c - Casting unsigned int to unsigned short int with bit operator -

i cast unsigned int (32bit) unsigned short int (16bit) b in following way:

• if <= 2^16-1 b=a
• if > 2^16-1 b=2^16-1

in other words cast if > of maximum allowed value 16bit set max value.

how can achieved bit operations or other non branching method?

find minimum of 2 integers without branching:

http://graphics.stanford.edu/~seander/bithacks.html#integerminormax

on rare machines branching expensive , no condition move instructions exist, above expression might faster obvious approach, r = (x < y) ? x : y, though involves 2 more instructions. (typically, obvious approach best, though.)

just kick things off, here's brain-dead benchmark. i'm trying 50/50 mix of large , small values "at random":

#include <iostream> #include <stdint.h>  int main() {     uint32_t total = 0;     uint32_t n = 27465;     (int = 0; < 1000*1000*500; ++i) {         n *= 30029; // worst prng in world         uint32_t = n & 0x1ffff;         #ifdef empty             uint16_t b = a; // gives wrong total, of course.         #endif         #ifdef normal             uint16_t b = (a > 0xffff) ? 0xffff : a;         #endif         #ifdef ruslik             uint16_t b = (-(a >> 16) >> 16) | a;         #endif         #ifdef bithack             uint16_t b = ^ ((0xffff ^ a) & -(0xffff < a));         #endif         total += b;     }     std::cout << total << "\n"; } 

on compiler (gcc 4.3.4 on cygwin -o3), normal wins, followed ruslik, bithack, respectively 0.3, 0.5 , 0.9 seconds slower empty loop. benchmark means nothing, haven't checked emitted code see whether compiler's smart enough outwit me somewhere. ruslik's anyway.