Macro Miscalculation

[SkunkieButt]

So I'm working on a library and I'm writing a lot of cryptographic classes, such as SHA1. I'm using this code for bit rotation:

#ifndef XSPEEDOPT
// Makes external functions
static uint rotater(uint val, int shift) { return ((uint)val >> shift) | ((uint)val << (32 - shift)); }
static uint rotatel(uint val, int shift) { return ((uint)val << shift) | ((uint)val >> (32 - shift)); }
#else
// Inlines a math only function
#define rotater(val, shift) (uint)((uint)val >> shift) | ((uint)val << (32 - shift))
#define rotatel(val, shift) (uint)((uint)val << shift) | ((uint)val >> (32 - shift))
#endif 

XSPEEDOPT is defined if I feel like optimizing speed (like inlining etc). When I do that, the macro does not work 0.o The output hash of, such as a SHA1, is invalid, but if I use the actual functions, it works... lolwut?

[Athar]

What on earth makes you think a macro would be faster than an inlined function?
Ditch the macro - worrying about why it doesn't work is a waste of time.

[SkunkieButt]

To my understanding, a macro inlines that function

I'm working on a cross platform library so i want to use a macro on math functions that want to be inlined otherwise i would have done that

[Disch]

a macro inlines that function

inlined functions also inline the function, but they don't destroy the global namespace in the process.

Avoid macros wherever possible.

Just use the nice inline keyword:

static inline uint rotater(uint val, int shift) { return ((uint)val >> shift) | ((uint)val << (32 - shift)); }
static inline uint rotatel(uint val, int shift) { return ((uint)val << shift) | ((uint)val >> (32 - shift)); }

[firedraco]

A macro is no faster. In fact, it can be slower since it *forces* the compiler to inline the function, even if it would slow down the program. Just use an the inline keyword and let the compiler do the hard work for you.

[SkunkieButt]

I'm working on a cross platform project... windows only can do inlining... so I would like to do a macro where i can cause that can be interpreted on other platforms

And I disabled inline checks, i have it inline everywhere i choose to

[Disch]

windows only can do inlining

Where did you hear that? That's not true at all.

And I disabled inline checks, i have it inline everywhere i choose to

So you think you know more about what the compiler is doing than the compiler, then, eh? ;P

[SkunkieButt]

I've read it at a few places but if that's the case then ok, I will use inline but I just wanna know why a macro wouldn't work anyways D:

[Disch]

Probably because you're compounding statements when you pass them, causing the macro to expand in a way you didn't expect (yet another reason why they're evil)

#define TEST(a,b) a*b

int main()
{
  int foo = 3;

  int test = TEST(3,foo+4);   // you expect test to be 21, right?

  cout << test;  // SURPRISE, it's 13!

  // TEST(3,foo+4) gets expanded to 3*foo+4 which evaluates as (3*foo)+4
  //  not as 3*(foo+4) as you'd expect
}

[SkunkieButt]

@disch
Nope that's not it...

[Athar]

Nope that's not it...

So does that mean you already put an extra pair of braces around the entire term, plus around each occurrence of val?

[warrant]

Nothing wrong.