学术文献库

The growing memory footprints of cloud and big data applications mean that data center CPUs can spend significant time waiting for memory. An attractive approach to improving performance in such centralized compute settings is to employ prefetchers that are customized per application, where gains can be easily scaled across thousands of machines. Helper thread prefetching is such a technique but has yet to achieve wide adoption since it requires spare thread contexts or special hardware/firmware support. In this paper, we propose an inline software prefetching technique that overcomes these restrictions by inserting the helper code into the main thread itself. Our approach is complementary to and does not interfere with existing hardware prefetchers since we target only delinquent irregular load instructions (those with no constant or striding address patterns). For each chosen load instruction, we generate and insert a customized software prefetcher extracted from and mimicking the application's dataflow, all without access to the application source code. For a set of irregular workloads that are memory-bound, we demonstrate up to 2X single-thread performance improvement on recent high-end hardware (Intel Skylake) and up to 83% speedup over a helper thread implementation on the same hardware, due to the absence of thread spawning overhead.