 |
|
|
|
|
7.6.10 Adaptive Speculation |
|
|
|
| |
|
|
|
|
|
7.6.11 Results |
|
|
|
| |
|
|
|
|
|
7.7 Comparing Vector and Multiple-Issue Processors |
|
|
|
| |
|
|
|
|
|
7.7.1 Cost Comparison |
|
|
|
| |
|
|
|
|
|
7.7.2 Performance Comparison |
|
|
|
| |
|
|
|
|
|
7.7.3 Alternative Organizations |
|
|
|
| |
|
|
|
|
|
7.8 Conclusions |
|
|
|
| |
|
|
|
|
|
7.9 Some Areas for Further Research |
|
|
|
| |
|
|
|
|
|
7.10 Data Notes |
|
|
|
| |
|
|
|
|
|
7.11 Annotated Bibliography |
|
|
|
| |
|
|
|
|
|
7.12 Problem Set |
|
|
|
| |
|
|
|
|
|
8
Shared Memory Multiprocessors |
|
|
|
| |
|
|
|
|
|
8.1 Basic Issues |
|
|
|
| |
|
|
|
|
|
8.2 Partitioning |
|
|
|
| |
|
|
|
|
|
8.3 Scheduling |
|
|
|
| |
|
|
|
|
|
8.3.1 Run-Time Scheduling Techniques |
|
|
|
| |
|
|
|
|
|
8.4 Synchronization and Coherency |
|
|
|
| |
|
|
|
|
|
8.5 The Effects of Partitioning and Scheduling Overhead |
|
|
|
| |
|
|
|
|
|
8.5.1 Grain Size and Overhead |
|
|
|
| |
|
|
|
|
|
8.6 Types of Shared Memory Multiprocessors |
|
|
|
| |
|
|
|
|
|
8.7 Multithreaded or Shared Resource Multiprocessing |
|
|
|
| |
|
|
|
|
|
8.8 Memory Coherence in Shared Memory Multiprocessors |
|
|
|
| |
|
|
|
|
|
8.9 Shared-Bus Multiprocessors |
|
|
|
| |
|
|
|
|
|
8.9.1 Snoopy Protocols |
|
|
|
| |
|
|
|
|
|
8.9.2 Bus-Based Models |
|
|
|
| |
|
|
|
|
|
8.10 Scalable Multiprocessors |
|
|
|
| |
|
|
|
|
|
8.11 Directory-Based Protocols |
|
|
|
| |
|
|
|
|
|
8.11.1 Directory Structure |
|
|
|
| |
|
|
|
|
|
8.11.2 Invalidate Protocols |
|
|
|
| |
|
|
|
|
|
8.11.3 Update Protocols |
|
|
|
| |
|
|
|
|
|
8.12 Evaluating Some Systems Alternatives |
|
|
|
| |
|
|
|
|
|
8.13 Interconnections |
|
|
|
| |
|
|
|
|
|
8.14 Static Networks |
|
|
|
| |
|
|
|
|
|
8.14.1 Links and Nodes |
|
|
|
| |
|
|
|
|
|
8.15 Dynamic Networks |
|
|
|
| |
|
|
|
|
|
8.16 Evaluating Interconnect Networks |
|
|
|
| |
|
|
|
|
|
8.16.1 Direct Static vs. Indirect Dynamic |
|
|
|
| |
|
|
|
|
|
8.16.2 Network Dimensionality and Link-Limited Network |
|
|
|
| |
|
|
|
|
|
8.17 Hotspots and Combining |
|
|
|
| |