Instruction replication for clustered microarchitectures

Abstract

This work presents a new compilation technique that uses instruction replication in order to

Virtual Cluster Scheduling Through the Scheduling Graph

“…By taking several decisions a final schedule and assignment is obtained. In our approach, a decision may be one of the following actions: (1) establish a distance relation between two instructions in the final schedule (the final schedule is found when all instructions are assigned to a particular cycle and to a physical cluster), (2) schedule an instruction in a particular cycle, (3) assign a set of instructions to the same physical cluster, or (4) assign two sets of instructions to different physical clusters.…”

“…Assuming a target processor with four clusters, the mapping process works as follows: (1) After fusing a pair of VCs, a new VC is created that contains all nodes of the original VCs, and an edge to any VC previously linked to the VCs fused. In the previous example, VC' 2 obtained from the fusion of VC 2 and VC 3 inherits all instructions previously mapped to them, and all edges from VCs linked to VC 2 or VC 3 .…”

“…Finally, several proposals to cluster scheduling based on modulo scheduling have been proposed [23] [2]. These proposals focus on loops, whereas our technique targets general code regions.…”

Partitioning Multi-Threaded Processors with a Large Number of Threads

“…The only drawback of such an architecture is the intercluster communication cost. Various groups [2,3,4,6,9,10] have studied cluster assignment mechanisms for one thread to reduce the overhead of inter-cluster communication. We extend their ideas to a clustered architecture with multiple threads.…”

AGAMOS: A Graph-Based Approach to Modulo Scheduling for Clustered Microarchitectures

“…The information considered is obtained through an estimate of the final schedule, which is called a pseudoschedule [3]. Our approach also features novel instruction replication heuristics to further reduce the number of communications [4], [5].…”