fix binary search http://googleresearch.blogspot.de/2006/06/extra-extra-...

[kuebler]

Fixes calculating the binary search pivot index calculation for large values, c.f. Extra, Extra - Read All About It: Nearly All Binary Searches and Mergesorts are Broken

[kgcrom]

I compare two operation.

test.o:     file format elf64-x86-64


Disassembly of section .text:

0000000000000000 <main>:
#include <stdio.h>

int main() {
   0:   55                      push   rbp
   1:   48 89 e5                mov    rbp,rsp
        int min = 10, max = 16;
   4:   c7 45 f4 0a 00 00 00    mov    DWORD PTR [rbp-0xc],0xa
   b:   c7 45 f8 10 00 00 00    mov    DWORD PTR [rbp-0x8],0x10
        int mid = 0;
  12:   c7 45 fc 00 00 00 00    mov    DWORD PTR [rbp-0x4],0x0
        mid = (min+max) / 2;
  19:   8b 45 f8                mov    eax,DWORD PTR [rbp-0x8]
  1c:   8b 55 f4                mov    edx,DWORD PTR [rbp-0xc]
  1f:   01 d0                   add    eax,edx
  21:   89 c2                   mov    edx,eax
  23:   c1 ea 1f                shr    edx,0x1f
  26:   01 d0                   add    eax,edx
  28:   d1 f8                   sar    eax,1
  2a:   89 45 fc                mov    DWORD PTR [rbp-0x4],eax
        mid = ((unsigned int)min+(unsigned int)max)>>1;
  2d:   8b 55 f4                mov    edx,DWORD PTR [rbp-0xc]
  30:   8b 45 f8                mov    eax,DWORD PTR [rbp-0x8]
  33:   01 d0                   add    eax,edx
  35:   d1 e8                   shr    eax,1
  37:   89 45 fc                mov    DWORD PTR [rbp-0x4],eax
        return 0;
  3a:   b8 00 00 00 00          mov    eax,0x0
}
  3f:   5d                      pop    rbp
  40:   c3                      ret

I think this has the advantage of reducing the operation.
👍

[mattsta]

Fixed by #1741 and 4aa3300