kbrafford,
Float to int can be executed in a single instruction. Double to int is not 100% supported in SDK 2.1, but the current non-conforming implementation executes it in 2 instructions(dbl->float then float->int). In the upcoming release, double to int will be fully supported and should take ~12-15 instructions for 32bit ints and ~20-23 instructions for 64bit ints.