I suck at searching and finding the things that I want so hopefully this isn't a common question that I just can't find.
I've been pondering over the question of initializing a struct. Is it faster to initialize a struct using a list, by accessing each individual element or is there no discernable difference at all (e.g., it's just a matter of coding style)?
Code:myStruct tmpStruct = {1, 2, 3}; tmpStruct.a = 1; tmpStruct.b = 2; tmpStruct.c = 3;
I see no difference in terms of assembler code. This is a style matter, only. I did not even build an exe.Code:; 8 : struct foo bar = {1, 2, 3}; mov DWORD PTR _bar$[ebp], 1 mov DWORD PTR _bar$[ebp+4], 2 mov DWORD PTR _bar$[ebp+8], 3 ; 9 : struct foo baz; ; 10 : baz.a = 1; mov DWORD PTR _baz$[ebp], 1 ; 11 : baz.b = 2; mov DWORD PTR _baz$[ebp+4], 2 ; 12 : baz.c = 3; mov DWORD PTR _baz$[ebp+8], 3
Edit: woops, but still no difference.
Last edited by whiteflags; 03-11-2009 at 10:59 PM.
No difference in assembler output for gcc 4.2.4 either, with or without optimisations enabled. If you are talking about which is faster to type, I'd say that using an initialiser usually is faster
That said, it is not always a matter of style, e.g., you need it for const objects.
Look up a C++ Reference and learn How To Ask Questions The Smart WayOriginally Posted by Bjarne Stroustrup (2000-10-14)
Great. Thanks for the replies. I don't actually use structs in my own code as I'm sticking with strictly OOP objects in C++ but I'm using several API's that are written in C. I only ask because I know it can be a lot faster to use an initialization list for a class so I wondered if the same benefits applied to C structs.
For me having fewer lines of code makes sense when setting up these structs so I'll stick with that.
Thanks again!
The compiler is free to do pretty much anything. It could even vary from one initialization to another, by the same compiler. Initialization of a large struct could even be accomplished by a call to memcpy().
I think it's unlikely that any decision the compiler makes about this sort of thing is going to significantly impact performance.
Code://try //{ if (a) do { f( b); } while(1); else do { f(!b); } while(1); //}
I didn't think so. I assumed that it would all be pretty much the same and any differences would be negligable.
I'm using the SDL struct SDL_Rect which only has four values. In this case it makes far more sense to have a single line setting all of those values than having five lines to define the struct and set each value. I guess you could call it a case of readability vs. any real improvement in performance.
