Question: Codegen: floating point constants vs. symbolic constants

Codegen: floating point constants vs. symbolic constants Let's say in a complex-valued function, constants such as 1/2, 1/4, etc. are involved. I found, upon some experts' advice, the symbolic computations are different. If one expression is too hard to evaluate or simplify, changing the constants back and forth between floating point constants 0.5, 0.25, etc. and symbolic constants 1/2, 1/4 or one way or the other will often help. However, our end goal is code generation and translation into C or Fortran. Will the efficiency and accuracy be different in the translated end results for the floating point constants and symbolic constants cases? I roughly found no difference. Of course I would like to hear your experience and opinions. Thanks a lot!
Please Wait...