How to find TMax without using shifts

1

Using ONLY

 ! ~ & ^ | +

How can I find out if a 32 bit number is TMax?

TMax is the maximum, two's complement number.

My thoughts so far have been:

int isTMax(int x)
{
  int y = 0;

  x = ~x;
  y = x + x;

  return !y;
}

That is just one of the many things I have unsuccessfully have tried but I just cant think of a property of TMax that would give me TMax back. Like adding tmax to itself would be unique compared to all the other integers.


Here is the actual problem:

/*
 * isTMax - return 1 if x is the maximum, two's complement number,
 *     and 0 return otherwise. 
 *   Legal ops: ! ~ & ^ | +
 *   Max ops: 10
 *   Rating: 1
 */
int isTMax(int x) {
        int y = 0;

  x = ~x;
  y = x + x;

  return !y;
}

int is 32 bits so the max signed would probably be 0x7FFFFFFF

c++
c
binary
logic
bit-manipulation
asked on Stack Overflow Sep 4, 2011 by David • edited Jan 6, 2019 by 陳 力

7 Answers

4

As far as I know, there is no way to determine if a particular value is the max value of a signed type without already knowing the maximum value of that type and making a direct comparison. This is because signed expressions experience undefined behavior on overflow. If there were an answer to your question, it would imply the existence of an answer to a serious unsolved problem that's been floating around on SO for some time: how to programmatically determine the max value for a given signed type.

2
int isTmax(int x) {


    //add one to x if this is Tmax. If this is Tmax, then this number will become Tmin
    //uses Tmin = Tmax + 1
    int plusOne = x + 1;

    //add to x so desired input becomes 0xFFFFFFFF, which is Umax and also -1
    //uses Umax = 2Tmax + 1
    x = x + plusOne;

    plusOne = !(plusOne);

    //is x is 0xffffffff, then this becomes zero when ~ is used
    x = ~x;
    x = x | plusOne;
    x = !x;

    return x;
}
answered on Stack Overflow Jul 5, 2013 by N.W. • edited Jan 30, 2018 by sallaben
2

Spend 3 hours on this problem. I know this problem comes from csapp's data lab and its newest requirement is

1. Integer constants 0 through 255 (0xFF), inclusive. You are
  not allowed to use big constants such as 0xffffffff
....

* isTmax - returns 1 if x is the maximum, two's complement number,
 *     and 0 otherwise
 *   Legal ops: ! ~ & ^ | +
 *   Max ops: 10
 *   Rating: 1

So, shift operator(<</>> and 0x7FFFFFFF from accepted answer is forbidden now)

Below is my way:

TDD-style:

isTmax(2147483647) == isTmax(0b011111111111...1) == 1
isTmax(2147483646) == isTmax(0b011111111111...0) == 0
isTmax(-1) == isTmax(0b111111111...1) == 0
isTmax(-2147483648) == isTmax(0b100000000...0) == 0 

the return should be either 0 or 1. In, c, ! + all nonzero will return 0. So ! is a must, otherwise we cannot guarantee getting 0 for all numbers.

First naive try:

because 0b0111111...1(aka 2147483647) is the only argument which should make isTmax return 1 and 2147483647 + 1 should be 10000000...0(aka -2147483648)

0b011111111...1 xor 0b1000000000...0 is 0b11111111111...111. Because we must use !, what we hope to see is 0(aka 0b0000000000000...0). Obviously, just apply logic not(aka !) to 0b1111111...1), then we will get 0b000000000000):

!(~(x ^ (x + 1))

let's printf it

void print(int x)
{
     printf("%d\n", !(~(x ^ (x + 1))));
}
int main() {
        print (2147483647);
        print(2147483646);
        print(-1);
        print(-2147483648);
}

1

0

1

0

live demo

Not bad, only -1 doesn't work as we expected.

second try:

Let's compare -1 and 2147483647

11111111111111111111111111111111
01111111111111111111111111111111

We can find -1 + 1 = 0 while 2147483647 + 1 = -2147483648. Emphasize again, what we want is diff -1 and 2147483647, because both of them return 1 as above shows. Look back to the protety of logic not in c: all nonzero will return 0, so !-2147483648 == 0 and !(-1 + 1) != 0. Just modify left part of x ^ (x + 1)(x) into x + !(x + 1). If x is 2147483647, x + !(x + 1) will equal to x.

Run again:

void print(int x)
{
    printf("%d\n", !(~( x + !(x + 1) ^ (x + 1))));
}
int main() {
        print (2147483647);
        print(2147483646);
        print(-1);
        print(-2147483648);
}

1

0

0

0

live demo

Done!

answered on Stack Overflow Jan 6, 2019 by 陳 力 • edited Jan 11, 2019 by 陳 力
2

No shifts solution. Realizing to take advantage of the property that !(0b01111 + 1 = 0b10000) = 0 and !(0b11111 + 1 = 0b00000) = 1 is particularly tricky. This problem took me a long time.

/*
 * isTmax - returns 1 if x is the maximum, two's complement number,
 *     and 0 otherwise 
 *   Legal ops: ! ~ & ^ | +
 *   Max ops: 10
 *   Rating: 1
 */

int isTmax(int x) {

  int a = ~((x + 1) ^ x);
  int b = !(x + 1); 
  int c = !(a + b);

  return c;
}
answered on Stack Overflow Dec 15, 2020 by skihops
0

Something like this perhaps? 0x7FFFFFFF is the maximum positive signed 32 bit two's complement number.

int isTMax(int x){
    return !(x ^ 0x7FFFFFFF);
}

I am not sure, you may need to cast it to unsigned for it to work.

answered on Stack Overflow Sep 4, 2011 by Ben Goodrich
0
#include <stdio.h>
#include <stdlib.h>

int test(int n) {
  return !(n & 0x80000000) & !~(n | (n + 1));
}

// or just effectively do a comparison

int test2(int n) {
  return !(n ^ 0x7fffffff);
}

int main(int ac, char **av) {
  printf("is%s TMax\n", test(atoi(av[1])) ? "" : " not");
  return 0;
}
answered on Stack Overflow Sep 4, 2011 by DigitalRoss
0

if it is Tmax : 011111.....

then we xor it with 10000....

we get 11111....

then we ~ to get all 0s = 0 , !0 we get 1:

int isTmax(int x) {
  return !(~((1 << 31) ^ x ));
}
answered on Stack Overflow Sep 30, 2018 by XueYu

User contributions licensed under CC BY-SA 3.0