Extending a RISC Processor for Supporting the High Performance rchitecture
| Autor: | Hong-Sheng Yang, 楊宏盛 |
|---|---|
| Rok vydání: | 1994 |
| Druh dokumentu: | 學位論文 ; thesis |
| Popis: | 82 Using the thread package supported by the advanced operat- ing system, user can develop multithreaded application. Based on the multithreaded kernel, the computer system performance can be improved by the parallel execution of multithreaded applications. In this thesis, we propose two kinds of multi- threaded processor as the processor in uniprocessor system to realize the benefits of threads. Because of the rapid context- switching of threads, concurrent multithreaded processor (CMP) can overlap the cache-miss latency with useful computation of other threads. This latency-hiding technique can improve the processor utilization and throughput. Moreover, the parallel multithreaded processor (PMP) take advantage of the independ- ence between threads. In our PMP,instructions from two threads are issued simultaneously to multiple functional units and these instructions can begin execution unless there are functional unit conflicts. This parallel execution of two threads can decrease the CPI (cycle per instruction) of proce- ssor and the execution times of the applications. According to the performance analysis of our performance model,we extend a typical RISC architecture, DLX processor, to construct our CMP and PMP. We compare our processors with DLX processor, a conventional single-thread processor,by simulating with apply- ing some benchmarks. From the results, indeed, CMP can improve throughput and PMP can support the parallelism of threads to achieve speedup. Supported by our multithreaded processor, the benefits of thread can actually be realized to improve the uniprocessor computer system. |
| Databáze: | Networked Digital Library of Theses & Dissertations |
| Externí odkaz: |
|