Presented annually to the author(s) of a paper presented at the PLDI held 10 years prior to the award year. The award includes a prize of $1,000 to be split among the authors of the winning paper. The papers are judged by their influence over the past decade.
Selection Committee The award given in year N is for the most influential paper presented at the conference held in year N-10. The selection committee consists of the following members:
The SIGPLAN Chair shall adjudicate conflicts of interest, appointing substitutes to the committee as necessary.
This paper introduced Pin, a dynamic binary instrumentation framework that enables the creation of dynamic program analysis tools. Pin uses dynamic compilation to instrument executables and dynamically-linked libraries while they are running, permitting the tool writer to study the behavior of an application at the instruction level without significant perturbation to application behavior. The PLDI 2005 paper is highly cited and the system it describes is in widespread use in academia and industry. Pin’s ease of use and relative efficiency have made it the tool of choice for dynamic binary instrumentation.
(for 2004) Scalable Lock-Free Dynamic Memory Allocation
Maged Michael’s PLDI’04 paper is considered a landmark in memory allocation for multithreaded programs, presenting the first completely lock-free general-purpose user-space allocator. It provides good performance with respect to scalability, speed and space efficiency, while at the same time only relying on common hardware and OS support. The work is highly regarded and frequently referenced, and is also the basis of multiple memory allocator implementations, both in IBM products and in follow-on research.
This is the first publication to describe the design, implementation, and optimization of the language nesC, which is a variant of the C programming language especially well suited to the programming of embedded systems. Much as the C programming language made possible the development of UNIX, so has nesC made possible the development of TinyOS. A decade hence, nesC is still the language of choice for developing applications under TinyOS on “motes” and other embedded systems platforms. Central to the design of nesC and TinyOS is the notion of an asynchronous, or non-blocking, call. This form of software architecture nicely accommodates the timing uncertainties and failures experienced by embedded systems applications in the real world. The success and longevity of nesC can be attributed to its precise and useful formulations of concurrency, atomicity, and modularization with regard to its intended application audience. These formulations allow for efficient implementations while offering encapsulation mechanisms that encourage software sharing and reuse. The nesC paper has amassed hundreds of citations, and the language continues to be in widespread use for teaching, research, and industry in the embedded systems community.
(for 2002) Extended Static Checking for Java
This paper marks a turning point in the field of static checking, describing pragmatic design decisions that promote practicality over completeness. Pioneered in ESC/Modula-3, techniques from ESC/Java are now widely used in various forms in Microsoft’s development tools, notably as part of Code Contracts which ships with VisualStudio. Recent innovations strongly influenced by ESC/Java include refinement types for Haskell, and verification of Eiffel programs.
The paper, “Automatic Predicate Abstraction of C Programs” by Thomas Ball, Rupak Majumdar, Todd D. Millstein, and Sriram K. Rajamani presented the underlying predicate abstraction technology of the SLAM project for checking that software satisfies critical behavioral properties of the interfaces it uses and to aid software engineers in designing interfaces and software that ensure reliable and correct execution. The technology is now part of Microsoft’s Static Driver Verifier in the Windows Driver Development Kit. This is one of the earliest examples of automation of software verification on a large scale and the basis for numerous efforts to expand the domains that can be verified.
Dynamo pioneered the technique of monitoring, analyzing, and optimizing binary code on-the-fly while the program executes, without relying on any program modifications, compiler hints, profile data from prior runs, or special purpose hardware. Contrary to intuition, one could use Dynamo to substantially improve the performance of a binary’s execution, even when it was generated by a state-of-the-art optimizing compiler. By continuously monitoring its own overhead, Dynamo could suspend itself (and resume later) when the optimization gains were not sufficient to offset the cost of its own operation. The ability for a software-only system to speed up a program binary without any kind of modification or externally provided hints challenged much entrenched thinking at the time, and catalyzed a rethinking of the compiler-architecture interface. Hardware designs were shifting more of the performance optimization burden from the hardware to the compiler. At the same time software was moving towards greater use of dynamic binding, resulting in a shrinking optimization scope for the compiler. These two trends were in tension with one another. By operating at binary execution time, Dynamo helped bridge this growing gap by complementing the compiler’s traditional strength in static control flow based optimization with instruction-level runtime trace-based optimization. This paper was the first in a series of publications that continue to this day on similar trace-based systems for dynamic binary optimization, dynamic binary instrumentation, and dynamic binary translation.
(for 1999) A Fast Fourier Transform Compiler
The 1999 PLDI paper “A Fast Fourier Transform Compiler” by Matteo Frigo describes the implementation of genfft, a special-purpose compiler that produces the performance critical code for a library, called FFTW (the “Fastest Fourier Transform in the West”), that computes the discrete Fourier transform. FFTW is the predominant open fast Fourier transform package available today, as it has been since its introduction a decade ago. genfft demonstrated the power of domain-specific compilation – FFTW achieves the best or close to best performance on most machines, which is remarkable for a single package. By encapsulating expert knowledge from the FFT algorithm domain and the compiler domain, genfft and FFTW provide a tremendous service to the scientific and technical community by making highly efficient FFTs available to everyone on any machine. As well as being the fastest FFT in the West, FFTW may be the last FFT in the West as the quality of this package and the maturity of the field may mean that it will never be superseded, at least for computer architectures similar to past and current ones.
The 1998 PLDI paper “Implementation of the Cilk-5 Multithreaded Language” by Matteo Frigo, Charles E. Leiserson, and Keith H. Randall introduced an efficient form of thread-local deques to control scheduling of multithreaded programs. This innovation not only opened the way to faster and simpler runtimes for fine-grained parallelism, but also provided a basis for simpler parallel recursive programming techniques that elegantly extend those of sequential programming. The stack-like side of a deque acts just like a standard procedure stack, while the queue side enables breadth-first work-stealing by other threads. The work-stealing techniques introduced in this paper are beginning to influence common practice, such as the Intel Threading Building Blocks project, an upcoming standardized fork-join framework for Java, and a variety of projects at Microsoft.
(for 1992) Lazy Code Motion
(for 1991) A data locality optimizing algorithm
(for 1990) Profile guided code positioning