Inconsistent results for mod 0 operation

[Garic152]

Description:

When using the outerMod function on matrices filled with zeros, the result is unexpected large negative values, to be exact the minimum signed 64-bit integer value. We get -nan when doing the same for zero floats. When using mod on two 0 scalars, we get an error message as expected.

Examples:

1. Using the outerMod function with two matrices filled with zero integers:

   X = [0, 0];
   Y = [0, 0];
   result = outerMod(X, Y);
   print(result);

   Output:

   DenseMatrix(2x2, int64_t)
   -9223372036854775808 -9223372036854775808
   -9223372036854775808 -9223372036854775808

2. Using the outerMod function with two matrices filled with zero floats:

   X = [0.0, 0.0];
   Y = [0.0, 0.0];
   result = outerMod(X, Y);
   print(result);

   Output:

   DenseMatrix(2x2, int64_t)
   -nan -nan
   -nan -nan

3. Using the mod functionality with two scalar zeros:

   X = 0;
   Y = 0;
   result = X % Y;
   print(result);

   Output:

   [error]: Lowering pipeline error. CanonicalizerPass (mlir::daphne::EwModOp::fold) failed with the following message [ Can't compute mod 0 ]

To my knowledge, all of these mod usecases should return an error message like in the third example.