comparison

[ndun78@gmail.com]

I want to write something like
status_bit = (unsigned)(i - j) >31

as an equivalent code for
if(i < j){
{
status_bit = 1;
}else{
status_bit = 0;
}

This actually works for most of the cases but fails for two values
i = 0x80000000
j = 0x7fffffff

I want a linear code for the above i.e. without jumps, so
1. what could be the issue?
2. what could be the solution?

[ndun78@gmail.com]

Re: comparison

On Oct 21, 3:10 pm, ndu...@gmail.co m wrote:

I want to write something like
status_bit = (unsigned)(i - j) >31
>
as an equivalent code for

signed int i;
signed int j;

if(i < j){
{
status_bit = 1;
>
}else{
status_bit = 0;
}
>
This actually works for most of the cases but fails for two values
i = 0x80000000
j = 0x7fffffff
>
I want a linear code for the above i.e. without jumps, so
1. what could be the issue?
2. what could be the solution?

[Martin Ambuhl]

Re: comparison

ndun78@gmail.co m wrote:

I want to write something like
status_bit = (unsigned)(i - j) >31
>
as an equivalent code for
if(i < j){
{
status_bit = 1;
}else{
status_bit = 0;
}

Try the simple
status_bit = (i < j);

This actually works for most of the cases but fails for two values
i = 0x80000000
j = 0x7fffffff
>
I want a linear code for the above i.e. without jumps, so
1. what could be the issue?
2. what could be the solution?

Is there a reason for avoiding the obvious
status_bit = (i < j);
?

[Charlie Gordon]

Re: comparison

"Martin Ambuhl" <mambuhl@earthl ink.neta écrit dans le message de news:
5o0ngiFkk1juU1@ mid.individual. net...

ndun78@gmail.co m wrote:

>I want to write something like
>status_bit = (unsigned)(i - j) >31
>>
>as an equivalent code for
>if(i < j){
>{
> status_bit = 1;
>}else{
> status_bit = 0;
>}

>
Try the simple
status_bit = (i < j);
>

>This actually works for most of the cases but fails for two values
>i = 0x80000000
>j = 0x7fffffff
>>
>I want a linear code for the above i.e. without jumps, so
>1. what could be the issue?
>2. what could be the solution?

>
Is there a reason for avoiding the obvious
status_bit = (i < j);

In other words: if there is a way to compute this without tests and jumps,
the compiler should know.

--
Chqrlie.

[ndun78@gmail.com]

Re: comparison

Is there a reason for avoiding the obvious
status_bit = (i < j);

>

I think that it is not guaranteed to get only 1 or 0 as a result of i
< j, this may result in 0 or non zero value is it not?

[santosh]

Re: comparison

ndun78@gmail.co m wrote:

>

Is there a reason for avoiding the obvious
status_bit = (i < j);

>>

I think that it is not guaranteed to get only 1 or 0 as a result of i
< j, this may result in 0 or non zero value is it not?

No, the relational operators yield either 1 or 0.

[ndun78@gmail.com]

Re: comparison

No, the relational operators yield either 1 or 0.

Is it guaranteed by ANSI C standard?

[pete]

Re: comparison

ndun78@gmail.co m wrote:

>

Is there a reason for avoiding the obvious
status_bit = (i < j);

I think that it is not guaranteed to get only 1 or 0 as a result of i
< j, this may result in 0 or non zero value is it not?

No.
(i < j) is can only be equal to either one or zero.

--
pete

[santosh]

Re: comparison

ndun78@gmail.co m wrote:

>

>No, the relational operators yield either 1 or 0.

Is it guaranteed by ANSI C standard?

Yes. From n1256.pdf:

6.5.8 Relational operators

....

6 Each of the operators < (less than), (greater than), <= (less
than or equal to), and >= (greater than or equal to) shall yield 1
if the speci?ed relation is true and 0 if it is false.92)
The result has type int.

[Army1987]

Re: comparison

On Sun, 21 Oct 2007 10:10:51 +0000, ndun78 wrote:

I want to write something like
status_bit = (unsigned)(i - j) >31
>
as an equivalent code for
if(i < j){
{
status_bit = 1;
}else{
status_bit = 0;
}
>
This actually works for most of the cases but fails for two values
i = 0x80000000
j = 0x7fffffff
>
I want a linear code for the above i.e. without jumps, so
1. what could be the issue?

If *you* don't know that...

2. what could be the solution?

status_bit = (i < j);
Boolean operators evaluate to 0 if false and 1 if true.

--
Army1987 (Replace "NOSPAM" with "email")
A hamburger is better than nothing.
Nothing is better than eternal happiness.
Therefore, a hamburger is better than eternal happiness.

[abhy]

Re: comparison

On Oct 21, 3:10 pm, ndu...@gmail.co m wrote:

I want to write something like
status_bit = (unsigned)(i - j) >31
>
as an equivalent code for
if(i < j){
{
status_bit = 1;
>
}else{
status_bit = 0;
}
>
This actually works for most of the cases but fails for two values
i = 0x80000000
j = 0x7fffffff
>
I want a linear code for the above i.e. without jumps, so
1. what could be the issue?
2. what could be the solution?

you as well write this as status_bit = i - j 1 ? 0 : 1

[Martin Ambuhl]

Re: comparison

ndun78@gmail.co m wrote:

>>Is there a reason for avoiding the obvious
>> status_bit = (i < j);

I think that it is not guaranteed to get only 1 or 0 as a result of i
< j, this may result in 0 or non zero value is it not?
>

You are wrong. Check your elementary C textbook.

[Charlie Gordon]

Re: comparison

"abhy" <abhijitkrao283 @gmail.coma écrit dans le message de news:
1192982476.6191 73.30800@z24g20 00...legro ups.com...

On Oct 21, 3:10 pm, ndu...@gmail.co m wrote:

>I want to write something like
>status_bit = (unsigned)(i - j) >31
>>
>as an equivalent code for
>if(i < j){
>{
> status_bit = 1;
>>
>}else{
> status_bit = 0;
>}
>>
>This actually works for most of the cases but fails for two values
>i = 0x80000000
>j = 0x7fffffff
>>
>I want a linear code for the above i.e. without jumps, so
>1. what could be the issue?
>2. what could be the solution?

>
you as well write this as status_bit = i - j 1 ? 0 : 1

No, it does not work at all: the expression i - j 1 is not equivalent to
!(i < j) and even the more accurate i - j < 0 does not produce the correct
result if i - j overflows (as the OP correctly pointed out).

If you *really* want to compute the status bit with plain arithmetics then
you can use this:

status_bit = ((unsigned long long)((long long)i - j) >(CHAR_BIT *
sizeof(int) + 1)) & 1;

It should work if sizeof(long long) sizeof(int).
It may even be acceptably efficient on 64 bit architectures.

But it would be crazy to use such a convoluted expression when ``status_bit
= (i < j)'' is vastly more readable and quite unlikely to be a performance
killer.

--
Chqrlie

[Ben Pfaff]

Re: comparison

ndun78@gmail.co m writes:

I want to write something like
status_bit = (unsigned)(i - j) >31
>
as an equivalent code for
if(i < j){

You should buy a copy of the book _Hacker's Delight_ by Henry
S. Warren. It is a compendium of tricks like this. This
particular problem, and related inequalities, takes up a few
pages in chapter 2, for example.
--
Ben Pfaff

http://benpfaff.org

[Charlie Gordon]

Re: comparison

"Ben Pfaff" <blp@cs.stanfor d.edua écrit dans le message de news:
87lk9wm9d2.fsf@ blp.benpfaff.or g...

ndun78@gmail.co m writes:
>

>I want to write something like
>status_bit = (unsigned)(i - j) >31
>>
>as an equivalent code for
>if(i < j){

>
You should buy a copy of the book _Hacker's Delight_ by Henry
S. Warren. It is a compendium of tricks like this. This
particular problem, and related inequalities, takes up a few
pages in chapter 2, for example.

You will find this "Bit Twiddling Hacks" entertaining as well:

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

Good night!

--
Chqrlie.