Course Specification
CSCE432, CSCE832 Course Specification
Catalog Description:
CSCE432 assumes knowledge of computer architecture, pipelining, memory hierarchy, instruction level parallelism, and compiler principles.High performance computing at the processor level. The underlying principles and micro-architectures of contemporary high-performance processors and systems. State-of-the-art architectural approaches to exploiting instruction level parallelism for performance enhancements. Case studies of actual systems highlight real-world trade-offs and theories.
Textbook(s) and/or Other Required Materials:
- John P. Shen and Mikko H. Lipasti, Modern Processor Design - Fundamentals of Superscalar Processors, McGraw-Hill Higher Education, 2005.
- John L. Hennessy and David A. Patterson, Computer Architecture -- A Quantitative Approach, 3rd Edition, Morgan Kaufmann Publishers Inc., 2003.
Prerequisites by Topic:
- Mastery of: the principles of classical single-processor Computer Architectures, including:
- Memory hierarchies, esp. Cache & main memory,
- Scalar processor architecture, RISC vs. CISC philosophies, & scalar pipeline performance.
- Familiarity with: Probability, & stochastic performance modeling of scalar processors, given standard stochastic & timing parameters (e.g. hit ratios, access times, cycle times, CPI, branch probabilities).
Course Objectives:
- Mastery of: the principles and practice of high-performance processor architectures, including:
- Data parallelism vs. Instruction Level.Parallelism (ILP),
- Principles and practice of Vector processor architectures,
- Principles and practice of Superscalar Architectures.
- Principles and practice of Very Long Instruction Word (VLIW) Architectures
Topics Covered:
- Review of Scalar Processor Architecture Topics:
- Pipelining: concept, variations, performance,
- Scalar instruction pipelines: branching & load delays, ideal & expected performance,
- Superpipelining
- RISC vs. CISC philosophies.
- Vector Processors:
- Data Level Parallelism, regularity, predictability, chaining,
- Case studies: primarily the Cray family.
- Superscalar Processors:
- Instruction Level Parallelism,
- Branch processing and instruction flow techniques,
- In-order vs. out-of-order issue vs. out-of-order execution,
- Strong vs. weak consistency,
- Data dependencies and issue rules,
- Contemporary case studies.
- VLIW Processors:
- Concepts, HJP code scheduling: loop folding, trace scheduling, & percolation scheduling, b. Early case studies, e.g FPS AP-120B, Trace family, c. Contemporary case studies.

