Inline assembly SSE Instruction with GNU Compiler

[fredwurst]

Hello,

This simple program (below) compiles and runs, but it does not work correctly. The output I get is
> nan 2 3 4
but I should get something like
> 2 3 1 4

Why that ???
(Compiled with GNU Compiler GCC 2.96 and 3.2 ...)

Thanks for help,
Fredwurst



#include <iostream>
using namespace std;

int main(int argc, char* argv[])
{
float vec1[4] = {1,2,3,4};
float vec2[4];

__asm__ __volatile__ (&quot;
movaps %1, %%xmm0;
pshufd $0x78, %%xmm0, %%xmm1;
movaps %%xmm0, %0;
&quot;
: &quot;=m&quot;(vec2[0])
: &quot;m&quot;(vec1[0])
);

cout << vec2[0] << &quot; &quot; << vec2[1] << &quot; &quot; <<
vec2[2] << &quot; &quot; << vec2[3] << endl;

return 0;
}