Multicast Support - 2022.1 English

Versal ACAP AI Engine Programming Environment User Guide (UG1076)
Document ID
UG1076
Release Date
2022-05-25
Version
2022.1 English

Various multicast scenarios are supported in the graph, such as from a window to multiple windows, from stream to multiple streams, from input_plio to multiple windows etc. This section lists the supported types of multicast from a single source to multiple destinations. For additional details on input_plio/output_plio, and input_gmio/output_gmio, see Graph Programming Model.

Table 1. Multicast Support Scenarios
Scenario # Source Destination 1 Destination 2 Support
1 AI Engine Window AI Engine Window AI Engine Window Supported
2 AI Engine Window AI Engine Window AI Engine Stream Supported
3 AI Engine Window AI Engine Window output_plio/output_gmio Supported
4 AI Engine Window AI Engine Stream AI Engine Stream Supported
5 AI Engine Window AI Engine Stream output_plio/output_gmio Supported
6 AI Engine window output_plio/output_gmio output_plio/output_gmio Supported
7 AI Engine Stream AI Engine Window AI Engine window Supported
8 AI Engine Stream AI Engine Window AI Engine Stream Supported
9 AI Engine Stream AI Engine Window output_plio/output_gmio Supported
10 AI Engine Stream AI Engine Stream AI Engine Stream Supported
11 AI Engine Stream AI Engine Stream output_plio/output_gmio Supported
12 AI Engine Stream output_plio/output_gmio output_plio/output_gmio Supported
13 input_plio/input_gmio AI Engine Window AI Engine window Supported
14 input_plio/input_gmio AI Engine Window AI Engine Stream Not Supported
15 input_plio/input_gmio AI Engine Window output_plio/output_gmio Not Supported
16 input_plio/input_gmio AI Engine Stream AI Engine Stream Supported
17 input_plio/input_gmio AI Engine Stream output_plio/output_gmio Not Supported
18 input_plio/input_gmio output_plio/output_gmio output_plio/output_gmio Not Supported

Note the following:

  • All source and destination windows in the multicast connections are required to have the same size.
  • RTP and packet switching are not covered in this section.
  • If the multicast type is supported, the destination number is not limited if it can fit into the hardware.

When multiple streams are connected to the same source, the data is sent to all the destination ports at the same time and is only sent when all destinations are ready to receive data. This might cause stream stall or design hang if the FIFO depth of the stream connections are not deep enough. Refer to the examples in AI Engine Kernel Coding Best Practices Guide (UG1079) for more information about the stream stalls and potential solutions.

The following multicast example shows scenario number 10 from above table. Source and both destinations are stream.

In this graph.h code snippet, two sub-graphs named _graph0 and _graph1 are defined within the top graph named top_graph. This ensures that sending data to all destination ports at the same time.
class _graph0: public adf::graph {
private:
    adf::kernel kr;
public:
    adf::port<input> instream;
    adf::port<output> outstream;
    _graph0() {
        kr = adf::kernel::create(compute0);
        adf::runtime<ratio>(kr) = 0.9;
        adf::source(kr) = "compute0.cc";
        adf::connect<adf::stream> n0(instream, kr.in[0]);
        adf::connect<adf::stream> n1(kr.out[0], outstream);
    }
};

class _graph1: public adf::graph {
private:
    adf::kernel kr;
public:
    adf::port<input> instream;
    adf::port<output> outstream;
    _graph1() {
        kr = adf::kernel::create(compute1);
        adf::runtime<ratio>(kr) = 0.9;
        adf::source(kr) = "compute1.cc";
        adf::connect<adf::stream> n0(instream, kr.in[0]);
        adf::connect<adf::stream> n1(kr.out[0], outstream);
    }
};

class top_graph: public adf::graph {
private:

public:
    _graph0 g0;
    _graph1 g1;
    adf::input_plio  instream;
    adf::output_plio outstream0;
    adf::output_plio outstream1;
    top_graph()
    {
        instream   = adf::input_plio::create("aie_brodcast_0_S_AXIS",
                         adf::plio_32_bits,
                         "data/input.txt");
        outstream0 = adf::output_plio::create("aie_graph0_outstream",
                         adf::plio_32_bits,
                         "data/output0.txt");
        outstream1 = adf::output_plio::create("aie_graph1_outstream",
                         adf::plio_32_bits,
                         "data/output1.txt");

        adf::connect<adf::stream> n0(instream.out[0], g0.instream);
        adf::connect<adf::stream> n1(instream.out[0], g1.instream);
        adf::connect<adf::stream> n2(g0.outstream, outstream0.in[0]);
        adf::connect<adf::stream> n3(g1.outstream, outstream1.in[0]);
    }
};
In this graph.cpp code snippet, the graph calls are invoked from the top graph so all sub-graphs are receiving the same data at the same time.
using namespace adf;

top_graph top_g;

#if defined  (__AIESIM__)
int main () {
        top_g.init();
        top_g.run(3);
        top_g.wait();
        top_g.end();
        return 0;
}
#endif