Submission + - Making the transition to parallel
inverselimit writes: Intel today announces an 80-core chip consuming 62W. As Moore's law gives way to parallelism, do developers need to rethink their approach to software? Some progress has been made at places like Google, with their MapReduce framework, GFS, etc. Some claim that functional languages, or some other technology, will be the key to effectively utilizing parallelism. What new ideas, abstractions, or technologies do we need to use parallelism effectively?