CPU Performance Bottlenecks Limit Parallel Processing Speedups


Multi-core processors theoretically can run many threads of code in parallel, but some categories of operation currently bog down attempts to raise overall performance by parallelizing computing. Is it time to have accelerators for running highly parallel code? Standard processors have many CPUs, so it follows that cache coherency and synchronization can involve thousands of cycles of low-le... » read more

System Bits: Nov. 8


Optimizing multiprocessor programs for non-experts While ‘dynamic programming’ is a technique that yields efficient solutions to computational problems in economics, genomic analysis, and other fields, adapting it to multicore chips requires a level of programming expertise that few economists and biologists have. But researchers from MIT’s Computer Science and Artificial Intelligence La... » read more