FE-Logo
  • Home
  • Study Material
  • Non-Deterministic Finite Automation
    • Introduction to Compiler
    • The Structure of a Compiler
    • Intermediate Code Generation
    • Building a Compiler
    • Applications of Compiler
    • Optimizations for Computer Architectures
    • Design of New Computer Architectures
    • Program Translations
    • Software Productivity Tools
    • Programming Language Basics
    • Minimisation of DFAs
    • Explicit Access Control
    • Parameter Passing Mechanisms
    • Introduction to Lexical Analysis
    • Regular expressions
    • Short hands
    • Nondeterministic finite automata
    • Converting a regular expression to an NFA
    • Deterministic finite automata
    • Converting an NFA to a DFA
    • The subset construction
    • Dead states
    • Lexers and lexer generators
    • Splitting the input stream
    • Lexical errors
    • Properties of regular languages
    • Limits to expressive power
    • The Role of the Lexical Analyzer
    • Input Buffering
    • Specification of Tokens
    • Operations on Languages
    • Regular Definitions and Extensions
    • Recognition of Tokens
    • The Lexical-Analyzer Generator Lex
    • Finite Automata
    • Construction of an NFA from a Regular Expression
    • Efficiency of String-Processing Algorithms
    • The Structure of the Generated Analyzer
    • Optimization of DFA-Based Pattern Matchers

  • Basic Parsing Techniques
    • Introduction to Syntax analysis
    • Context-free grammars
    • Writing context free grammars
    • Derivation
    • Syntax trees and ambiguity
    • Operator precedence
    • Writing ambiguous expression grammars
    • Other sources of ambiguity
    • Syntax analysis and Predictive parsing
    • Nullable and FIRST
    • Predictive parsing revisited
    • FOLLOW
    • LL(1) parsing
    • Methods for rewriting grammars for LL(1) parsing
    • SLR parsing
    • Constructions of SLR parse tables
    • Conflicts in SLR parse-tables
    • Using precedence rules in LR parse tables
    • Using LR-parser generators
    • Properties of context-free languages
    • Introduction to Syntax-Directed Translator
    • Evaluating an SDD at the Nodes of a Parse Tree
    • Evaluation Orders for SDD\'s
    • Ordering the Evaluation of Attributes
    • A larger example of calculating FIRST and FOLLOW
    • Syntax Definition
    • Associativity of Operators
    • Parse Trees
    • Ambiguity
    • Syntax-Directed Translation
    • Synthesized Attributes
    • Tree Traversals
    • Parsing
    • Predictive Parsing
    • Use e-Productions
    • Translator for Simple Expressions
    • Semantic Rules with Controlled Side Effects
    • Applications of Syntax-Directed Translation
    • The Structure of a Type of syntax
    • Switch-Statements
    • Syntax-Directed Translation Schemes
    • Postfix Translation Schemes
    • SDT\'s With Actions Inside Productions
    • Eliminating Left Recursion from SDT\'s
    • SDT\'s for L-Attributed Definitions
    • Implementing L-Attributed SDD\'s
    • On-The-Fly Code Generation
    • L-Attributed SDD\'s and LL Parsing
    • Bottom-Up Parsing of L-Attributed SDD\'s

  • Syntax-directed Translation
    • Register Allocation and Assignment
    • Semantic Analysis
    • Introduction to Intermediate Code Generation
    • Variants of Syntax Trees
    • Variants of Syntax Trees
    • The Value-Number Method for Constructing DAG\'s
    • Three-Address Code
    • Quadruples
    • Triples
    • Static Single-Assignment Form
    • Types and Declarations
    • Type Equivalence
    • Sequences of Declarations
    • Translation of Expressions
    • Incremental Translation
    • Addressing Array Elements
    • Translation of Array References
    • Type Checking
    • Type Conversions
    • Overloading of Functions and Operators
    • Type Inference and Polymorphic Functions
    • Algorithm for Unification
    • Control Flow
    • Flow-of-Control Statements
    • Control-Flow Translation of Boolean Expressions
    • Boolean Values and Jumping Code
    • Back patching
    • Backpatching for Boolean Expressions
    • Flow-of-Control Statements
    • Break-, Continue-, and Goto-Statements
    • Introduction to Run-Time Environments
    • Stack Allocation of Space
    • Activation Records
    • Calling Sequences
    • Variable-Length Data on the Stack
    • Access to Nonlocal Data on the Stack
    • Displays
    • Heap Management
    • Locality in Programs
    • Reducing Fragmentation
    • Managing and Coalescing Free Space
    • Manual Deallocation Requests
    • Reachability
    • Introduction to Garbage Collection
    • Reference Counting Garbage Collectors
    • Introduction to Trace-Based Collection
    • Basic Abstraction
    • Optimizing Mark-and-Sweep
    • Mark-and-Compact Garbage Collectors
    • Copying collectors
    • Short-Pause Garbage Collection
    • Incremental Reachability Analysis
    • Partial-Collection Basics
    • The Train Algorithm
    • Parallel and Concurrent Garbage Collection
    • Partial Object Relocation
    • Introduction Code Generation
    • Issues in the Design of a Code Generator
    • Instruction Selection
    • Register Allocation
    • The Target Language
    • Addresses in the Target Code
    • Stack Allocation
    • Run-Time Addresses for Names
    • Basic Blocks and Flow Graphs
    • Basic Blocks
    • Next-Use Information
    • Representation of Flow Graphs
    • Optimization of Basic Blocks
    • Dead Code Elimination
    • Representation of Array References
    • Pointer Assignments and Procedure Calls
    • A Simple Code Generator
    • The Code-Generation Algorithm
    • Design of the Function getReg
    • Peephole Optimization
    • Algebraic Simplification and Reduction in Strength
    • Register Assignment for Outer Loops
    • Instruction Selection by Tree Rewriting
    • Code Generation by Tiling an Input Tree
    • Pattern Matching by Parsing
    • General Tree Matching
    • Optimal Code Generation for Expressions
    • Evaluating Expressions with an Insufficient Supply of Registers
    • Dynamic Programming Code-Generation

  • Data Flow Analysis
    • The Lazy-Code-Motion Algorithm
    • Introduction to Machine-Independent Optimizations
    • The Dynamic Programming Algorithm
    • The Principal Sources of Optimization
    • Semantics-Preserving Transformations
    • Copy Propagation
    • Induction Variables and Reduction in Strength
    • Introduction to Data-Flow Analysis
    • The Data-Flow Analysis Schema
    • Reaching Definitions
    • Live-Variable Analysis
    • Available Expressions
    • Foundations of Data-Flow Analysis
    • Transfer Functions
    • The Iterative Algorithm for General Frameworks
    • Meaning of a Data-Flow Solution
    • Constant Propagation
    • Transfer Functions for the Constant-Propagation Framework
    • Partial-Redundancy Elimination
    • The Lazy-Code-Motion Problem
    • Loops in Flow Graphs
    • Depth-First Ordering
    • Back Edges and Reducibility
    • Natural Loops
    • Speed of Convergence of Iterative Data-Flow Algorithms
    • Region-Based Analysis
    • Necessary Assumptions About Transfer Functions
    • An Algorithm for Region-Based Analysis
    • Handling Non-reducible Flow Graphs
    • Symbolic Analysis
    • Data-Flow Problem Formulation
    • Region-Based Symbolic Analysis

  • Code Generation
    • Introduction to Software Pipelining of Loops
    • Matrix Multiply: An In-Depth Example
    • Software Pipelining of Loops
    • Introduction Instruction-Level Parallelism
    • Multiple Instruction Issue
    • A Basic Machine Model
    • Code-Scheduling Constraints
    • Finding Dependences Among Memory Accesses
    • Phase Ordering Between Register Allocation and Code Scheduling
    • Speculative Execution Support
    • Basic-Block Scheduling
    • List Scheduling of Basic Blocks
    • Global Code Scheduling
    • Upward Code Motion
    • Updating Data Dependences
    • Advanced Code Motion Techniques
    • Software Pipelining
    • Register Allocation and Code Generation
    • A Software-Pipelining Algorithm
    • Scheduling Cyclic Dependence Graphs
    • Improvements to the Pipelining Algorithms
    • Conditional Statements and Hardware Support for Software Pipelining
    • Basic Concepts of Parallelism and Locality
    • Parallelism in Applications
    • Loop-Level Parallelism
    • Introduction to Affine Transform Theory
    • Optimizations
    • Iteration Spaces
    • Affine Array Indexes
    • Controlling the Order of Execution
    • Changing Axes
    • Intermediate Code for Procedures
    • Data Reuse
    • Self Reuse
    • Self-Spatial Reuse
    • Array Data-Dependence Analysis
    • Integer Linear Programming
    • Heuristics for Solving Integer Linear Programs
    • Solving General Integer Linear Programs
    • Finding Synchronization-Free Parallelism
    • Affine Space Partitions
    • Space-Partition Constraints
    • Solving Space-Partition Constraints
    • A Simple Code-Generation Algorithm
    • Eliminating Empty Iterations
    • Synchronization Between Parallel Loops
    • The Parallelization Algorithm and Hierarchical Time
    • Pipelining
    • Solving Time-Partition Constraints by Farkas' Lemma
    • Code Transformations
    • Parallelism With Minimum Synchronization
    • Locality Optimizations
    • Partition Interleaving
    • Putting it All Together
    • Uses of Affine Transforms
    • Interprocedural Analysis
    • Context Sensitivity
    • Cloning-Based Context-Sensitive Analysis
    • Importance of Interprocedural Analysis
    • SQL Injection
    • A Logical Representation of Data Flow
    • Execution of Datalog Programs
    • Problematic Datalog Rules
    • A Simple Pointer-Analysis Algorithm
    • Flow Insensitivity
    • Context-Insensitive Interprocedural Analysis
    • Context-Sensitive Pointer Analysis
    • Adding Context to Datalog Rules
    • Datalog Implementation by BDD's
    • Relational Operations as BDD Operations

Branch : Computer Science and Engineering
Subject : Compiler design
Unit : Code Generation

Upward Code Motion


Introduction: We now examine carefully what it means to move an operation up a path. Suppose we wish to move an operation from block src up a control-flow path to block dst. We assume that such a move does not violate any data dependences and that it makes paths through dst and src run faster. If dst dominates src, and src post-dominates dst, then the operation moved is executed once and only once, when it should. If src does not postdominate dst

Then there exists a path that passes through dst that does not reach src. An extra operation would have been executed in this case. This code motion is illegal unless the operation moved has no unwanted side effects. If the moved operation executes "for free" (i.e., it uses only resources that otherwise would be idle), then this move has no cost. It is beneficial only if the control flow reaches src. If dst does not dominate src

Then there exists a path that reaches src without first going through dst. We need to insert copies of the moved operation along such paths. We know how to achieve exactly that from our discussion of partial redundancy elimination in Section 9.5- We place copies of the operation along basic blocks that form a cut set separating the entry block from src. At each place where the operation is inserted, the following constraints must be satisfied:

1.  The operands of t he operation must hold the same values as in the original,

2.  The result does not overwrite a value that is still needed, and

3.  It itself is not subsequently overwritten befpre reaching src. These copies render the original instruction in src fully redundant, and it thus can be eliminated.

We refer to the extra copies of the operation as compensation code. As discussed in Section 9.5, basic blocks can be inserted along critical edges to create places for holding such copies. The compensation code can potentially make some paths run slower. Thus, this code motion improves program execution only if the optimized paths are executed more frequently than the non-optimized ones.

Downward Code Motion

Suppose we are interested in moving an operation from block src down a controlflow path to block dst. We can reason about such code motion in the same way as above.

If src does not dominate dst

Then there exists a path that reaches dst without first visiting src. Again, an extra operation will be executed in this case. Unfortunately, downward code motion is often applied to writes, which have the side effects of overwriting old values. We can get around this problem by replicating the basic blocks along the paths from src to dst, and placing the operation only in the new copy of dst. Another approach, if available, is to use predicated instructions. We guard the operation moved with the predicate that guards the src block. Note that the predicated instruction must be scheduled only in a block dominated by the computation of the predicate, because the predicate would not be available otherwise.

If dst does not postdominate src

As in the discussion above, compensation code needs to be inserted so that the operation moved is executed on all paths not visiting dst. This transformation is again analogous to partial redundancy elimination, except that the copies are placed below the src block in a cut set that separates src from the exit.

Summary of Upward and Downward Code Motion: From this discussion, we see that there is a range of possible global code motions which vary in terms of benefit, cost, and implementation complexity. Figure 10.13 shows a summary of these various code motions; the lines correspond to the following four cases:

1. Moving instructions between control-equivalent blocks is simplest and most cost effective. No extra operations are ever executed and no compensation code is needed.

2. Extra operations may be executed if the source does not postdominate (dominate) the destination in upward (downward) code motion. This code motion is beneficial if the extra operations can be executed for free, and the path passing through the source block is executed.

3. Compensation code is needed if the destination does not dominate (postdominate) the source in upward (downward) code motion. The paths with the compensation code may be slowed down, so it is important that the optimized paths are more frequently executed.

4. The last case combines the disadvantages of the second and third case: extra operations may be executed and compensation code is needed.

Questions of this topic


Ask your question

<
>