(Downloads - 0)
For more info about our services contact : help@bestpfe.com
Table of contents
1 Introduction
1.1 The Java Execution and Runtime
1.2 Two JIT Compilers in the HotSpot VM
1.3 The Goal of This Thesis
1.4 Outline
2 Background
2.1 Java Bytecode
2.2 JIT Compilation
2.3 Intermediate Representations
2.4 Control Flow Graph
2.5 Static Single Assignment Form
2.6 Liveness Analysis and Live Ranges
3 Register Allocation
3.1 Local Register Allocation
3.2 Global Register Allocation
4 Register Allocation using Graph Coloring
4.1 Pruning the Interference Graph
4.2 Reconstructing the Interference Graph
5 The Linear Scan Algorithm
5.1 Block Order, Row Numbering and Intervals
5.2 Assigning Physical Registers to Virtual Registers
5.3 Resolve Stack Intervals
5.4 Lifetime Holes
5.5 Interval Splitting
6 Implementing the Linear Scan Algorithm in the Server Compiler
6.1 The Intermediate Representation in the Server Compiler
6.2 Mapping Nodes to Machine Instructions
6.2.1 Fixed Registers
6.3 Building Intervals
6.3.1 The Interval Class
6.3.2 Create Live Ranges
6.4 Assigning Registers and Memory Positions
6.5 Resolving the Intervals that are Assigned Memory Positions
6.6 Resolving Registers at Calls
6.6.1 Removing the SSA Form
7 Performance Testing and Results
7.1 Performance
7.2 Compilation Time
7.2.1 Register Allocation Time
7.2.2 Time Distribution inside the Linear Scan Register Allocator in the Server Compiler
7.2.3 Total Compilation Time
8 Future work
8.1 Improving the Runtime Performance
8.1.1 Interval Splitting
8.1.2 Use Profile Data for Spilling Heuristics
8.1.3 Improving the Phi Resolver
8.2 Reducing the Compilation Time
8.2.1 Reducing the Total Time being Spent in the Server Compiler
8.2.2 Replacing the Liveness Analysis
9 Conclusions
