Chapter Draft of February 8, 2001
3
CHAPTER 1
Compiler Challenges for High-Performance Architectures
17
1.1
Overview and Goals
17
1.2
Pipelining
21
1.2.1 Pipelined Instruction Units 21
1.2.2 Pipelined Execution Units 23
1.2.3 Parallel Functional Units 24
1.2.4 Compiling for Scalar Pipelines. 25
1.3
Vector Instructions
29
1.3.1 Vector Hardware Overview 29
1.3.2 Compiling for Vector Pipelines 30
1.4
Superscalar and VLIW Processors
32
1.4.1 Multiple-Issue Instruction Units 32
1.4.2 Compiling for Multiple-Issue Processors 33
1.5
Processor Parallelism
35
1.5.1 Compiling for Asynchronous Parallelism 37
1.6
Memory Hierarchy
39
1.6.1 Compiling for Memory Hierarchy 41
1.7
A Case Study: Matrix Multiplication
42
1.8
Advanced Compiler Technology
47
1.8.1 Dependence 48
1.8.2 Transformations 50
1.9
Chapter Summary
51
1.10
Case Studies
51
1.11
Historical Comments and References
52
1.12
Exercises
53
1.13
References
54
CHAPTER 2
Dependence: Theory and Practice
57
2.1
Introduction
57
2.2
Dependence and its Properties
58
2.2.1 Load-Store Classification 60
2.2.2 Dependence in Loops 61
2.2.3 Dependence and Transformations 63
2.2.4 Distance and Direction Vectors 68
2.2.5 Loop-carried and Loop-independent Dependences 72
2.2.5.1 Loop-Carried Dependence 72
2.2.5.2 Loop-Independent Dependences 76
2.2.5.3 Iteration Reordering 78
剩余832页未读，继续阅读
评论8