I got a series of defines who represent all bits 1 for different types.
#define ALLBITS_8 ( 0xff )
#define ALLBITS_16 ( 0xffff )
#define ALLBITS_32 ( 0xffffffffL )
2 questions
1. In the code these defines are used as signed types too, is that portable and if not what problems could it cause?
2. Wouldnt it be better to define these macros like:
#define ALLBITS_8 ((unsigned char)~0)
#define ALLBITS_16 ((unsigned short)~0)
#define ALLBITS_32 ((unsigned long)~0)
What would happen in your version if sizeof(short) > 16 bits?
the macro would have to be changed to the type that is 16 bits on that platform.Originally Posted by Thantos
I was more thinking about it in terms of that 0xffff is used as a SIGNED type. Could that cause problems on some platforms when converting it to an unsigned type?
that certainly depends on you implementation. a 16 bit integer is a 16 bit integer, signed or unsigned. your maskings will still be the same.
pretend we're working with 4bit integers. not possible without bitfields, bit just pretend....
if we have
Code:__int4 n = 1000b unsigned __int4 m = 1000b __int4 x = 1000b unsigned __int4 y = 1000b then n & m & x & y always == true
now, if we have
Code:__int4 x = 1000 __int4 y = random; and x & y == true we know that y is negative.
but in this case
it's not so much the values you get. it's what the values you get mean.Code:__int4 x = 1000 unsigned __int4 y = random; and x & y == true we know that y is greater than 2^3 - 1. orther wise, if false, y < 2^3 - 1
the portability issues arise when you assume the size of integers. that's why it's best to use __intN when declaring "size critical" integers to ensure they are the size you need them to be across every platform.
Last edited by misplaced; 04-18-2005 at 05:26 AM.
i seem to have GCC 3.3.4
But how do i start it?
I dont have a menu for it or anything.
And thats where your pretty little macros fail. In theory I should be able to take any code and go from platform to platform and compile and run it without modification.
> #define ALLBITS_8 ((unsigned char)~0)
That would depend on whether the user was after just 8 bits, or all the bits in the underlying data type.
Whilst most common desktop machines have 8-bit unsigned char, some embedded DSP processors for example have 32-bit unsigned char.
0xFF and ~0 would then obviously have different answers.
If you dance barefoot on the broken glass of undefined behaviour, you've got to expect the occasional cut.
If at first you don't succeed, try writing your phone number on the exam paper.
