|
The Gluon Project
People:
Li-Ling Chen, Roy Ju, Shih-wei Liao, Tin-fook Ngai, Youfeng Wu
Purpose:
GLUON is an IA-64 research compiler with a flexible infrastructure and a set of powerful supporting components allowing researchers to quickly prototype new compiler technologies for existing and future architectures. We work closely with Architecture/Circuit Labs to drive IA-64 micro-architecture research and enhance Intel's leadership in processors and platforms.
Our compiler incorporates state-of-the-art IA-64 code optimization and generation technologies, a flexible machine model, a set of powerful edge/path/data/value profilers and annotation tools.
Description:
We currently focus on two research projects:
|
 | Speculative multithreading | |
| Memory optimizations | |
Speculative Multithreading
The Speculative Multithreading Project investigates the new concept of speculative multithread execution and develops the corresponding compiler technologies. It studies the behavior of speculative multithreaded programs, assesses their performance potentials and develops new program analysis, code generation and optimization algorithms. The existing Gluon compiler is being enhanced with the compilation model and supports for speculative multithread execution.
Memory Optimizations
The Memory Optimization Project focuses on memory sub-system of IA-64 and develops the corresponding optimization technologies. It studies memory-intensive benchmarks, analyzes the memory behavior of programs, and develops compiler transformations or new micro-architecture features to improve the performance of memory sub-systems. The GLUON compiler is used to prototype emerging techniques for improving reference localities and hiding load latencies.
People:
Weldon Washburn, Rick Hudson, Michal Cierniak, Sreenivas Subramoney, Ken Lueh, Ali Adl-tabatabai, Tatiana Shpeisman, Bratin Saha, Anwar Ghuloum, Vijay Menon, Brian R. Murphy, Jayashankar Bharadwaj, Brian T. Lewis, Marsha Eng, Neal (Arthur) Glew
Purpose:
The Open Source Dynamic Computing Research Platform is a platform for bytecode system research. To date, an open source research platform that allows interchangeable modules implementing Just-In-Time (JIT) compilers and Garbage Collection (GC) has not been available to the academic research community. ORP (Open Runtime Platform) is an open source research infrastructure project that provides these features, thus enabling the rapid evolution of systems research in dynamic compilation and memory management.
Description:
ORP implements advanced JIT and GC interfaces. These interfaces allow the JIT, GC and remaining runtime systems to be developed in complete isolation and at independent rates. A major benefit is that JIT and GC researchers are relieved of the intractable task of developing a full system before they can innovate in their area of expertise. Initially ORP interoperates with GNU Classpath. GNU Classpath is an open source library for the Java* language. Current ORP source code compiles and runs on IA-32 Linux and IA-32 Windows 2000 operating systems. Future updates will include JIT and GC for IA-64. GNU Make makefiles are included for Linux builds. Microsoft VC++ 5.0 project files are included for the Windows* 2000 build.
Application Performance
The basic system incorporates a fast code generating JIT, as well as an optimizing JIT. It also includes several GC algorithms, ranging from a simple mark-sweep algorithm to an advanced train algorithm.
MRL academic papers on these topics are listed at Publications.
People:
Dong-Yuan Chen
Purpose:
Profile-guided optimizations used in static compilers have contributed to significant performance improvement on modern microprocessors. This project focuses on exploiting the potential of continuously monitoring or collecting the execution behavior profiles using hardware performance monitoring unit (PMU) and using the profiles to guide optimizations in dynamic compilation or re-optimization environments, such as Java runtime or dynamic binary re-optimization system. The areas of research that we are interested in included:
|
| Light-weight continuous profiling using PMU | |
| New PMU events or support for profile collection | |
| Dynamic compilation or re-optimization infrastructure | |
| PMU profile-guided optimizations | |
| Architecture or microarchitecture support for dynamic optimization systems | |
| Dynamic behavior of real-world applications | |
People:
Kai Yu Chen, William Chen, Buqi Cheng, Roy Ju, Qingyu Zhao
Purpose:
The Open Research Compiler (ORC) is an Itanium® Processor Family (IPF) (IA-64) open source research compiler available to research and open source communities for conducting compiler and architecture research. It is offered as part of Intel's initiative to lead research on instruction-level parallelism, thread-level parallelism and beyond. ORC incorporates advanced program analysis and optimization techniques, including sophisticated uses of IPF architectural features. It provides a robust, flexible research compiler with competitive performance to enable researchers to explore and investigate advanced research problems.
ORC allows research groups around the world to focus on studying advanced research issues without substantial up-front costs to build a compiler infrastructure. The compiler research efforts at the Microprocessor Research Labs (MRL) to study advanced compilation and architectural features on Intel Architectures are based on ORC. Intel continues to lead in microprocessors and high-performance computing.
Description:
To leverage well-developed technology, ORC is built upon an existing open source compiler, Pro64. ORC 1.0 has been released as open source since Jan 2002. The first release focused on the development of advanced optimizations to effectively exploit IPF features, such as predication, speculation, and explicit parallelism. It also provided the research infrastructure support to explore different design choices of compilers and hardware, such as controlling optimization scope using program regions, parameterizing machine descriptions, and supporting various types of profiling feedback. We are actively working on additions to ORC, to include further exploitation of IPF features and revamp key parts of machine-independent optimizations, such as interprocedural analysis and optimization, to achieve competitive execution performance. A number of important research problems have been targeted on ORC, such as the compiler and architectural support for various forms of multi-threading. Optimizations for memory hierarchy is also an important research topic. Additional releases of ORC have been planned for the year of 2002.
We would like to invite researchers around the world to collaborate on ORC-based emerging research issues, such as integrated static and dynamic compilation frameworks, software-driven power management, compilation for specific domains of applications, code size, program analysis and performance tools, etc.
Related Links:
| 
