It has been 32 (=2^5) years since the first microprocessor-based general-purpose computer was built. In the 1970s, microprocessors were low-end, marginal computing engines. Today, microprocessors are the central computing element in the vast majority of computers. These years have seen remarkable progress in the techniques used to make fast microprocessors and in the resulting performance gains. This talk examines the major factors that have contributed to this performance growth, focusing on the exploitation of instruction-level parallelism. We assess the current state-of-the-art and future challenges in maintaining this growth in processor performance. We show that diminishing returns in both instruction level parallelism appear to lead to lower rates of performance improvement in the future. We speculate on what approaches may be most successful in the near term, while arguing that more radical approaches will be needed in the not too distant future.
Apr 15, 2003
Free and open to the public