Can we and how safe is to “signed” to “unsigned” trick to save one comparison in this case?

问题

For example

bool CheckWithinBoundary(int x, int b) {
  return (x>=0) && (x <= b);
}
bool CheckWithinBoundary2(int x, int b) { 
  return static_cast<uint32>(x) <= static_cast<uint32>(b);
}

CheckWithinBoundary2 can save one comparison.

My question is:

1. Can today's compiler optimize code using this? Or how can I make the compiler do this kind of optimization?

2. Is there any danger to use this trick?

回答1:

The answer to 2 is, yes, there is, these two are not the same. It seems that you are silently assuming that b >= 0, too. Consider e.g x == 1 and b == -1, this would give false for the first case and true for the second.

(I switch to C notation, this is easier to me, and since you also seem to be interested in it)

So we have that in fact

static_assert(INT_MAX < UINT_MAX);

bool CheckWithinBoundary(int x, int b) {
  return (b >=0) && (x>=0) && (x <= b);
}
bool CheckWithinBoundary2(unsigned x, unsigned b) { 
  return (b >=0) && (x <= b);
}

if it compiles, are equivalent on all architectures where INT_MAX < UINT_MAX, and then the implicit conversion int --> unsigned does the right thing.

But be careful, you note that I use unsigned and not uint32_t, because you have to be sure to use an unsigned type with the same width. I don't know if there are architectures with 64 bit int, but there your method would fail.

来源：https://stackoverflow.com/questions/31347420/can-we-and-how-safe-is-to-signed-to-unsigned-trick-to-save-one-comparison-in