Loading...
Please wait, while we are loading the content...
Similar Documents
Dataflow graph partitioning for the id compiler.
| Content Provider | CiteSeerX |
|---|---|
| Author | Chery, Yonald Searle, Campbell L. |
| Abstract | The multi-threaded compilation of Id poses problems due to its non-strictness. Non-strictness permits a yet to be computed result of a function application to be fed back as an argument to the computing function. As a result, a more efficient compile-time scheduling of the sub-expressions of a computation is not always possible and left to be resolved at run-time. The partitioning algorithms implemented [15] provide a mechanism for performing compiletime analysis to resolve some sub-expression ordering decisions. These algorithms attempt to group instructions into larger partitions of instructions sharing the same synchronization requirements. Larger partitions would in turn be compiled to longer sequential instruction threads targeted for execution on a multi-threaded dataflow machine. Post-partitioning phases of compilation to facilitate multi-threaded style code generation have also been implemented. The traditional program graph dataflow representation proved to be insufficient fo... |
| File Format | |
| Access Restriction | Open |
| Subject Keyword | Dataflow Graph Partitioning Id Compiler Sequential Instruction Thread Id Pose Problem Function Application Synchronization Requirement Efficient Compile-time Scheduling Multi-threaded Compilation Algorithm Attempt Compiletime Analysis Post-partitioning Phase Insufficient Fo Larger Partition Traditional Program Graph Dataflow Representation Group Instruction Multi-threaded Style Code Generation Multi-threaded Dataflow Machine Sub-expression Ordering Decision |
| Content Type | Text |
