Possible bug in generated HLS

[wyanzhao]

Hi, I am trying to use Halide-HLS to write a simple HLS program. But I found in some cases, the tool will produce incorrect HLS results.
Here's the given Halide C++ code:

MyPipeline() : input(UInt(64), 1, "input"), A("A"), B("B"), C("C"), hw_output("hw_output")

{

    // define the algorithm

    A = BoundaryConditions::repeat_edge(input);

    Expr constant(((uint64_t)(274877906943)));

    output(x) =  A(x) & constant;

    // define common schedule: tile output

    args.push_back(input);

}

This program supposed to clear the 39th and 40th bit of A. However, in the generated HLS, the code will be:

 for (int _p2_output_s0_x = _21; _p2_output_s0_x < _21 + _137; _p2_output_s0_x++)

  {

   int32_t _138 = _14 + _15;

   int32_t _139 = _138 + -1;

   int32_t _140 = min(_p2_output_s0_x, _139);

   int32_t _141 = _140 - _14;

   int32_t _142 = max(_141, 0);

   uint64_t _143 = ((const uint64_t *)_input)[_142];

   uint64_t _144 = _143 & 63;

   int32_t _145 = _p2_output_s0_x - _21;

   ((uint64_t *)_p2_output)[_145] = _144;

  } // for _p2_output_s0_x

  int32_t _146 = _136 - _132;

  for (int _p2_output_s0_x = _132; _p2_output_s0_x < _132 + _146; _p2_output_s0_x++)

  {

   int32_t _147 = _p2_output_s0_x - _14;

   uint64_t _148 = ((const uint64_t *)_input)[_147];

   uint64_t _149 = _148 & 63;

   int32_t _150 = _p2_output_s0_x - _21;

   ((uint64_t *)_p2_output)[_150] = _149;

  } // for _p2_output_s0_x

Where the constant expr will become 63, instead of the original 274877906943. And this will produce incorrect results.