Compiler Optimization for Superscalar Systems: Global Instruction Scheduling without Copies