Intel and Chinese Academy Of Sciences Deliver Advanced Compiler Framework

"Less than one year into their research collaboration, Intel Labs and the Chinese Academy of Sciences have delivered an advanced compiler framework for academic and corporate use in support of cutting-edge microarchitecture research," said Justin Rattner, Intel Fellow and director of Microprocessor Research at Intel Labs. "This is a strong demonstration of our enhanced relationship with the Chinese Academy of Sciences and our continued commitment to engage the top Chinese computer scientists in the global research community." Intel Labs and the Institute of Computing Technology (ICT) of the Chinese Academy of Sciences teamed up in April of 2001 for a two-year research project to develop the next generation of compilers. Mr. Li Guojie, director of ICT, CAS, praised the milestone, saying, "The Itanium processor family architecture compiler framework will provide a robust and flexible research platform for global compiler research community. It will also help China's research and development in processor and compiler technologies." Compilers are like translators; they translate programming languages like C, C++ and Fortran, and turn them into machine language (the language computers understand). The more efficiently this translation can be done, the more performance a computer can provide. Modern processors like those in the Itanium processor family have multiple instruction units and can perform multiple tasks in parallel. Modern compilers must be able to identify parallelism in the source code and assign tasks to keep all the instruction units busy. As microprocessor architecture advances, it is important to move compiler techniques along in parallel. The compiler framework released today uses robust, modular compiler components that permit experimentation. Each "module" performs a specific optimization task. This allows researchers to quickly test different techniques against the basic compiler framework. The optimization techniques tested in this way include global instruction scheduling, predication and speculation, and software pipelining, among others. The compiler framework also comes with a rich set of infrastructure additions such as profiling and region formation. The release of the advanced compiler framework was announced at the 34th Annual International Symposium on Microarchitecture (MICRO-34 2001) in December of 2001. The entire compiler source code will be made available from http://ipf-orc.sourceforge.net.* Intel also offers fully supported commercial compiler products for Windows and Linux. More information on those products, including evaluation versions, is available at www.intel.com/software/products. The Institute of Computing Technology team is led by Professor Zhaoqing Zhang, who has been a leading researcher in high-performance compilers for more than 10 years working on both academic and industrial projects. The team from Intel Labs is lead by Dr. Jesse Fang, who has also worked in research and development of high performance compilers for more than 10 years and has been involved in building optimizing compilers for multiple generations of microprocessor. For more information visit www.intel.com/research/university/index.htm