Dynamic storage allocation continues to play an important role in the performance and correctness of systems ranging from user productivity software to high-performance servers. While algorithms for dynamic storage allocation have been studied for decades, much of the literature is based on measuring the performance of benchmark programs unrepresentative of many important allocation-intensive workloads. Furthermore, to date no standard has emerged or been proposed for publishing and exchanging representative allocation workloads. In this paper, we describe a preliminary design of a trace format for such workloads and investigate its e#ectiveness at representing large allocation traces. Our proposal allows for a flexible encoding of information in the trace to achieve greater compression. We evaluate our preliminary design in two dimensions. First, we measure how e#ective these encodings are at reducing trace size. Second we consider how a meta-level specification language could be used...

We have measured the energy efficiency of different memory management strategies on a high performance pocket computer. We conducted our study by measuring the energy consumption of eight C programs with four different memory management strategies each. The memory management strategies are: no deallocation, explicit deallocation, conservative mark-and-sweep garbage collection, and conservative mark-and-sweep incremental garbage collection. Our measurements show that different memory management strategies have very different energy requirements. In the most extreme case, one program consumed 40 times as much energy with incremental garbage collection than with explicit deallocation. We demonstrate that, although overall energy use is strongly correlated with execution time, the processor and peripheral energies separately do not correlate well with execution time.

We study the cost of storage management for garbage-collected programs compiled with the Standard ML of New Jersey compiler. We show that the cost of storage management is not the same as the time spent garbage collecting. For many of the programs, the time spent garbage collecting is less than the time spent doing other storage-management tasks. 1 Authors' addresses: David Tarditi, Computer Science Department, Carnegie Mellon University, 5000 Forbes Avenue, Pittsburgh, PA, 152133891. e-mail: dtarditi@cs.cmu.edu. Amer Diwan, Department of Computer Science, University of Massachusetts, Amherst, MA 01003-4610. e-mail: diwan@cs.umass.edu. This research is sponsored by the Defense Advanced Research Projects Agency, DoD, through ARPA Order 8313, and monitored by ESD/AVS under contract F19628-91-C-0168. David Tarditi is also supported by an AT&T PhD Scholarship. Views and conclusions contained in this document are those of the authors and should not be interpreted as representing the offic...

. This paper is concerned with on-line storage allocation to processes in a dynamic environment. This problem has been extensively studied in the past. We provide a new, tighter bound, for the competitive ratio of the well known First Fit algorithm. This bound is obtained by considering a new parameter, namely the maximum number of concurrent active processes. We observe that this bound is also a lower bound on the competitive ratio of any deterministic on-line algorithm. Our second contribution is an on-line allocation algorithm which uses coloring techniques. We show that the competitive ratio of this algorithm is the same as that of First Fit. Furthermore, we indicate that this algorithm may be advantageous in certain applications. Our third contribution is to analyze the performance of randomized algorithms for this problem. We obtain lower bounds on the competitive ratio which are close to the best deterministic upper bounds.

We describe the rationale behind the design of key features of Titanium—an explicitly parallel dialect of Java for high-performance scientific programming—and our experiences in building applications with the language. Specifically, we address Titanium’s partitioned global address space model, single program multiple data parallelism support, multi-dimensional arrays and array-index calculus, memory management, immutable classes (class-like types that are value types rather than reference types), operator overloading, and generic programming. We provide an overview of the Titanium compiler implementation, covering various parallel analyses and optimizations, Titanium runtime technology and the GASNet network communication layer. We summarize results and lessons learned from

This paper presents the performance evaluation of five memory allocation strategies for the PerDiS Persistent Distributed Object store in the context of allocating two different web sites in the store. The evaluation was conducted using (i) a web gathering tool, to log the web objects graph, and (ii) a PerDiS memory simulator that implements the different allocation strategies. Our results show that for all the strategies and parameters we have evaluated, reference and cycle locality are quiet poor. The best policy seems to be first sequential fits. Results are linear with the size of the garbage collection Unit (a bunch). There is no clear optimum, but 64K to 128K appear to be good choices.

This paper describes an incremental conservative garbage collector that significantly reduces pause time of an existing collector by Boehm et al. Like their collector, it is a true conservative collector that does not require compiler cooperation but uses virtual memory primitives (page protection) of operating systems for write barriers. While much successful work has been done on incremental collectors in general, achieving small pause time of the order of a few milliseconds in such uncooperative settings has been challenging. Our collector combines several ideas that bound pause times without introducing significant overheads. They in- clude: (1) bounding the number of dirty (writable) pages during concurrent marking, (2) adaptive repetition of concurrent marking phases, and (3) allocating objects in black (marked) in later stages of a collection. With these tech- niques, we have achieved the maximum pause time of 2.6 4.5ms for five application benchmarks on 400MHz UltraSPARC processors. They are up to forty times shorter than the basic incremental collector similar to Boehm et al.'s. The overhead (total work time) of our collector is 1.2 53% to the stop-the-world collector and 9% or less to the basic incremental collector.

. OSUIF is an extension to SUIF 2.0 that provides support for the compilation of object-oriented languages. OSUIF extends standard SUIF in three main areas: symbol table, intermediate language, and exception handling. The resulting system should be able to support compilers for many (but not all) object-oriented languages. The two initial OSUIF front ends will support C++ and Java. 1. Introduction The original SUIF system [SCG94] is a research compiler for procedural languages like C and Fortran. While these languages cover a large spectrum of compiler research, many software developers (including the SUIF developers) use object-oriented languages today. As a result, compilation of object-oriented languages has become a relevant research area. Object-oriented languages like C++ and Java are quite different from procedural languages. The most striking language difference may be the support for polymorphism that allows pieces of code to be reused with objects of different types. This p...

Garbage collection (GC) plays a more and more dominant role in the landscape of modern computer science languages. Coupled with a proper object oriented language GC can help the developer to focus on the true problem domain, leaving the details of memory deallocation to a system cleanup process. However, the transition to a GC collected language might introduce new areas of problems, both in terms of application performance and scalability. This article will aim at describing these positive and negative effects of using a GC language. It will also illuminate the strong and weak side of both GC and non-GC language in such a way that a developer will have a clear understanding of the two alternatives.

This paper evaluates six different types of interprocedural pointer analyses on 22 telecommunication and media applications and describes their application to an SRAM power reduction technique. This configurable SRAM provides differentiation of data access time and port counts within a single on-chip structure. Scheduling for configurable SRAM relies on inter-procedural dependence analysis for safe assignment of program data objects to on-chip storage regions with little or no performance degradation. It thus provides a meaningful vehicle for exploring the applicability of and need for various interprocedural analysis techniques, as well as a demonstration of how compilation and hardware techniques can be combined to achieve optimization beyond that for performance. 1.