Fine-grain multi-thread processor architecture for massively parallel processing

Abstract

Latency, caused by remofe memo y access and remote procedure call, is one of the most serious problems in massively parallel computers. In order to eliminate the processors' idle time caused by these latencies, processors must perform fast context switching among fine-grain concurrent processes.In this paper, we propose a processor architecture, called Datarol-11, that promotes eficient finegrain multi-thread execution by performing fast context switching among fine-gram concurrent processes. In the Datarol-11 processor, an implicit register load/store mechanism is embedded in the execution pipeline in order to reduce memo y access overhead caused by contezt switching. In order to reduce local memory access latency, a two-level hierarchical memory system and a load control mechanism are also introduced. We describe the Datarol-11 processor architecture, and show ifs evaluation results.

A continuation-based noninterruptible multithreading processor architecture

“…The dataflow-based approach aims to improve the fine-grained dataflow parallelism toward coarser-gained, so-called macro-dataflow, that augments dataflow with control-flow, and has evolved into multithreading. This approach was active in the 1980s and 1990s [5][6][7][8]. Unfortunately, this approach seems to have declined.…”

“…The core concept of the Fuce thread execution model is the event-driven computation [8]. Every thread receives continuation signals from other threads and its execution is triggered on the receipt of all corresponding continuation's.…”

“…A thread is defined as a block of instructions that are sequentially executed without being interrupted. Every execution of thread is triggered by events called continuation [8]. Executions among threads are partially ordered, and their order is specified by continuation, that is, the continuation-based multithreaded program.…”

“…Fuce is evolved from dataflow architectures (e.g. [5][6][7][8]) to make it more practical. A thread is defined as a block of instructions that are sequentially executed without being interrupted.…”

OS Mechanism for Continuation-based Fine-grained Threads on Dedicated and Commodity Processors

“…Fine-grained multithreading code can be hand-coded based on an appropriate model or generated from very high level languages such as functional or logic programming languages. Many architectures dedicated to such multithreaded programs have been proposed [7,8,9,10,11,12,13]. However, in this paper, we discuss issues on operating systems for fine-grained multithread programs.…”