Researchers at Arizona State University have developed a compiler that can automatically compile a given application to execute on a scratchpad memory based multi-core architecture. A novel stack data management technique comprises a modified compiler infrastructure and optimized stack data management runtime library. The compiler can manage limitless stack data, solve invalid pointers, perform stack management with smaller footprint on local memory, and scale with the number of cores. Benchmark results ran on an IBM Cell processor show overhead of only 3.23% in performance.
- Chip Manufacturers – reduce programming overhead of implementing scratchpad memory based multi-core processors.
- Application Developers – create motivation to efficiently utilize the multitude of multi-core processors available in the market.
Benefits and Advantages
- Automated – compiler can automatically compile applications for scratchpad memory on multi-core architecture.
- Unlimited Stack Data in Local Memory – the compiler can request dynamic memory allocation in the global memory and support unlimited stack data in the local memory.
- Better Stack Pointer Management – pointer problems caused by pointing to other stack frames in global memory will no longer become an issue.
- Reduced Number of Direct Memory Access (DMA) – stack frames are managed at a coarser granularity.
- Improved Performance – stack data is managed in linear fashion, which reduces the management overhead on maintaining memory status if memory is fragmented by circular fashion. Therefore, our management overhead is only 3% on average.
For more information about the inventor(s) and their research, please see
Dr. Aviral Shrivastava's directory webpage