During my last 2 weeks, I was a little bit busy for my project of Advance Computer Architecture course. In this project, the performance and power dissipation of superscalar processor are observed. The suitable processor for running gcc application is examined by tuning the parameter of single-core simplescalar architecture simulator. The performances and power issue of the processor are analyzed.
The development of processor design nowadays is driven by the parallelism at multiple level with energy and cost being the primary constraints. Superscalar architecture is one the processor architecture that implement instruction-level parallelism which exploits data-level parallelism using compiler, out-of order issue, pipelining, and speculative execution.
To develop superscalar efficiently, further studies and observation about performances and energy efficiency of processor configurations should be exploited. In this assignment, SimpleScalar and Wattch tools are used to find out the most suitable superscalar architecture configuration for gcc benchmark. 2 metrics IPC and EDP are observed to assess the performance and energy efficiency of processor. To analyze the IPC and EDP of processor, a testbench C code fmath which is available in the SimpleScalar toolbox is used.
Several configurable parameters in the processor performances are branch predictor, memory system, datapath, and function units. Using the toolboxes, changes of processor parameters are observed and 2 configurations of superscalar processor are designed to to increase the IPC and decrease the EDP of processor running the fmath application. The first configuration is designed for high IPC while the second configuration is designed for low EDP. From the simulation, it could be observed that the maximum IPC configuration is 60% higher than default IPC and the minimum EDP configuration is 38,3% lower than default IPC using specific application.
Through this course, I could learn more about processor configuration and performance analysis. Using a proper and formal analysis, a broad knowledge of processor characteristic could be obtained and used for implementation, which is important to reduce the research time for implementing high performance system. Next assignment is multi-core processor analysis using Intel Parallel Studio which will be done to observed the implementation of multiple core for running specific application.