A Dynamic Programming Approach to Optimal Integrated Code Generation
Christoph W. Keßler, Andrzej Bednarski
Abstract
Christoph W. Keßler, Andrzej Bednarski
Abstract
Phase-decoupled methods for code generation are the state of the art in compilers for standard processors but generally produce code of poor quality for irregular target architectures such as many DSPs. In that case, the generation of efficient code requires the simultaneous solution of the main subproblems instruction selection, instruction scheduling, and register allocation, as an integrated optimization problem.
A significance statement is not available in the OpenAlex record.
A contribution statement is not available in the OpenAlex record.
Method details are not available in the OpenAlex metadata.
Findings are not separately available in the OpenAlex metadata.
Limitations are not available in the OpenAlex metadata.
Application details are not available in the OpenAlex metadata.
Phase-decoupled methods for code generation are the state of the art in compilers for standard processors but generally produce code of poor quality for irregular target architectures such as many DSPs. In that case, the generation of efficient code requires the simultaneous solution of the main subproblems instruction selection, instruction scheduling, and register allocation, as an integrated optimization problem.
Key concepts: Computer science, Register allocation, Compiler, Code generation, Instruction scheduling, Parallel computing, Code (set theory), Scheduling (production processes)