The AI Engine kernel
can be made to run continuously by using graph::run(-1). This means the kernel restarts automatically after the last
iteration is complete.
graph::run() without an argument runs the AI Engine kernels for a previously specified
number of iterations (which is infinity by default if the graph is run without any
arguments). If you run the graph with a finite number of iterations, for example,
mygraph.run(3); mygraph.run();, the second run call
also runs for three iterations.However, it requires input buffers and output buffers to be ready before it can start. Thus, there is a small overhead between kernel execution iterations. This section describes a method to construct a type of kernel that has zero overhead and runs forever. It is called a free-running AI Engine kernel.
Free-running kernels can only have streaming interfaces. Loops with infinite iterations can be inside the kernel. See the following example:
void free_running_aie(input_stream<int32> *in, output_stream<int32> *out) {
while(true){//for(;;) is acceptable for C++
int32 tmp=readincr(in);
chess_separator_scheduler();//make sure stream is flushed
writeincr(out,tmp+1);
chess_separator_scheduler();//make sure stream is flushed
}
}The free-running kernel must have its own graph defined. This graph must have no other non-free-running kernels, because the graph never stops and non-free-running kernels lose control after being started. The graph containing the free-running kernel must be a top-level graph that can be connected to other graphs, or it can be connected to PLIO or GMIO. The following code sample shows a sample connection between the free-running graph and other graphs.
passingGraph mygraph;
freeGraph mygraph_free;
connect<> net0(mygraph.out1,mygraph_free.in);
connect<> net1(mygraph_free.out,mygraph.in2);
The free-running graph can be started using mygraph_free.run(-1) or automatically started after loading.