boost multiprecision gmp_float convert_to() gives compilation error

Question

I am using boost multiprecision library floating point number. I need to map a gmp_float to an tanh and then take it as a double because value of tanh will be [0, 1). When I use convert_to<double>() I get compilation error as shown in the snippet bellow.

typedef boost::multiprecision::number<boost::multiprecision::gmp_float<4>> float_type;

float_type previous_v = agent->_velocity(i, j);
float_type sigmapped_v = boost::multiprecision::tanh(previous_v);
double sigmoid_velocity = sigmapped_v.convert_to<double>();
// expected primary-expression before ‘double’ ^^

double v_probable = abs(sigmoid_velocity);

However explicitly casting it to double (double)sigmapped_v works

(If my hunch is correct, you left out important information from the question. If so, please consider including a /tested/ SSCCE next time) — sehe, Aug 30 '16 at 21:24

score 1 · Answer 1 · edited May 23 '17 at 10:29

Going out on a limb, you are probably in a template context and float_type is a depends on a template argument.

You need to give the compiler type hints here:

double sigmoid_velocity = sigmapped_v.template convert_to<double>();
//                                    ^^

Without the disambiguation hint the compiler will parse < as operator<

See also Where and why do I have to put the "template" and "typename" keywords?

boost multiprecision gmp_float convert_to() gives compilation error

1 Answers1