An efficient heuristic for instruction scheduling on clustered vliw processors

Abstract

Clustering is a well-known technique for improving the scalability of classical VLIW processors. A clustered VLIW processor consists of multiple clusters, each of which has its own register file and functional units. This paper presents a novel phase coupled priority-based heuristic for scheduling a set of instructions in a basic block on a clustered VLIW processor. Our heuristic converts the instruction scheduling problem into the problem of scheduling a set of instructions with a common deadline. The priority of each instruction vi is the lmax(vi)-successor-tree-consistent deadline which is the upper bound on the latest completion time of vi in any feasible schedule for a relaxed problem where the precedence-latency constraints between vi and all its successors, as well as the resource constraints are considered. We have simulated our heuristic, UAS heuristic and Integrated heuristic on the 808 basic blocks taken from the MediaBench II benchmark suite using six processor models. On average, for the six processor models, our heuristic improves 25%, 25%, 33%, 23%, 26%, 27% over UAS heuristic, respectively, and 15%, 16%, 15%, 9%, 20%, 8% over Integrated heuristic, respectively.

CAeSaR: Unified cluster-assignment scheduling and communication reuse for clustered VLIW processors

“…CAeSaR runs just before register allocation, as shown in Figure 6. To evaluate CAeSaR's performance, we measure the total size (in cycles) of the schedules generated by the compiler under CAeSaR and compare it against two state-of-the-art clustering algorithms (UAS [28] and CS [37]). …”

“…CAeSaR can work with various clustering heuristics, but the implementation shown makes use of the Start-Cycle heuristic [9,17] which, according to [31], is the best for clustered architectures with low inter-cluster communication delays (like the 1-cycle delay we consider). Other heuristics such as the Completion-Cycle [9] or the Critical-Successor [37] could also be used instead. The main body of the CAeSaR algorithm is listed in Algorithm 1.…”

“…The actual heuristic is orthogonal to our approach, since ICC reuse is supported by our framework, no matter the decision of the clustering heuristic. Therefore, we can plug in any other clustering heuristic, such as the Completion-Cycle (CC) [9] or the Critical Successor (CS) [37].…”

“…We directly compare CAeSaR against the two state-of-the-art unified cluster assignment and scheduling algorithms: (UAS) [28], and Critical-Successor (CS) [37]. We also measure the intermediate scheme: decoupled UAS + ICC reuse (as shown in Figure 3.e).…”

“…Recently, a new clustering heuristic was introduced by [37]. This differs from the previously mentioned ones in that, under certain conditions, the clustering decision is based on earliest schedule cycle of the most critical successor of the current instruction.…”