Improving GPU Performance: Reducing Memory Conflicts and Latency