GCC Faster with Automatic Parallelization
The upcoming GNU Compiler Collection (GCC) version 4.4 will combine the existing Graphite loop optimization with autopar automatic parallelization support.
Soon-to-be-released GCC 4.4 will include the Graphite framework that should optimize high-level loops in multi-core systems, even though it has required specific compiler configuration.
IBM's Razya Ladelsky has since proposed plans in GCC mail for automating this optimization. The Graphite framework will integrate with autopar, the automatic parallelization code generator, so that upcoming versions of GCC will identify next to simple loops also more complex structures for better performance on multi-core systems.
Source: Phoronix