DocumentCode :
2774654
Title :
Balanced microprocessor design keeps performance peaked
Author :
Iacobovici, Sorin ; Iacobelli, Franco
Author_Institution :
Nat. Semicond., Santa Clara, CA, USA
fYear :
1989
fDate :
17-19 May 1989
Firstpage :
371
Lastpage :
375
Abstract :
The factors that determine a CPU´s (central processing unit´s) performance are presented, and the design solutions featured by the NS32532 microprocessor are described. As the CPU´s operating frequency increases, the decisions on the CPU´s pipeline configuration and on which support mechanisms to integrate on chip become crucial for the CPU´s sustained performance in a system. In order to sustain an 8 to 10 MIPS average performance at 30 MHz in a realistic system, the NS32532 microprocessor was designed as a balanced machine. It integrates an MMU (memory management unit) and twin caches on chip, in order to supply instructions and operands at the speed requested by the fast instruction execution pipeline. Pipeline mechanisms like operand prefetch and branch prediction also help maximize the system performance. Data dependencies are solved in NS32532 by hardware, in order to be able to run Series 3200 code without any modification and get a very significant performance boost. A carefully designed bus interface gives full support to high-performance system design. In lower cost systems wait states are necessary, but their penalty is cushioned by the lower performance degradation per wait state (only 3 to 5% per wait state, compared to about 20% for other microprocessors)
Keywords :
computer architecture; microprocessor chips; pipeline processing; 30 MHz; 8 to 10 MIPS; CPU operating frequency; NS32532 microprocessor; Series 3200 code; balanced machine; branch prediction; bus interface; fast instruction execution pipeline; internal architecture; memory management unit; microprocessor design; operand prefetch; pipeline configuration; support mechanisms; twin caches; wait states; Central Processing Unit; Costs; Frequency; Hardware; Memory management; Microprocessors; Pipelines; Prefetching; System performance; System-on-a-chip;
fLanguage :
English
Publisher :
ieee
Conference_Titel :
VLSI Technology, Systems and Applications, 1989. Proceedings of Technical Papers. 1989 International Symposium on
Conference_Location :
Taipei
Type :
conf
DOI :
10.1109/VTSA.1989.68648
Filename :
68648
Link To Document :
بازگشت