Evaluating the XMT parallel programming model

Abstract

Explicit-multithreading (XMT) is a parallel programming model designed for exploiting on-chip parallelism. Its features include a simple thread execution model and an efficient prefix-sum instruction for synchronizing shared data accesses. By taking advantage of low-overhead parallel threads and high on-chip memory bandwidth, the XMT model tries to reduce the burden on programmers by obviating the need for explicit task assignment and thread coarsening. This paper presents features of the XMT programming model, and evaluates their utility through experiments on a prototype XMT compiler and architecture simulator. We find the lack of explicit task assignment has slight effects on performance for the XMT architecture. Despite low thread overhead, thread coarsening is still necessary to some extent, but can usually be automatically applied by the XMT compiler. The prefix-sum instruction provides more scalable synchronization than traditional locks, and the simple run-until-completion thread execution model (no busywaits) does not impair performance. Finally, the combination of features in XMT can encourage simpler parallel algorithms that may be more efficient than more traditional complex approaches.

Towards a First Vertical Prototyping of an Extremely Fine-Grained Parallel Programming Approach

“…Our experiments show that the asynchronous algorithm is outperformed by the synchronous one [14] due to the amount of storage and extra work that it involves. Asynchronous algorithms are more useful for irregular computations, as we show next.…”

“…In this section we discuss how these features are manifested in the programming and performance of two different types of computation domains -regular computations, and irregular dynamic computations. (Other types of computations, such as divide-and-conquer and sorting algorithms, are discussed in [14]). …”

“…The introduction of an XMT compiler, presented here, allows us to evaluate XMT for the first time as a complete environment ("vertical prototyping"), using a much larger benchmark suite (with longer codes) than before. Due to the rather broad nature of our work, specialized parts of the work -the evaluation of the XMT compiler and evaluation of the programming model -were published in two respective specialized workshops [13], [14]. This paper incorporates the feedback from these workshops, and is the first one to present the whole work, including the integrated results, as well as the interplay between the programming model and the other components of XMT (compiler and architecture).…”