Unified Assign and Schedule: A New Approach to Scheduling for Clustered Register File Microarchitectures
Emre Özer, Sanjeev Banerjia, Thomas M. Conte
Abstract
Emre Özer, Sanjeev Banerjia, Thomas M. Conte
Abstract
Recently, there has been a trend towards clustered microarchitectures to reduce the cycle time for wideissue microprocessors. In such processors, the register file and functional units are partitioned and grouped into clusters. Instruction scheduling for a clustered machine requires assignment and scheduling of operations to the clusters. In this paper, a new scheduling algorithm named unified-assign-and-schedule (UAS) is proposed for clustered, statically-scheduled architectures. UAS merges the cluster assignment and instruction scheduling phases in a natural and straightforward fashion. We compared the performance of UAS with various heuristics to the well-known Bottom-up Greedy (BUG) algorithm and to an optimal cluster scheduling algorithm, measuring the schedule lengths produced by all of the schedulers. Our results show that UAS gives better performance than the BUG algorithm and is quite close to optimal. 1 Introduction Many high performance processors are designed with wide is...
OpenAlex reports 174 citations for this work. Citation counts describe recorded attention and do not establish research quality.
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.
Recently, there has been a trend towards clustered microarchitectures to reduce the cycle time for wideissue microprocessors. In such processors, the register file and functional units are partitioned and grouped into clusters. Instruction scheduling for a clustered machine requires assignment and scheduling of operations to the clusters. In this paper, a new scheduling algorithm named unified-assign-and-schedule (UAS) is proposed for clustered, statically-scheduled architectures. UAS merges the cluster assignment and instruction scheduling phases in a natural and straightforward fashion. We compared the performance of UAS with various heuristics to the well-known Bottom-up Greedy (BUG) algorithm and to an optimal cluster scheduling algorithm, measuring the schedule lengths produced by all of the schedulers. Our results show that UAS gives better performance than the BUG algorithm and is quite close to optimal. 1 Introduction Many high performance processors are designed with wide is...
Key concepts: Computer science, Parallel computing, Register allocation, Heuristics, Scheduling (production processes), Register file, Gang scheduling, Fair-share scheduling