Check the OpenCL spec and see if your calculation can be substituted by one of its built-in functions.
I used to write
y = x * (1 << n);
for x is float and n is integer. But I just found that there is a function doing this job! And FYI, C <math.h> has this function too!