Fix/solverprecision

[abhinavsns]

Allowing for petsc to work in float and double both

[syaskovets]

Which use case problem does it solve? In this implementation the point expression and the other operand T d either have the same type, or the type of the operand is implicitly converted to double (which works for all base types). I don't see why it has to be compared with float here

[abhinavsns]

Which use case problem does it solve? In this implementation the point expression and the other operand T d either have the same type, or the type of the operand is implicitly converted to double (which works for all base types). I don't see why it has to be compared with float here

The Dcpse class uses point expressions which ambiguously match for the division operator to both cases when the type is Float, resulting in compilation failure when compiling float versions.

[syaskovets]

Could you please provide a reproducible example? I would like to understand the issue and fix it for all point_expression_op's. For now it seems adding typename check=typename std::enable_if<std::is_same<T,double>::value >::type to the other operator point_expression_op of the same kind should be a more straightforward solution

[abhinavsns]

typedef float float_type;
float_type eps=1;
Point<3, float_type> xq = {1,2,3},xp={4,5,6};
Point<3, float_type> normalizedArg = (xp - xq) / eps;

Compiling the above results in:

error: use of overloaded operator '/' is ambiguous (with operand types 'point_expression_op<Point<3U, float>, Point<3U, float>, Point<3U, float>, 2>' and 'float_type' (aka 'float'))
  577 |         Point<3, float_type> normalizedArg = (xp - xq) / eps;
      |                                              ~~~~~~~~~ ^ ~~~
openfpm_data/src/Space/Shape/Point_operators.hpp:1199:23: note: candidate function [with orig = Point<3, float>, exp1 = Point<3, float>, exp2 = Point<3, float>, op1 = 2, T = float]
 1199 | CREATE_POINT_OPERATOR(operator/,POINT_DIV)
      |                       ^
openfpm_data/src/Space/Shape/Point_operators.hpp:1199:23: note: candidate function [with orig = Point<3, float>, exp1 = Point<3, float>, exp2 = Point<3, float>, op1 = 2, T = float, check = void]