Intel on Thursday announced a partnership with the Institute of Computing Technology of the Chinese Academy of Sciences (CAS) in Beijing to conduct advanced software compiler research.
The focus of the two-year research agreement will be on improving the general performance and efficiency of Intel's Itanium processor, the company said in a statement. Compilers take software written in high-level programming languages, such as C and C++, and convert them into machine code that can be understood by microprocessors.
"This is not only a symbol, this is a real agreement," said Dr. Jesse Fang, director of the Programming System Lab at Intel's Microprocessor Research Lab, in a phone interview. The aim of the partnership is to provide technology that allows researchers around the world to develop new compiler technology, he said.
Research work to be carried out under the agreement will focus on the development of modular compiler components that allow researchers to test different optimization techniques, Intel said. Optimization techniques to be tested in this way will include loop nesting, inter-procedural analysis, scalar optimization, global instruction scheduling, profiling, software pipeline, register allocation, predication and speculation. In addition, a number of simulator tools will be developed to enable validation and comparison of these modules, it said.
Research will initially focus on compilers for the C, C++ and Fortran programming languages, Fang said. However, nothing precludes this from expanding to include Java and other languages, he said.
Likewise, while the compiler research initially will focus on Intel's 64-bit processor architecture, IA-64, the infrastructure that Intel and CAS build will be flexible enough to cover other architectures in the future, Fang said.
"IA-64 is the first step," he said.
Fang cited the extensive experience that CAS researchers have had working with compilers as one reason Intel entered into the research partnership.
"The CAS group has over 20 years of experience on compiler research," Fang said, adding that while CAS researchers may not be familiar with the most recent microarchitecture features, they have very good software engineering skills.
"They have a quite good track record," said Fang.