In this paper we present a parallel implementation of the Bowyer-Watson (BW) algorithm using the task-parallel programming model. The BW algorithm constitutes an ideal mesh refinement strategy for implementing a large class of unstructured mesh generation techniques on both sequential and parallel computers, by preventing the need for global mesh refinement. Its implementation on distributed memory multicomputes using the traditional data-parallel model has been proven very inefficient due to excessive synchronization needed among processors. In this paper we demonstrate that with the task-parallel model we can tolerate synchronization costs inherent to data-parallel methods by exploring concurrency in the processor level. Our preliminary performance data indicate that the taskparallel approach: (i) is almost four times faster than the existing data-parallel methods, (ii) scales linearly, and (iii) introduces minimum overheads compared to the "best" sequential implementation of the BW...

In this paper we present a parallel runtime substrate, the Mobile Object Layer (MOL), that supports data or object mobility and automatic message forwarding in order to ease the implementation of adaptive and irregular applications on distributed memory machines. The MOL uses global logical name space for message passing and distributed directories to assist in the translation of logical to physical addresses. The latency of the MOL primitives is within 10% to 14% of the the latency of the underlying communication substrate. The MOL is a light-weight, portable library designed to minimize maintenance costs for very large-scale parallel adaptive applications. Keywords: Parallel, message passing, load balancing, runtime software, adaptive mesh generation. 1

Data locality is critical to good performance on the distributed memory machines, so load balancing in dynamic, irregular applications requires the movement of both data and computation. The Mobile Object Layer (MOL) is a light-weight, portable library built on top of one-sided communication systems that handles the message forwarding between processors in the presence of object migration. In this paper we present: (1) The interface to MOL, as well as a detailed description of the implementation and a proof of the correctness of MOL's forwarding protocol, (2) preliminary performance data that demonstrate MOL's e ciency, and (3) a preliminary evaluation of the MOL in the context of an adaptive application.

Parallel mesh generation is a relatively new research area between the boundaries of two scientific computing disciplines: computational geometry and parallel computing. In this chapter we present a survey of parallel unstructured mesh generation methods. Parallel mesh generation methods decompose the original mesh generation problem into smaller subproblems which are meshed in parallel. We organize the parallel mesh generation methods in terms of two basic attributes: (1) the sequential technique used for meshing the individual subproblems and (2) the degree of coupling between the subproblems. This survey shows that without compromising in the stability of parallel mesh generation methods it is possible to develop parallel meshing software using off-the-shelf sequential meshing codes. However, more research is required for the efficient use of the state-of-the-art codes which can scale from emerging chip multiprocessors (CMPs) to clusters built from CMPs. 2

In this paper we present a parallel runtime substrate that supports a global addressing scheme, object mobility, and automatic message forwarding required for the implementation of adaptive applications on distributed memory machines. Our approach is application driven; the target applications are characterized by very large variations in time and length scales. Preliminary performance data from parallel unstructured adaptive mesh refinement on an SP2 suggest that the flexibility and general nature of the approach we follow does not cause undue overhead. To appear in Proc. of the 2 nd Intl. Symposium on Computing in Object-oriented Parallel Environments, Dec. 8-11, 1998, Santa Fe, NM, as part of the Lecture Notes in Computer Science Series, Springer-Verlag. y This work supported by NSF grant #9726388 and JPL award #961097 z Supported by the Arthur J. Schmidt Fellowship and by NSF grant #9726388 x Supported by NSF grant #9726388 -- Chris Hawblitzel was supported by an N...