Would setting a value to true if it's not true be more effecient than checking if the value is true then setting it to true

Question

I know there really won't be a big change but I'm generally curious. Which is faster one or two?

1:

if (foobool)
{
    // FOOBOOL!
}
else
{
    foobool = true;
}

2:

if (foobool)
{
    // FOOBOOL!
}
// else
// {
    foobool = true;
//}

I figure there is more going into making a if else statement than setting a bool to true. However if I don't put the else statement and the bool is already true then it would be setting it twice. But would adding the else statement would also make it slower?

Does this answer your question? [What does a branchless if in C++ look like?](https://stackoverflow.com/questions/6133322/what-does-a-branchless-if-in-c-look-like) — Sergey Kolesnik, Apr 01 '22 at 23:48
@SergeyKolesnik well with return its a different story because everything stop there but in this case after the if statement is done it will still continue — SplatTab, Apr 01 '22 at 23:53
Is this one time or in a loop? Is foobool volatile and does it have to be synchronized between processors? A lot of questions to answer to save very little if anything. Whatever you are saving , I bet it probably makes less than 1-microsecond difference in your application, so it is not something to worry about. Write something that looks clear to you when you look at it in 6 months or 4 years from now. — Tony BenBrahim, Apr 01 '22 at 23:53
@TonyBenBrahim Thank you and the second does seem clearer to me so its a double win It is pretty interesting too because just looking at the code it seems not setting it to true twice would be more efficient — SplatTab, Apr 01 '22 at 23:57
If we're throwing out small changes, `if (std::exchange(foobool, true))` could potentially be clear depending on the code it's in. Depends whether you want `foobool` to be thrown out of the reader's mind sooner or kept in mind after the body. It also depends what exactly that piece of code is trying to say, e.g., "`foobool` is tested here, but all future tests will pass" or "`foobool` is tested and then reset to its default of `true`". — chris, Apr 02 '22 at 00:01
*Also it is in a loop so it adds up* How do you know it "adds up"? After compiling your code at a high level of optimization, you've profiled both versions of the code, obtained detailed performance data, and identified this boolean check as a performance bottleneck? — Andrew Henle, Apr 02 '22 at 00:04
Why not just throw both in a bench and get a “definitive” answer? — Taekahn, Apr 02 '22 at 00:05
@Taekahn Would be nice to see an exaple that would not be completely optimized out in a bench :) — Sergey Kolesnik, Apr 02 '22 at 00:07
New title `Would setting a value to true if it's not true be more effecient than checking if the value is true then setting it to true` makes no sence. You choose between "reassinging it in else branch" and "reassigning regardless of its previous value" — Sergey Kolesnik, Apr 02 '22 at 00:10
https://www.quick-bench.com/q/ImcjS4wOSvThkRce-tcNqqFUBeo Looks like the second one is cheaper — Fatih BAKIR, Apr 02 '22 at 00:27
@FatihBAKIR I set to C++17 and GCC 11.2 and both versions were the same — Sergey Kolesnik, Apr 02 '22 at 00:40
Is there code in any of the branches? If not `foobool = true;` (without the conditional) would clearly be the fastest. If there is code there, that code would be relevant to running any benchmarks because pretty much everything impacts the optimiser. — bitmask, Apr 02 '22 at 01:10

Would setting a value to true if it's not true be more effecient than checking if the value is true then setting it to true

0 Answers0