Skip to content

dywsjtu/2-Way-Superscalar-P6-Processor

Repository files navigation

Overview

High-level Processor Design Overview

The project has built a P6-structure out of order pipeline to handle RISC-V instructions. Some advanced features such as superscalar execution are added to enhance processor performance. This report covers the design details of our project, the final implementation and testing results, and evaluation and analysis of performance for different optional features. Here is the report and slides

Run instruction

make assembly   # to generate instructions
make            # to run the code

Or

make assembly   # to generate instructions
make syn        # to make synthesis

Run test script

sh test.sh test     # to test simulation
sh test.sh syn      # to test synthsis

Final Performance

Final Performance

  • Synthesized clock period: 11ns
  • Average CPI: 2.23
  • #total cycles / #total instructions: 1.44

Pipeline performance compared with peer

Pipeline performance compared with peer