BNN-PYNQ处理中的位num的硬编码值是多少?

时间:2019-01-11 10:05:10

标签: conv-neural-network fpga

我想基于BNN-PYNQ构建自定义网络。 但是,我不知道BNN-PYNQ中的硬编码值。

源代码如下。

void DoCompute(ap_uint<64> *in, ap_uint<64>* out, const unsigned int numReps) {
#pragma HLS DATAFLOW
  stream<ap_uint<64>> inter0("DoCompute.inter0");
  stream<ap_uint<192>> inter0_1("DoCompute.inter0_1");
  stream<ap_uint<24>> inter0_2("DoCompute.inter0_2");
#pragma HLS STREAM variable=inter0_2 depth=128
  stream<ap_uint<64>> inter1("DoCompute.inter1");
#pragma HLS STREAM variable=inter1 depth=128
  stream<ap_uint<64>> inter2("DoCompute.inter2");
  stream<ap_uint<64>> inter3("DoCompute.inter3");
#pragma HLS STREAM variable=inter3 depth=128
  stream<ap_uint<128>> inter4("DoCompute.inter4");
#pragma HLS STREAM variable=inter4 depth=128
  stream<ap_uint<128>> inter5("DoCompute.inter5");
  stream<ap_uint<128>> inter6("DoCompute.inter6");
#pragma HLS STREAM variable=inter6 depth=81
  stream<ap_uint<256>> inter7("DoCompute.inter7");
#pragma HLS STREAM variable=inter7 depth=1
  stream<ap_uint<256>> inter8("DoCompute.inter8");
#pragma HLS STREAM variable=inter8 depth=1
  stream<ap_uint<64>> inter9("DoCompute.inter9");
#pragma HLS STREAM variable=inter9 depth=128
  stream<ap_uint<64>> inter10("DoCompute.inter10");
#pragma HLS STREAM variable=inter10 depth=3
  stream<ap_uint<64>> memOutStrm("DoCompute.memOutStrm");

所有源代码:https://github.com/Xilinx/BNN-PYNQ/blob/d84f6e887d224791b6a20e7a907123bf9e15ee41/bnn/src/network/cnvW1A1/hw/top.cpp

我认为此源代码对于并行计算是必需的。 ap_uint和depth的值如何确定?

0 个答案:

没有答案