Extending Automatic Parallelization to Optimize High-Level Abstractions for Multicore