Is not 'long long' of 8 bytes in gcc?

ishandutta2007 · 01-12-2010, 01:54 AM

Both 'long long' and 'double' are 64 bit data types in gcc but
why can't I insert a value greater than 2^31-1 (ie 32bit)
in a long long type variable.

secondly in the following code,

long long v=1000;
while(v++>=0){
printf("\n%lld",a[0]);
}

the program continues to print negative value of v
even after 2^31-1.Why?

Sergei Steshenko · 01-12-2010, 02:12 AM

Quote:

Originally Posted by ishandutta2007

Both 'long long' and 'double' are 64 bit data types in gcc but
why can't I insert a value greater than 2^31-1 (ie 32bit)
in a long long type variable.

secondly in the following code,

long long v=1000;
while(v++>=0){
printf("\n%lld",a[0]);
}

the program continues to print negative value of v
even after 2^31-1.Why?

Where did you describe 'a' ?

Publish your full source code using CODE tags and publish gcc output when you compile your code like

Code:

gcc -Wall -Wextra your_source.c

carbonfiber · 01-12-2010, 03:26 AM

`long long int` was introduced by ISO 9899/1999, pass "-std=c99" to gcc.

ishandutta2007 · 01-12-2010, 08:28 AM

correction:

Both 'long long' and 'double' are 64 bit data types in gcc but
why can't I insert a value greater than 2^31-1 (ie 32bit)
in a long long type variable.

secondly in the following code,

Code:

long long v=1000;
while(v++>=0){
printf("\n%lld",v);
}

the program continues to print negative value of v
even after 2^31-1.Why?

ForzaItalia2006 · 01-12-2010, 09:34 AM

Quote:

Originally Posted by carbonfiber

`long long int` was introduced by ISO 9899/1999, pass "-std=c99" to gcc.

long long int is also supported by the GNU extensions (gnu89 & gnu99) whereof gnu89 is used by default IF no -std= option is given on the command-line. But yes, -std=c99 should be safe

I can't reproduce this error, neither on a 32-bit nor a 64-bit system. Some thoughts:

* maybe you are using an old C library
* maybe you are using an old & buggy compiler

carbonfiber · 01-12-2010, 11:52 AM

Quote:

Originally Posted by ForzaItalia2006

long long int is also supported by the GNU extensions (gnu89 & gnu99)

Thank you for the information. I suppose I really should read the GCC manual one of these days.

I don't really understand what the OP means by "can't insert a value greater than 2^31-1..". How are you attempting to do this and what makes you declare that you failed?

smeezekitty · 01-12-2010, 12:17 PM

BTW when you are trying to get a variable of the exact bit size, use the stdint types instead.

graemef · 01-12-2010, 07:49 PM

Using the following code:

Code:

#include <stdio.h>
int main()
{
   long long v=9223372036854775800LL;
   while(v++>=0)
   {
      printf("\n%lld",v);
   }
   return 0;
}

I get (as I would expect) the following output

Code:

9223372036854775801
9223372036854775802
9223372036854775803
9223372036854775804
9223372036854775805
9223372036854775806
9223372036854775807
-9223372036854775808

The last number is negative because the comparison is done before the increment. If you want no negative number to be displayed then change the condition of your while loop to:

Code:

while(++v>=0)

sundialsvcs · 01-12-2010, 07:57 PM

As a rule of thumb, don't assume that a particular data type is of a particular size. gcc runs on multiple dozens of hardware platforms, and it must make an "appropriate" decision for each.

There are standard data-type definitions available which stipulate that the variable will be of "at least" a certain size and/or numeric capacity. If your application needs to make such assertions, it should use one of those types. But if it does not, then it should leave all such matters to the discretion of the compiler.

graemef · 01-12-2010, 08:20 PM

For GCC the long long int data type is assumed to be at least 64bits in length. I can't talk for other compilers.

graemef · 01-12-2010, 08:24 PM

Quote:

Originally Posted by ishandutta2007

why can't I insert a value greater than 2^31-1 (ie 32bit)
in a long long type variable.

By this did you mean that you are having difficulty assigning a literal greater than 32bits? if so see my code (post 8) and notice the suffix LL which is essential to add numbers greater than 32 bits. Refer to the link I provided (post 10) from the gcc documentation on the long long data type.

Dan04 · 01-12-2010, 09:51 PM

Quote:

Originally Posted by graemef

For GCC the long long int data type is assumed to be at least 64bits in length. I can't talk for other compilers.

The C99 standard requires that "long long" be at least 64 bits.

paulsm4 · 01-12-2010, 11:00 PM

ishandutta2007 -

I don't know if you caught what graemef said, but he probably hit the nail right on the head:

Quote:

By this did you mean that you are having difficulty assigning a literal greater than 32bits? if so see my code (post 8) and notice the suffix LL which is essential to add numbers greater than 32 bits. Refer to the link I provided (post 10) from the gcc documentation on the long long data type.

Every time you use a long long in an expression with an "int", or with a literal that doesn't explicitly use the "LL" suffix, you risk the entire expression being DOWNCAST to 32-bits.

I'm pretty sure that if you review all your code, you'll confirm that's what's happening.

IMHO .. PSM