| Abstract | .... 2 | |
| 1 | Introduction | .... 2 |
| 2 | KL1 and its Parallel Execution | .... 2 |
| 2.1 | Overview of KL1 | .... 2 |
| 2.2 | Parallel Execution Model | .... 2 |
| 2.3 | Memory Access Feature | .... 3 |
| 3 | Typical Parallel Inference Machines | .... 4 |
| 3.1 | Memory Architecture within a Cluster | .... 4 |
| 3.2 | Inter-Cluster Message Passing Architecture | .... 4 |
| 3.3 | KL1 System on a Typical PIM | .... 4 |
| 4 | Possible Inter-cluster Shared Memory Architecture | .... 5 |
| 4.1 | Hierarchical coherent cache Memory | .... 5 |
| 4.2 | Memory Access Based Inter-cluster Network | .... 6 |
| 5 | KL1 System under the Proposed Memory Architecture | .... 6 |
| 5.1 | Direct Access to the Remote Objects | .... 6 |
| 5.2 | Garbage Collection | .... 6 |
| 6 | Conclusion | .... 7 |
| Acknowledgement | .... 7 | |
| References | .... 7 |
ICOT研究論文(TR)一覧に戻る / Back to the list of ICOT Technical Reports