Rice University researchers have received a $16 million grant from the Defense Advanced Research Projects Agency to develop a universal compiler that can run on heterogeneous hardware and multicore platforms, which are found in supercomputers and embedded systems, such as those used in routers and game consoles. If the effort is successful it could result in software that takes advantage of the underlying hardware to create faster and more energy-efficient computers and devices. As devices evolve, they contain more semiconductors and use more power, and writing code that takes full advantage of the hardware becomes more difficult. Programmers can either customize code running on the heterogeneous hardware or use a compiler that can translate the code into the 0s and 1s the specific type of chip can process. However, developing custom code is time consuming, expensive, and can only be applied to one specific device, so most chipmakers or outside vendors develop compilers for each chip, which can take between three to five years, but can be used with the chip in a variety of devices. The Rice-led project hopes to build a universal compiler that will improve software performance by developing a software suite that maps out the limitations of and capabilities of the hardware it is running on; creating a planned compiler to examine the source code of the application and attempting to automatically partition the source code to run on multicore processors; and creating a runtime tool that measures the performance of the application on the system, and possibly even changes the code on the fly if the software is written for an x86 processor. Rice University professor Keith Cooper says the researchers will have 54 months to achieve this objective.
For more information please visit: http://www.cpccci.com
