3.5.1.9 Loop Pipelining

Loop pipelining is an optimization that can automatically extract loop-level parallelism to create an efficient hardware pipeline. It allows executing multiple loop iterations concurrently on the same pipelined hardware.

To use loop pipelining, the user needs to specify the loop pipeline pragma above the applicable loop:

#pragma HLS loop pipeline
for (i = 1; i < N; i++) {
    a[i] = a[i-1] + 2
}

An important concept in loop pipelining is the initiation interval (II), which is the cycle interval between starting successive iterations of the loop. The best performance and hardware usage is achieved when II=1, which means that successive iterations of the loop can begin every clock cycle. A pipelined loop with an II=2 means that successive iterations of the loop can begin every two clock cycles, corresponding to half of the throughput of an II=1 loop.

By default, SmartHLS™ always attempts to create a pipeline with an II=1. However, this is not possible in some cases due to resource constraints or cross-iteration dependencies. Refer to Optimization Guide for more examples and details on loop pipelining. When II=1 cannot be met, SmartHLS's pipeline scheduling algorithm will try to find the smallest possible II that satisfies the constraints and dependencies.