[pathra]

I'm trying to look into C++'s operators to see if it's possible to create a math function that executes faster than those in the STL, but less accurate, for when high precision isn't necessary (and also to expand my knowledge of some of the ones I haven't used as much, like the bitwise operators). A consideration, though, seems to be that not all operators are equal; I've heard that floating point multiplication can be very slow compared to bitwise operations.

Are there any other C++ operators that could possibly take a very long amount of time to execute, assuming that only int and float data types are used?

Also, does it take a long time to cast an int to a float and vice versa?

And, as a separate note, (for example) if there is a way to perform bitwise operations on floats, or other strange combinations, is it possible that it'll give a different result on a different machine than mine?

[Skydiver]

Good luck, if you are talking about replacing all these functions:
http://en.cppreferen...pp/numeric/math
http://en.cppreferen...numeric/complex
http://en.cppreferen.../numeric/random

then good luck. You've got a big project ahead of you. I'd like to see your performance numbers and too see if the weeks/months that you spend is worth the milliseconds of speed that you will gain.

As for getting different results on different machines, it's really a matter of several factors:
- raw speed of the processor
- processor architecture (ex. RISC vs CISC)
- processor family/generation with the family
- variants of the processor (same processor family, but different cache sizes)
- quality of your compiler
- quality of your runtime library

If you are willing to hold those factors stable to be able to do apples to apples comparison, then it should be the same results.