PThread + Feval

[Surfing]

Hi all,
I'm writing some code in C++, compiling it with mkoctfile from Octave.

I used pthread to obtain a multithread calculations and it works for 
almost everything.
But now I've a problem ...

I want to do simply this:

double func1(double first, double second)
{
    octave_value_list fargs;
    fargs(0) = octave_value(first);
    fargs(1) = octave_value(second);
    return feval(func2,fargs)(0).double_value();
}

If I execute this in a single thread mode, it works... but when I run in 
in multithread mode (so more threads calls that function in the same 
time), it gives me a segmentation fault at the assignment fargs(0) = 
...  after some iterations...

If I try to assign those value to a simple double array it works...but 
obviously then I can't pass that array to feval ....

Any suggestions?

Then...I execute that function lots of time and I know that feval is 
very slow if it's called so many time...(and I noticed it). Is there a 
way to speed up things? The function I have to call is always the same, 
but with different arguments...

Thanks