Question: How can I use hardware floating-point computations throughout the code?

Dear All,

I want all calculations in Maple code to be done with double precision. Using Digits:=18 and evalf greatly affects the calculation speed. But using the evalhf performs calculations much faster. The problem is that the evalhf cannot work with expressions that contain symbolic terms. For example, the evalhf(1/3*x+1/5*exp(-x^2)) generates an error. How can I perform calculations on expression with double-precision floating point without significantly slowing down the calculation speed? Is there a method that can be used at the beginning of the code to use double precision (preferably hardware) in all computation steps without slowing down the computations, unlike software floating-point computations?
Best wishes
 

Digits:=18;
evalf(1/3.);
evalhf(1/3.);

str:=time[real]();
for i from 1 to 100000 do
evalf(sqrt(3)*i):
end do:
time[real]()-str;

str:=time[real]();
for i from 1 to 100000 do
evalhf(sqrt(3)*i):
end do:
time[real]()-str;