ECE 512
ECE 512 - Computer Microarchitecture
See full schedule from Course Explorer
Subject Area
- Hardware Systems
Course Director
Description
Design of high performance computer systems; instruction level concurrency; memory system implementation; pipelining, superscalar, and vector processing; compiler back-end code optimization; profile assisted code transformations; code generation and machine dependent code optimization; cache memory design for multiprocessors; synchronization implementation in multiprocessors; compatibility issues; technology factors; state-of-the-art commercial systems.
Notes
Same as CSE 528.
Topics
- Introduction and architecture
- Retargetable optimizing code generators
- Compile-time instruction memory access optimization: inlining and trace placement
- Instruction sequencing: branch target buffers, slot-based methods, exception handling, complexity and efficiency
- Data register efficiency: register allocation, renaming, windowing, interprocedural allocation
- Instruction concurrency: interlocking, forwarding compaction, multiple issue, state recovery
- Execution efficiency: hyperblock global code transformation
- Caches and interconnection for multiprocessors: directory/snooping nvalidate/update, multilevel, compilation, bus architectures
- Synchronization for multiprocessors: atomicity and locii, processor chips, processor boards, buses and memory boards
- Advanced system design tools: benchmark characterization, execution driven simulation, and fast system simulation
Detailed Description and Outline
Topics:
- Introduction and architecture
- Retargetable optimizing code generators
- Compile-time instruction memory access optimization: inlining and trace placement
- Instruction sequencing: branch target buffers, slot-based methods, exception handling, complexity and efficiency
- Data register efficiency: register allocation, renaming, windowing, interprocedural allocation
- Instruction concurrency: interlocking, forwarding compaction, multiple issue, state recovery
- Execution efficiency: hyperblock global code transformation
- Caches and interconnection for multiprocessors: directory/snooping nvalidate/update, multilevel, compilation, bus architectures
- Synchronization for multiprocessors: atomicity and locii, processor chips, processor boards, buses and memory boards
- Advanced system design tools: benchmark characterization, execution driven simulation, and fast system simulation
Same as CSE 528.
Texts
Lecture notes and collection of technical papers.
Last updated
2/13/2013