Elements of High-Performance RISC Processors - Design and Synthesis

News

Attendance is MANDATORY in this course.

Lab Course Overview

In this lab course a basic 32-bit RISC processor is extended with techniques common to high-performance processors.

The  extensions include:

• Pipelining, forwarding
• 4-way SIMD
• Pipelined hardware multiplier
• Cache, writeback queue, etc.

The resulting processor architecture is quite similar to the one used in the synergistic processing element of the Cell Broadband Engine, used e.g. in Playstation 3.

In order to achieve high performance,  proper design techniques and software tools for synthesis and analysis play an important role. The students learn how timing analysis, pipelining and retiming can be used to optimize the synthesis results. Finally, the processor is emulated on a Virtex-5 FPGA prototyping board.

The performance gain of the student's design is measured for the Mandelbrot set computation with respect to the basic RISC architecture. Because software has to be specifically tailored to high-performance processor architectures, the lab course also deals with scheduling techniques that avoid pipeline stalls.

Schedule

Literature

• D. A. Patterson, J. L. Hennessy: Computer Organization & Design . The Hardware / Software Interface (3rd Edition); San Francisco, Ca.: Morgan Kaufmann Publishers Inc., 2004
• J. L. Hennessy and D. A. Patterson: Computer Architecture - A Quantitative Approach (5th Edition); San Francisco, Ca.: Morgan Kaufmann Publishers Inc., 2012

Tutors

• Dr. Rafal Baranowski
• M. Sc. Chang Liu
• Dipl.-Inf. Laura Rodriguez Gomez