An evaluation of memory consistency models for shared-memory systems with ILP processors

The memory consistency model of a shared-memory multiprocessor determines the extent to which memory operations may be overlapped or reordered for better performance. Studies on previous-generation shared-memory multiprocessors have shown that relaxed memory consistency models like release consistency (RC) can significantly outperform the conceptually simpler model of sequential consistency (SC). Current and next-generation multiprocessors use commodity microprocessors that aggressively exploit instruction-level parallelism (ILP) using methods such as multiple issue, dynamic scheduling, and non-blocking reads. For such processors, researchers have conjectured that two techniques, hardware-controlled non-binding prefetching and speculative reads, have the potential to equalize the hardware performance of memory consistency models. These techniques have recently begun to appear in commercial microprocessors, and re-open the question of whether the performance benefits of release consistency justify its added programming complexity. This thesis performs the first detailed quantitative comparison of several implementations of sequential consistency and release consistency optimized for aggressive ILP processors. Our results indicate that although hardware prefetching and speculative reads dramatically improve the performance of sequential consistency, the simplest RC version continues to significantly outperform the most optimized SC version. Additionally, the performance of SC is highly sensitive to the cache write policy and the aggressiveness of the cache-coherence protocol, while the performance of RC is generally stable across all implementations. Overall our results show that RC hardware has significant performance benefits over SC hardware, and at the same time, requires less system complexity with ILP processors. Memory write latencies that hardware prefetching and speculative loads are unsuccessful in hiding are the main reason for the performance difference between SC and RC