exp with doubles anf floats

[riched158]

hi

i am running C++ under the GCC compiler on an embedded application. My iterative program (amongst other things) calls the exp function several hundred times. However i find that when using doubles as arguments to the exp function it runs approx 10% faster than when i use floats. surely floats should make the program run faster?

cheers
richard

[jsmith]

The exp() function takes a double as a parameter and returns a double.
Whether you pass it a float or a double will make no difference on
execution time of the function. The difference lies in that if you pass a
float, the compiler has to generate a code to convert the float to a
double.

[Bazzy]

The exp() function takes a double as a parameter and returns a double.

exp has an overload which takes and returns float instead of double

[helios]

If you need a lot of performance and you can afford the memory, use a lookup table to store precomputed values of exp(). But you need to be careful to make sure that accessing the table actually saves you time compared to just calling exp().

[jsmith]

I see expf() for float.

[Bazzy]

http://www.cplusplus.com/reference/clibrary/cmath/exp/