Top Function Definition - 2023.2 English

Vitis Tutorials: AI Engine (XD100)

Document ID
XD100
Release Date
2024-03-05
Version
2023.2 English

Use the dataflow pragma for concurrently scheduling the three functions mm2s0, dmaHls_rowsToCols, and s2mm1.

int dma_hls(
      hls::stream<qdma_axis<128, 0, 0, 0>> &strmOut_to_rowiseFFT,
      hls::stream<qdma_axis<128, 0, 0, 0>> &strmInp_from_rowiseFFT,
      hls::stream<qdma_axis<128, 0, 0, 0>> &strmOut_to_colwiseFFT,
      hls::stream<qdma_axis<128, 0, 0, 0>> &strmInp_from_colwiseFFT,
      int matSz, int rows, int cols, int iterCnt
     )
{
   #pragma HLS INTERFACE axis port=strmOut_to_rowiseFFT
   #pragma HLS INTERFACE axis port=strmInp_from_rowiseFFT
   #pragma HLS INTERFACE axis port=strmOut_to_colwiseFFT
   #pragma HLS INTERFACE axis port=strmInp_from_colwiseFFT
   
   #pragma HLS INTERFACE s_axilite port=matSz bundle=control
   #pragma HLS INTERFACE s_axilite port=rows bundle=control
   #pragma HLS INTERFACE s_axilite port=cols bundle=control
   #pragma HLS INTERFACE s_axilite port=iterCnt bundle=control
   #pragma HLS INTERFACE s_axilite port=return bundle=control  
   
   #pragma HLS DATAFLOW
   
   int stg0_errCnt = 0, stg1_errCnt = 0;
   
   ap_uint<128> goldenVal;

   goldenVal.range(127, 64) = 0x0000000100000001;
   goldenVal.range( 63,  0) = 0x0000000100000001;
   

   LOOP_ITER_MM2S0:for(int i = 0; i < iterCnt; ++i)
   {
      #pragma HLS loop_tripcount min=1 max=8
      
      mm2s0(strmOut_to_rowiseFFT, matSz);
   }
   
   LOOP_ITER_S2MM0_TO_MM2S1:for(int i = 0; i < iterCnt; ++i)
   {
      #pragma HLS loop_tripcount min=1 max=8
      
      dmaHls_rowsToCols(strmInp_from_rowiseFFT, strmOut_to_colwiseFFT, \
                        matSz, rows, cols, stg0_errCnt, goldenVal);
   }
   
   LOOP_ITER_S2MM1:for(int i = 0; i < iterCnt; ++i)
   {
      #pragma HLS loop_tripcount min=1 max=8
      
      s2mm1(strmInp_from_colwiseFFT, matSz, stg1_errCnt, goldenVal);
   }

   return (stg0_errCnt + stg1_errCnt);
}

The dma_hls kernel also specifies HLS pragmas to help optimize the kernel code and adhere to interface protocols. See this page for detailed documentation of all HLS pragmas. A summary of the HLS pragmas used in this kernel is given in the following table.

Switch Description
#pragma HLS INTERFACE In C/C++ code, all input and output operations are performed, in zero time, through formal function arguments. In a RTL design, these same input and output operations must be performed through a port in the design interface and typically operate using a specific input/output (I/O) protocol. For more information, see this page.
#pragma HLS PIPELINE II=1 Reduces the initiation interval (II) for a function or loop by allowing the concurrent execution of operations. The default type of pipeline is defined by the config_compile -pipeline_style command, but can be overridden in the PIPELINE pragma or directive. For more information, see this page.
#pragma HLS dataflow The DATAFLOW pragma enables task-level pipelining, allowing functions and loops to overlap in their operation, increasing the concurrency of the RTL implementation and increasing the overall throughput of the design. For more information, see this page.
#pragma HLS loop_tripcount When manually applied to a loop, this pragma specifies the total number of iterations performed by a loop. The LOOP_TRIPCOUNT pragma or directive is for analysis only, and does not impact the results of synthesis. For more information, see this page.
PS Host Application