Hardware Support for Data Dependence Speculation in Distributed Shared-Memory Multiprocessors Via Cache-block Reconciliation

Data dependence speculation allows a compiler to relax the constraint of data-independence to issue tasks in parallel, increasing the potential for automatic extraction of parallelism from sequential programs. This paper proposes hardware mechanisms to support a data-dependence speculative distributed shared-memory (DDSM) architecture that enable speculative parallelization of programs with irregular data structures and inherent coarse-grain parallelism. E cient support for coarse-grain tasks requires large bu ers for speculative data; DDSM leverages cache and directory structures to provide large bu ers that are managed transparently from applications. The proposed cache and directory extensions provide support for distributed speculative versions of cache blocks, run-time detection of dependence violations, and program-order reconciliation of cache blocks. This paper describes the DDSM architecture and presents a simulation-based evaluation of its performance on ve benchmarks chosen from the Spec95 and Olden suites. The proposed system yields simulated speedups of up to 12.5 in a 16-node con guration for programs with coarse-grain speculative windows (millions of instructions and hundreds of KBytes of speculative data).