c compiler problem in 64-bit version.

hold_breal · 04-22-2006, 05:36 PM

can someone try this out on a 64bit machine and tell me if it also goes wrong if you increase the size of the matrix?

#include <stdio.h>
#include <stdlib.h>

int main()
{
int i, j, zeile, spalte;

int **matrix;

printf("wie viele zeilen?: ");
scanf("%d",&zeile);
printf("wie viele spalten?: ");
scanf("%d",&spalte);

matrix = (int **)malloc(zeile*sizeof(int));
if(NULL == matrix)
{
printf("pwnd!\n");
exit(0);
}

for(i=0; i < zeile; i++)
{
matrix[i] = (int *)malloc(spalte*sizeof(int));
if(NULL == matrix[i])
{
printf("pwnd!\n");
exit(0);
}
}

for(i=0; i < zeile; i++)
for(j=0; j < spalte; j++)
matrix[i][j] = i+j;

for(i=0; i < zeile; i++)
{
for(j = 0; j < spalte; j++)
printf("%d ", matrix[i][j]);
printf("\n");
}

for(i =0; i < zeile; i++)
free(matrix[i]);

free(matrix);
return 0;
}

i also compiled it on my 32 bit machine, and it didn't have any troubles, so unless i've done something stupid, there seems to be a problem with malloc on my 64 bit machine.

both machines are running gcc 4.0.2

much obliged...

hold

ataraxia · 04-22-2006, 06:15 PM

Works here on Debian Sid AMD64 with gcc 4.0.3:

Code:

~ % uname -a                                                             7:13PM
Linux firiesen 2.6.16-1-amd64-k8 #2 Thu Apr 20 20:10:46 CEST 2006 x86_64 GNU/Linux

~ % gcc -v                                                               7:13PM
Using built-in specs.
Target: x86_64-linux-gnu
Configured with: ../src/configure -v --enable-languages=c,c++,java,f95,objc,ada,treelang --prefix=/usr --enable-shared --with-system-zlib --libexecdir=/usr/lib --without-included-gettext --enable-threads=posix --enable-nls --program-suffix=-4.0 --enable-__cxa_atexit --enable-clocale=gnu --enable-libstdcxx-debug --enable-java-awt=gtk --enable-gtk-cairo --with-java-home=/usr/lib/jvm/java-1.4.2-gcj-4.0-1.4.2.0/jre --enable-mpfr --disable-werror --enable-checking=release x86_64-linux-gnu
Thread model: posix
gcc version 4.0.3 (Debian 4.0.3-1)

~ % gcc -Wall -o t t.c                                                   7:11PM

~ % ./t                                                                  7:12PM
wie viele zeilen?: 3
wie viele spalten?: 4
0 1 2 3 
1 2 3 4 
2 3 4 5 

~ % ./t                                                                  7:12PM
wie viele zeilen?: 0
wie viele spalten?: 0

~ % ./t                                                                  7:12PM
wie viele zeilen?: r
wie viele spalten?:                                                            

~ % ./t                                                                  7:13PM
wie viele zeilen?: -1
wie viele spalten?: -5
pwnd!

hold_breal · 04-23-2006, 06:07 AM

someone told me the problem which occurs when the number of zeilen(rows?) is increased to above 4.

**matrix is itself an array where every element points to an array. every element has to be large enough to store an address, which on a 64bit system is of course 8-byte long. so the line:

matrix=(int **)malloc(zeile*sizeof(int))

is too small to hold all the addresses. (sizeof(int) is only 4-bytes large). this can be cured by
a/ finding something 8bytes large to go in sizeof(int)
or
b/ writing it explicitly: matrix=(int **)malloc(zeile*8);
or
c/ being crafty and making sure the computer checks how long the pointer has to be, for example:
i/ matrix=(int **)malloc(zeile*sizeof(int*));
or
ii/

int main()
{
...
int *ptr;
...

matrix=(int **)malloc(zeile*sizeof(ptr));
...
}

which would make pretty general code one could use everywhere, i suppose.

thanks for your help

hold

ataraxia · 04-23-2006, 10:17 AM

Ah, I didn't try it for big (or even medium-sized) numbers.

I suggest you use

Code:

matrix=(int **)malloc(zeile*sizeof(void*));

It's general enough that it won't confuse somebody like the int* version might - void* looks more like "size of pointer" than int* to me.

This code reminds me of tensor analysis...

hold_breal · 04-23-2006, 04:45 PM

good idea, ataraxia, though for the utter beginner (which i am) i think my idea of defining another trivial pointer and then taking sizeof(ptr) may be even clearer

don't worry, it's not tensor analysis (yet), though i am gradually working out numerical methods to do lots of interesting things (for example, laws of motion through using small steps in time to find out where something is, or charge distributions by placing charges singularly in a large 3d-matrix).

in case you or someone else has been playing about with it, i'm trying to think up methods to solve for the path of least effect (hamiltonian stuff) numerically, the aim being of course to build it up gradually to the path of most proper time, so i can calculate the course of a space probe on my system, but i'm struggling to think of good algorithms to maximize (minimize) an integral. i think the best method would be to have the program make many passes, where each pass already regards the path as being made up of n-elements, then letting it gradually adjust to fit. maybe i should find a specialist forum for this sort of thing or even *gasp* *shock* buy_a_book!

thanks again