I noticed that there is some redundant and useless slowing code in i and k-rate versions
of linrand, trirand,...,etc. 

For example the function:


void iklinear(PRAND *p)
{
    float *out = p->out;
    *out++ = linrand(*p->arg1);
}

would be faster if rewrited in this way:

void iklinear(PRAND *p)
{
    *p->out = linrand(*p->arg1);  
}

In the last case a local variable creation and a pointer increment are avoided mantaining
the same functionality. I already changed the sources of my version in order to fix this
redundance.

-- 
Gabriel Maldonado

http://www.agora.stm.it/G.Maldonado/home2.htm