Loop fusion for clustered VLIW architectures

Embedded systems require maximum performance from a processor within significant constraints in power consumption and chip cost. Using software pipelining, high-performance digital signal processors can often exploit considerable instruction-level parallelism (ILP), and thus significantly improve performance. However, software pipelining, in some instances, hinders the goals of low power consumption and low chip cost. Specifically, the registers required by a software pipelined loop may exceed the size of the physical register set.The register pressure problem incurred by software pipelining makes it difficult to build a high-performance embedded processor with a single, multi-ported register bank with enough registers to support high levels of ILP while maintaining clock speed and limiting power consumption. The large number of ports required to support a single register bank severely hampers access time. The port requirement for a register bank can be reduced via hardware by partitioning the register bank into multiple banks connected to disjoint subsets of functional units, called clusters. Since a functional unit is not directly connected to all register banks, wasted energy and resources can result due to delays incurred when accessing "non-local" registers.The overhead due to partitioning of the register set can be ameliorated by using high-level compiler loop optimization techniques such as unrolling, unroll-and-jam and fusion. High-level loop optimizations spread data-independent parallelism across clusters that may not require "non-local" register accesses and can provide work to hide the latency of any such register accesses that are needed.In this paper, we examine the effects of loop fusion on DSP loops run on four simulated, clustered VLIW architectures and the Texas Instruments TMS320C64x. Our experiments show a 1.3 -- 2 harmonic mean speedup.