high performance parallel and distributed applications. In this paper, we use LogGP as a conceptual framework for evaluating the performance of MPI communications on three platforms: Cray-Research T3D, Convex Exemplar 1600SP, and a network of workstations (NOW). We develop a simple set of communication benchmarks to extract the LogGP parameters. Our objective in this is to compare the performance of MPI communication on several platforms and to identify a performance model suitable for MPI performance characterization. In particular, two problems are addressed: how LogGP quantifies MPI performance and what extra features are required for modeling MPI, and how MPI performance compare on the three computing platforms: Cray Research T3D, Convex Exemplar 1600SP, and workstations clusters.

Falcon: A Production Quality Distributed Memory Reservoir Simulator

Summary We describe a new production model, Falcon, that has achieved speeds on parallel computers that are 100 times faster on real world problems than current production models on a vector computer. Falcon has been used to conduct the largest, geostatistical reservoir study ever conducted within Amoco. In this paper we discuss the following: Falcon's data parallel paradigm with FORTRAN 90 and high performance FORTRAN (HPF); its single program, multiple data (SPMD) paradigm with message passing; efficient memory management that enables simulation of enormous studies; a numerical formulation that reconciles the generalized compositional approach (based on component masses and pressure) with earlier approaches (based on pressures and saturations), in a more general and more efficient approach. We also discuss Falcon's scalability up to 512 processor nodes and performance (timings and memory) achieved on a number of parallel platforms, including Cray Research's T3D and T3E, SGI's Power Challenge and Origin 2000, Thinking Machines' CM5, and IBM's SP2. Falcon also runs on single processor computers such as PC's and IBM's RS6000. We discuss a new parallel linear solver technology based on a fully parallel scalable implementation of incomplete lower-upper (ILU) preconditioning coupled with a GMRES or Orthomin iteration process. This naturally ordered global ILU preconditioner is scalable to hundreds of processors, efficiently solving the matrix problems arising from large scale simulations The use of the techniques described in this paper has enabled us to run problem sizes of up to 16.5 million gridblocks. Falcon was used to simulate fifty geostatistically derived realizations of a large, black oil waterflood system. The realizations, each with 2.3 million cells and 1,039 wells, took an average of 4.2 hours to execute on a 128-node CM5 computer, thus enabling the simulation study to finish in less than a month. In this field study, we bypassed upscaling through the use of fine vertical resolution gridding. Our focus has been on the applicability of Falcon to real world problems. Falcon can be used for modeling both small and very large reservoirs, including reservoirs characterized by geostatistics. It can be used to simulate black oil, gas/water, and dry gas reservoirs. And, a fully compositional feature is being developed.

A new deterministic parallel sorting algorithm with an experimental evaluation

We introduce a new deterministic parallel sorting algorithm for distributed memory machines based on the regular sampling approach. The algorithm uses only two rounds of regular all-to-all personalized communication in a scheme that yields very good load balancing with virtually no overhead. Moreover, unlike previous variations, our algorithm efficiently handles the presence of duplicate values without the overhead of tagging each element with a unique identifier. This algorithm was implemented in SPLIT-C and run on a variety of platforms, including the Thinking Machines CM-5, the IBM SP-2-WN, and the Cray Research T3D. We ran our code using widely different benchmarks to examine the dependence of our algorithm on the input distribution. Our experimental results illustrate the efficiency and scalability of our algorithm across different platforms. In fact, the performance compares closely to that of our random sample sort algorithm, which seems to outperform all similar algorithms known to the authors on these platforms. Together, their performance is nearly invariant over the set of input distributions, unlike previous efficient algorithms. However, unlike our randomized sorting algorithm, the performance and memory requirements of our regular sorting algorithm can be deterministically guaranteed.

A Randomized Parallel Sorting Algorithm with an Experimental Study

Previous schemes for sorting on general-purpose parallel machines have had to choose between poor load balancing and irregular communication or multiple rounds of all-to-all personalized communication. In this paper, we introduce a novel variation on sample sort which uses only two rounds of regular all-to-all personalized communication in a scheme that yields very good load balancing with virtually no overhead. Moreover, unlike previous variations, our algorithm efficiently handles the presence of duplicate values without the overhead of tagging each element with a unique identifier. This algorithm was implemented in Split-C and run on a variety of platforms, including the Thinking Machines CM-5, the IBM SP-2, and the Cray Research T3D. We ran our code using widely different benchmarks to examine the dependence of our algorithm on the input distribution. Our experimental results illustrate the efficiency and scalability of our algorithm across different platforms. In fact, it seems to outperform all similar algorithms known to the authors on these platforms, and its performance is invariant over the set of input distributions unlike previous efficient algorithms. Our results also compare favorably with those reported for the simpler ranking problem posed by the NAS Integer Sorting (IS) Benchmark.

Practical parallel algorithms for personalized communication and integer sorting

A fundamental challenge for parallel computing is to obtain high-level, architecture independent, algorithms which efficiently execute on general-purpose parallel machines. With the emergence of message passing standards such as MPI, it has become easier to design efficient and portable parallel algorithms by making use of these communication primitives. While existing primitives allow an assortment of collective communication routines, they do not handle an important communication event when most or all processors have non-uniformly sized personalized messages to exchange with each other. We focus in this paper on the h-relation personalized communication whose efficient implementation will allow high performance implementations of a large class of algorithms. While most previous h-relation algorithms use randomization, this paper presents a new deterministic approach for h-relation personalized communication with asymptotically optimal complexity for h>p 2 . As an application, we present an efficient algorithm for stable integer sorting. The algorithms presented in this paper have been coded in Split-C and run on a variety of platforms, including the Thinking Machines CM-5, IBM SP-1 and SP-2, Cray Research T3D, Meiko Scientific CS-2, and the Intel Paragon. Our experimental results are consistent with the theoretical analysis and illustrate the scalability and efficiency of our algorithms across different platforms. In fact, they seem to outperform all similar algorithms known to the authors on these platforms.

Parallel algorithms for image enhancement and segmentation by region growing, with an experimental study

This paper presents efficient and portable implementations of a powerful image enhancement process, the Symmetric Neighborhood Filter (SNF), and an image segmentation technique that makes use of the SNF and a variant of the conventional connected components algorithm which we call δ-Connected Components. We use efficient techniques for distributing and coalescing data as well as efficient combinations of task and data parallelism. The image segmentation algorithm makes use of an efficient connected components algorithm based on a novel approach for parallel merging. The algorithms have been coded in Split-C and run on a variety of platforms, including the Thinking Machines CM-5, IBM SP-1 and SP-2, Cray Research T3D, Meiko Scientific CS-2, Intel Paragon, and workstation clusters. Our experimental results are consistent with the theoretical analysis (and provide the best known execution times for segmentation, even when compared with machine-specific implementations). Our test data include difficult images from the Landsat Thematic Mapper (TM) satellite data.

Distributed data access in AC

We have modified the C language to support a programming model based on a shared address space with physically distributed memory. With this model users can write programs in which the nodes of a massively parallel processor can access remote memory without message passing. AC provides support for distributed arrays as well as pointers to distributed data. Simple array references and pointer dereferencing are sufficient to generate low-overhead remote reads and writes. We have implemented these ideas in a compiler based on the GNU C compiler and targeted at Cray Research's T3D. Initial performance measurements show that AC generates code for remote accesses which is considerably faster than that of the native compiler for structures up to about 16 words in size and virtually equivalent for larger transfers.

A TECHNICAL DESCRIPTION OF SOME PARALLEL COMPUTERS

An important part of parallel computations, for any group interested in this field, is the purchase of one of the available machines. This new machine becomes the major platform for the group’s efforts in algorithmic development. This paper is an overview of many of the currently available parallel computers. Its original intent was to help in the review, evaluation phase of a machine purchase. The current machine offerings, Cray Research T3D, Fujitsu AP1000 and VPP500, IBM SP1, Intel Paragon, Kendall Square Research KSR1, nCUBE 2 and 3, NEC Cenju-3, Parsytec GigaCube-2, Performance Computing Industries (Meiko) CS-2, Tera, and Thinking Machines Corporation CM-5 are discussed, with emphasis on the unique features and problems of each particular machine and/or company. In addition to the text numerous tables are included to give a overview of the hardware capabilities, the software available and benchmark performance (where available) of each of the computers.