Power-Efficient Predication Techniques for Acceleration of Control Flow Execution on CGRA

Abstract

Coarse-grained reconfigurable architecture typically has an array of processing elements which are controlled by a centralized unit. This makes it difficult to execute programs having control divergence among PEs without predication. However, conventional predication techniques have a negative impact on both performance and power consumption due to longer instruction words and unnecessary instruction-fetching decoding nullifying steps. This article reveals performance and power issues in predicated execution which have not been well-addressed yet. Furthermore, it proposes fast and power-efficient predication mechanisms. Experiments conducted through gate-level simulation show that our mechanism improves energy-delay product by 11.9% to 23.8% on average.

Acceleration of control flows on reconfigurable architecture with a composite method

“…Control flows are divided into several typical patterns [3], such as short-if (the body's length of the branching is shorter than four), long-if (the body's length of the branching is longer than four), nested-if, unbalanced-if (the lengths of if-and else-paths are unbalanced) and loop structures. They are used to verify the performance of these techniques.…”

“…Previously, many methods were proposed to address the above problem, such as predication execution [1,2], dual-issue-singleexecution (DISE) [3], Triggered Instruction Architecture (TIA) [4,5], parallel condition [6] and so on. However, there is still no single technique which is optimal across all types of control flow, so composite methods are the only choice [3,7].…”

“…However, there is still no single technique which is optimal across all types of control flow, so composite methods are the only choice [3,7].…”

Branch-Aware Loop Mapping on CGRAs

“…The new value is valid only if the path of the branch in which the new value is computed should have been executed. Architectural support and the idea of partial predication was presented in [11]. Figure 3(b) shows the partial predication transformation of the loop body in figure 3(a).…”

“…This method also does not require select operation. The idea of dual-issue scheme was presented recently in [11]. Figure 5(a) shows the DFG after partial predication transformation.…”