Cheat Engine

[Polynomial]

Here's a quick bit of C++ code that I'm trying to optimise. It is being called repeatedly as part of a primality check test, and I'm trying to get this bit to run as fast as possible. I'm actually hoping that one of you will be able to come along and give me some crazy inline ASM that'll do this very quickly.

I'm assuming that combining the n%2 and n%3 lines will save a conditional jump (therefore increasing speed), but will this be negated by the fact that both n%2 and n%3 will need to be calculated every time?

[sven3107]

First thing I would think of is a switch block...

[Polynomial]

A switch block would be completely useless in this case. I'm doing 3 checks that return true/false, not one check that returns many values.

[smartz993]

[Polynomial]

What about multiplication? I can stick all of my checks into one line like this:

[Slugsnack]

no it evaluates them one by one and then if there is a match at any stage then it will jmp to a return 0 piece of code ( lazy evaluation )

[Jani]

Why not to test it yourself if one is faster? Just compile it, call it hella lot times and time the results. Do it N times until your statistical data base is large enough to draw any conclusions.

[Slugsnack]

would be better to test clock cycles instead of time though. since smaller code can be faster when looped even if each loop alone would have been slower. example is if more of the looped code could be cached. whereas realistically that code would be executed once only so the caching would not occur to the extent it does

solution : time by clock cycles

[hcavolsdsadgadsg]

This is the dumbest thing to optimize ever.

You will never have a humanly noticeable difference unless you run them a couple trillion times, if even then.

I checked the dissassembly and you would be hard pressed to write anything better on your own. It optimized most of the divides into multiplications, etc, etc. Mashing everything into one if does nothing besides make shit unreadable, they have to be evaluated anyway.



This is already faster than you would ever need for a real time application... a couple billionth of a second isn't ever going to matter besides wasting your time. Don't get hung up on trivial crap like optimizing IF STATEMENTS.

[Bswap]

Code optimisation can refer to either speed of execution or size of code... which are you after?

Here's a quick effort at optimising for size;

[DoomsDay]

Tell me if I'm wrong:
Pick any random R, and then compute N = R * 6.
N+0: Would fail because {(N+0) % 2 == 0}
N+1: Would process because {(N+1-1) % 6 == 0}
N+2: Would fail because {(N+2) % 2 == 0}
N+3: Would fail because {(N+3) % 3 == 0}
N+4: Would fail because {(N+4) % 2 == 0}
N+5: Would process because {(N+5+1) % 6 == 0}
N+6: == (R+1) * 6

[Polynomial]

I'm optimising for speed.

I'll give these a try and see what happens.

[Deltron Z]

If you need a list of primes you could use Sieve of Eratosthenes and then Binary Search on that list. faster, however takes up more space. it is useful when you need to check a lot of numbers a few times, instead of checking the same number a few times.

But as slovach said, you won't see any difference unless you execute this function millions or of times.