using uint64_t with GCC

[Richard Cavell]

uint64_t i64_BitMask = (uint64_t) (1 << 33) - 1 ;
int a = (i64_BitMask >> 32);
printf ("BitMask %x\n", a);

This produces ffffffff. I'm pretty sure, watching the stages in
debugging, that 1 << 33 is resolving to zero, since it's being shifted
off the MSB of a 32 bit integer.

I am in general having problems with uint64_t with GCC. In particular,
I can't send one to printf and get a sensible output. The shift
operators appear to only work with 32-bit integers. Also, if I have
intermediate steps, it always seems to want to truncate the output to
32-bits. Putting in casts to uint64_t sometimes fixes it, but not
always. What do I do?

(using XCode, Apple's stdint.h)

[jacob navia]

Richard Cavell wrote:

uint64_t i64_BitMask = (uint64_t) (1 << 33) - 1 ;

The problem is that the expression
(1 << 33)
is calculated (as it should) in integer mode, i.e. 32 bits.

The result (zero) is *THEN* cast to uint64_t.

Fix:

uint64_t i64_BitMask = (uint64_t) ((uint64_t)1 << 33) - 1 ;

jacob

[Ron Natalie]

jacob navia wrote:
=

uint64_t i64_BitMask = (uint64_t) ((uint64_t)1 << 33) - 1 ;

You have an unneeded cast there. The left-most cast serves no
purpose.

[jacob navia]

Ron Natalie wrote:

jacob navia wrote:
=

uint64_t i64_BitMask = (uint64_t) ((uint64_t)1 << 33) - 1 ;

You have an unneeded cast there. The left-most cast serves no
purpose.

True.

uint64_t i64_BitMask = ((uint64_t)1 << 33) - 1 ;

is better.

Also good is

uint64_t i64_BitMask = (1LL << 33) - 1 ;

jacob

[bachelor]

i don't know why you need shifts, in some cases <bitset> is more
elegant IMHO